The Advertising Association and WARC have published their annual Expenditure Report, and the most consequential thing in it is not the £46.7 billion headline figure for 2025 (up 6.4% year-on-year). It is a reclassification that reframes what every previous dataset was measuring.

For the first time, retail media and social media are being reported as standalone channels. They had previously been absorbed into broader categories, making it difficult to track their true scale. Pulling them out changes the comparability of every channel total in the report.

What “Other Online Display” tells you

“Other Online Display” fell 20% in 2025 and is forecast to drop a further 25% in 2026. That category, which includes audio, gaming, video outstream, native advertising, and display embedded formats, now explicitly excludes revenues from social, retail media, TV, radio, and news and magazine brands. Strip those out and what remains is a contracting residual.

The same logic applies to search. Retail media has now been separated from the search total, which means the £17.9 billion attributed to search in 2025 is not directly comparable to previous years’ figures. Search still accounts for the largest share of UK advertising investment at 38.3%, but that number has been quietly recalibrated.

The channels growing quickly

Retail media recorded full-year growth of 17.5% to reach £3.75 billion, and its Q4 performance was particularly sharp, up 30.5% year-on-year during the festive season. That tracks the US trajectory, where retail platforms now compete with traditional media owners for brand budgets, not just performance spend.

Addressable TV grew 37% across the year to £1.84 billion, though total TV revenue declined 1.2% when linear is included. Addressable is taking share inside a shrinking total, buyers are reallocating within TV before they reallocate away from it.

Social media (including YouTube) grew 21% to £11.5 billion, 24.7% of all UK advertising investment, a larger share than TV and published media combined.

Where the money is leaving

Published media (national and regional newspapers and magazines, print and digital) declined 5.1% in 2025, with regional news brands falling 6%. Even online revenues within regional titles dropped 2.2%. National news brands fared slightly better, with online revenues down just 0.9%, but the overall picture for print-heritage publishers is one of structural decline. Direct mail was broadly flat, down 0.3%.

What comes next

The AA and WARC describe this as phase one of a broader update programme. Phase two will address three areas:

• How investment in the influencer and creator channel is captured

• How to categorise spend in the emerging generative AI and LLM advertising channel

• How to distinguish between the investment patterns of large brand advertisers and the estimated 3.5 million UK SMEs that advertise annually.

Stephen Woodford, Chief Executive of the Advertising Association, framed the refresh as a response to a more fragmented market: “This evolution of the AA/WARC Expenditure Report will ensure the industry has the best possible information to guide understanding of investment across the UK advertising and media landscape.”

The LLM channel question is worth watching. At present there is no agreed methodology for measuring what brands spend on AI-native placements sponsored outputs in chatbot responses, model-embedded brand integrations, and similar. Without a standard, AI-native placements stay invisible in benchmarks like this one, even if the underlying spend is already material.

For more information visit www.warc.com/expenditurereport

One number from WARC’s 2026 Consumer Trends report deserves attention before any other.

One in ten consumers globally now report having been in a relationship with an AI chatbot. Of those, 62% say they would turn to that AI rather than a human friend for personal advice.

The figure comes from GWI survey data, global in scope, published this week in WARC’s 2026 Consumer Trends report. At world population scale, one in ten represents hundreds of millions of people managing a meaningful part of their emotional lives through a piece of software.

The marketing industry has spent the past two years debating AI as a production tool: faster copy, cheaper creative, automated personalisation. WARC’s data suggests the more consequential AI story is happening on the consumer side, and it is moving faster than most brand strategists have noticed.

“Consumer sentiment in 2026 remains fragile, shaped by financial pressures, geopolitical turbulence, and a growing need for escape, with spend shifting to small comforts that bring joy. Brands that forge emotional connections will thrive, but must navigate rising scrutiny on technology, authenticity, and trust to make a meaningful impact.”

“Our annual Consumer Trends report provides a broader view of the major issues confronting our industry from the perspective of consumers, with suggestions to help businesses create the most impact in the year ahead.”

Stephanie Siew, Senior Research Executive, WARC

What AI companionship actually means for brands

The dynamic WARC identifies is straightforward: AI companions reduce loneliness while removing the friction of human interaction. No judgement, no reciprocal obligation, available at any hour. For isolated consumers, and there are more of them than most public discourse acknowledges, this is a genuinely useful proposition.

The commercial implications run in two directions. In categories where AI-mediated interaction can demonstrably improve outcomes (elderly care, mental health support, companionship devices), there is real product opportunity. Manufacturers are already moving: Companion devices aimed at older consumers are a growth category precisely because the emotional and safety need is real.

The risk is more structural. Consumers in AI relationships are becoming more trusting of AI generally. That sounds like an opportunity until you consider what happens when commercial intent is layered on top of emotional dependency. WARC is explicit: the combination is not stable, particularly with younger demographics. An AI companion that is also a sales mechanism is a trust problem waiting to surface, and the regulatory environment around AI and minors is tightening in most major markets.

The other four signals

The AI companionship finding sits alongside four other trends in the WARC report.

Comfort consumption. Consumer sentiment is fragile but not broken. Nearly half of employed consumers, 45%, report job security concerns. A third are saving more or cutting back. The response is selective rather than paralysed: spending redirected toward small, reliable sources of joy. Wellness spending, hobbies, and accessible pleasures. For brands, the opportunity is in credibly occupying that territory, which requires positioning precision rather than a discount strategy.

Youth social media restrictions. 64% of consumers believe social media is harmful to children. Half would support age verification. Legislation is moving in the same direction across multiple markets. The commercial consequence is a disrupted acquisition funnel for Gen Alpha (children born from roughly 2010 onward), a cohort that has done much of its product discovery on social platforms. Brands without owned channels and direct relationships are more exposed than they may realise.

China’s brand shift. 36% of consumers now describe Chinese apps and tech as innovative. One in four globally prefer to buy personal electronics from Chinese brands, behind only the US and Japan. The “Made in China” association has moved from commodity to credibility across a substantial portion of the global consumer base. Western brands competing in consumer electronics and lifestyle tech are facing a different competitive landscape than they were five years ago.

The AI content backlash. 85% of consumers say knowing something was made by a human makes it more meaningful. Seventy-eight percent demand clear labelling of AI-generated content, which has crossed from niche preference into baseline expectation. For publishers and media brands, editorial credibility built on human judgement is more commercially valuable than it has been in years. The industry’s appetite for AI as a cost-reduction tool is running directly against its audience’s rising demand for the thing AI cannot reliably produce.

Trust connects all five trends: who holds it, and who is losing it. Brands that understand what consumers are actually anxious about, and use technology with those anxieties rather than against them, are better positioned than the headline AI adoption numbers suggest.

Part of WARC’s Evolution of Marketing programme, WARC subscribers can read the 2026 Consumer Trends report in full here.

Print isn’t dead. It’s just become a niche.

Print has been declared dead so many times that the headlines barely register anymore. Dave Kirwan has heard them all. As Production Director at The Guardian, he’s closer to reality than most.

“Print, especially quality newspapers, delivers to a very specific demographic and a very high-value demographic. It commands trust in a way that many other media do not. In these times, with everything becoming more difficult politically and globally, a trusted print brand is still filling a valuable niche,” says Dave Kirwan.

That trust is what has kept quality newspapers off the critical list while other print titles have buckled. The Guardian and others serve audiences that actively choose them. Advertisers — watch brands, food retailers, quality consumer goods — still find value in that association. The decline is real, Kirwan says, but the pace has slowed. The brutal early shifts, the collapse of classified advertising, and the disappearance of recruitment pages are largely done. What remains is more stable.

“There’s been a decline in print advertising, but the rate of decline has slowed. I think a lot of the major shifts have already happened. You’re not going to see much difference over the next five years; it will gradually reduce, but still deliver a good contribution,”

Dave Kirwan, Production Director, The Guardian

A recent collaboration with Tesco for a food magazine supplement shows how the Guardian is pushing print harder and more creatively. A three-quarter cover format blended advertising with editorial, trying, as Kirwan puts it, to make the best of both worlds. “It’s a long tail. It might be a niche now rather than the mainstream, but it’s definitely still there.”

Before the tools: chasing copy by hand

Behind the brand sits a production operation that, until recently, ran on systems Kirwan describes as “almost ancient.” Ad copy was tracked manually. Files were pushed from one workflow stage to the next by hand. Chasing advertisers for artwork meant emails, phone calls, and a lot of repetition.

Papermule’s AdDesk changed that. The platform automates the copy-chasing process — from initial contact with advertisers through to preflight checking and approval. Kirwan’s team now manages a process that previously ate significant resources with a fraction of the effort.

Mike Hoy, Papermule’s Managing Director, describes AdDesk as a content and communications hub, one that removes the human touch-points from a process that has historically depended on them.

“Have we identified the right person? Do they know what they’re sending? Is what they’re expecting to send going to match what we’re expecting to receive? Do I know who to call at five o’clock if I haven’t got the content? That approval process should be one click. If it’s anything more than that, then we’ve done our job wrong,”

Mike Hoy, Managing Director, Papermule

The contrast with generic project management tools, such as Monday.com and Asana, is stark, Hoy argues. Those platforms track tasks. They don’t track outcomes.

“If I’ve downloaded an email, saved a PDF, put it in a folder and then gone to Monday.com and marked the job as done, that relies on both hands coordinating. If I’ve taken a phone call or gone for a coffee, that process is broken. It involves humans, and it takes time. A thousand moments make up hours,” says Mike Hoy.

Headcount Down. Output the Same.

Kirwan won’t be drawn on specifics. But he doesn’t dodge the core point.

“The tangible improvement is costs. We’ve reduced headcount and achieved significant savings without affecting workflow or service to the business.”

For a lean production operation, this matters, especially as energy and paper costs rise and global instability increases scrutiny of operational expenses.

“All other costs are going up. The costs are always under scrutiny, especially right now. There's a high chance of price increases across the supply chain.”

The shift to AdDesk has moved his team toward a more monitoring-focused role. The dashboard provides a live view of where every piece of copy sits. Sales and editorial both have access. The manual intervention that once characterised the process has largely been replaced by exception management.

PlanDesk: the flat plan that shares itself

The second product in the Papermule suite — PlanDesk — handles flat planning and sits alongside AdDesk. Before its implementation at the Guardian, sharing a flat plan with editorial or sales meant generating a PDF and emailing it. Now it’s a live link.

“We share live flat plans now, which is something we couldn’t do before. You don’t need a bespoke application on a Windows-only computer. I keep telling editorial: you don’t need us to send it to you, you can just go to this link and see the live flat plan for yourself. We’re moving them towards that. Some departments use it more than others,” says Dave Kirwin.

Old habits die hard. Getting editorial teams to pull information rather than wait for it to be pushed to them is a work in progress. But the infrastructure is there.

Hoy describes PlanDesk as a proper enterprise-level, web-based flat planning solution, built to replace legacy platforms that were never designed for the current technology environment. The pitch to publishers is as much about IT overheads as it is about usability.

“You’ve currently got probably two or three people who know about your planning system, with integrations and platform dependencies. We’re saying: get rid of those platforms, free up those skills, host it in the cloud. It’s one environment. You’re not distributing content across multiple platforms. It’s just one place,” says Mike Hoy.

The two products are engineered to work together, sharing identifiers, order references, and content IDs from the sales platform through to editorial. Hoy’s view is that publishers who start with one will quickly want the other.

The ISO certification that came from every RFP

Papermule recently achieved ISO 14001 environmental certification. For a publishing software company, it’s an unusual credential. Hoy’s explanation is straightforward.

“Every RFP I’ve done in the last five years has had an environmental questionnaire. You reach a section that asks, “Are you ISO 14001 accredited?” If no, continue with the questions. If yes, move to page fifty. So you’re essentially doing an ISO accreditation with every single RFP — and it’s horrendous,” says Mike Hoy.

The certification formalises what Papermule was already doing. The business had always operated with environmental considerations in mind. What changed is the procurement friction.

“We want to be easy to do business with. Being able to just say yes, we’ve done it, let’s move on, just helps streamline that new customer experience.”

Publishers are increasingly asking these questions of their suppliers. Kirwan confirms it’s now a standard part of the Guardian’s vendor relationship evaluation. For Papermule, having the certification removes a conversation that, in Hoy’s words, was simply getting in the way.

The risks Kirwan can’t plan for

The technology challenge at the Guardian is largely solved. The workflow is automated, the team is leaner, and the systems are modern. What keeps Kirwan up at night is different.

His team is smaller, which means everyone needs to cover more ground. He is working with Papermule on callout and preflight processes that enable people from a flat-planning background to handle copy management work they couldn’t previously handle.

“We can’t justify specialists in every area now. We need generalists who can cover the entire advertising production process. That’s the challenge the team is stepping up to.”

Then there are the things nobody can plan for. Kirwan points to the 2008 financial crisis, the kind of external shock that overnight restructures the economics of the business. The current geopolitical situation, he suggests, carries the same potential.

“We have to be aware that things like the conflict in the Middle East show you that external factors — economic, political — can hit us in ways we have no control over. Unknown unknowns are a definite worry. You just don’t know where you’re going to be hit.”

On artificial intelligence, Kirwan draws a sharp line. He is open to automating production workflows — copy chasing, image tonal correction, and preflight processes. He will not touch generative AI anywhere near published content.

“This is very much a hesitant area for me. Not generative AI to create anything, and indeed, The Guardian has strict guidelines around the use of AI. Image enhancement in terms of tonal correction, yes. But not creating elements within an image. I want to keep an eye on that and see where the benefits might be.”

The workflow is sorted. The team is smaller and more adaptable. On AI, he’s watching. On everything else, he’s waiting.

The Supermetrics' 2026 Marketing Data Report surveyed 435 marketers globally and found something the industry would rather not say out loud: the pressure to adopt AI is overwhelming, but the results are not.

Eight in ten marketers report feeling under pressure to introduce AI into their workflows. Just 6% say they’ve actually managed it. That gap, between mandate and reality, is the report’s central finding, and it points to a problem that new tooling alone won’t resolve.

The pressure is coming from the top. According to the report, 89% of the pressure for AI adoption originates with C-suite executives and boards. The logic is understandable: AI is moving fast, competitors are experimenting, and leadership wants results. But the survey data, collected across brands and agencies in the US, UK, Germany, Australia and Singapore, tells a more complicated story about what’s actually blocking progress.

It isn’t reluctance. Most marketers believe AI could help them, 87% say better data and analytics would improve their team’s effectiveness. The real problem is that the data infrastructure required to make AI useful simply isn’t in place. Over a third of respondents report not having access to high-quality, accessible data. A similar proportion cites a lack of system integration between analytics tools and activation platforms. More than half say they wait 1 to 3 business days for answers from their data teams, and 73% are unhappy with the frequency of that support.

Zach Bricker, North American Solutions Engineering Lead at Supermetrics, is direct about it in the report: “teams are being told to use AI without the support to make it effective. When clear goals are absent, adoption becomes performative rather than functional.”

The ROI problem that won’t go away

Proving return on investment remains the hardest thing marketing teams are asked to do. 45% named it their number one challenge, yet 61% are required to demonstrate it to justify the budget. The irony is that ROI, still the top KPI cited by 45% of respondents, is not actually provable in any precise sense. Attribution modelling is too imperfect, human behaviour too unpredictable, and the variables too numerous for a clean number to emerge.

Linehan’s argument in the report is that budget pressure on marketing is rarely about a business’s financial health; “it’s driven by a deficit of trust in marketing’s ability to demonstrate commercial value.”

Measurement priorities are moving. A/B testing remains the most common approach at 69%, while marketing mix modelling — long considered too resource-intensive for all but the largest advertisers — has become the top investment priority for the next twelve months, cited by 40% of respondents. Agencies are ahead here: 59% already use MMM, compared to 35% of brand-side teams. Incrementality testing, last year’s rising priority, has retreated — perhaps a sign that teams overestimated their readiness to implement it.

The shift toward MMM reflects something real: marketers are trying to construct a more defensible picture of what’s working, one that can survive a budget meeting. Whether they have the data foundation to make it work is a different question.

Who actually owns the data

One structural problem runs through almost every finding in the report. In most organisations, data strategy is controlled by people outside marketing. Over half of respondents said that data decisions, what gets collected, how it gets measured, and what it means, are made by external data or IT teams. Only 31% said the CMO plays a meaningful role.

This matters because the data collected reflects the priorities of whoever controls the collection. When IT or central data teams lead, the result is data that’s efficient to gather and difficult for marketing to act on. As Linehan argues in the report, without marketing involvement in data decisions, you end up with data that’s comprehensive in volume but weak in relevance.

The report draws a useful distinction between four stages of the data workflow: connecting data, managing it, analysing it, and activating it. The argument is that marketing should own activation outright, the translation of insight into campaign decisions, and should be meaningfully involved in connection and analysis. Data management is the responsibility of the engineering and data teams. The problem is that most marketing functions have ceded too much ground across the board, leaving themselves dependent on slow-moving external teams for answers they need quickly.

Too much data, not enough time

45% of respondents say they rarely have enough time to properly analyse what they’re looking at. A third check reports monthly or less. Supermetrics’ own platform data shows a 9% increase in data usage between 2024 and 2025, following a 230% increase in the prior period. Volume isn’t producing proportionate insight.

The companies that derive genuine value from AI in the report share a few characteristics: clear use cases, fixed data foundations, and established habits of acting on data before any AI layer was introduced. The sequence matters. Most organisations are attempting it in reverse.

You can download the Supermetrics 2026 Marketing Report HERE

With the UK local elections and the Welsh Government and Scottish Parliament going to the polls on 7 May, the advertising industry has launched an updated version of its political advertising literacy guide aimed at 16–24-year-olds. The campaign is a collaboration between Media Smart, the industry’s long-running education programme, and the Advertising Association.

The timing reflects a specific tension that has been building since the 2024 General Election, when the guide was first produced. Young voters are not simply uninformed about political advertising — they are actively suspicious of it. Research from Next-Gen Media’s Youth Panel, surveying more than 2,000 UK-based 16–24-year-olds in April 2026, found that 58% do not trust the political advertising they see. That figure alone is a problem for any campaign strategist. But the more striking number is this: 84% of respondents said they were worried about being manipulated by AI-generated imagery in political advertising.

That concern is not irrational. The rules governing political advertising in the UK differ from those governing commercial advertising. The Advertising Standards Authority, which regulates commercial ads, has no remit over election campaign material. Political parties are instead subject to electoral law, administered by the Electoral Commission, and to platform-specific policies that vary considerably. Stephen Woodford, CEO of the Advertising Association, makes the point directly: “Election advertising is not regulated by the ASA, unlike commercial ads.” For a generation that has grown up with a degree of consumer protection baked into the digital environments they use, that gap is not immediately obvious, and the current media environment makes it harder to spot.

What the guide covers

The 10-point guide, updated for 2026, explains how political advertising operates across different digital platforms, what transparency obligations exist, and how to evaluate what you see.

It includes specific guidance on AI-generated content and on imprints, the legal requirement for political material to identify who produced and paid for it. The imprint issue is not trivial: 46% of those surveyed said they did not know what an imprint is, meaning nearly half of the target audience cannot perform even the most basic authenticity check on the political content in their feeds.

“Political content is increasingly shaped by AI and digital tools, which can make it harder for young people to tell what is real and who is behind the political advertising they see.” “By updating this campaign, we want to help young voters better understand the rules that apply to election advertising and encourage the use of critical thinking skills when looking at political adverts.”

Rachel Barber-Mack, Executive Director of Media Smart

The campaign material was developed by Livity, a youth-focused creative agency that worked on the original 2024 version. This edition also includes expanded guidance for Scotland, reflecting the multi-jurisdictional nature of the May elections, involving three separate electoral processes running simultaneously, which creates a particular challenge for any educational resource trying to be accurate across the board.

Taking the message off-platform

The campaign’s distribution mechanism is worth noting. Rather than relying entirely on social media, itself a contested, algorithmically unpredictable environment for public information campaigns, the campaign has partnered with Next-Gen Media to run six creative executions across a network of digital out-of-home screens at universities and colleges. The network claims a daily reach of around 250,000 young people. Guy Thurlow, Co-Founder of Next-Gen Media, describes the rationale: “Through our digital out-of-home network, we can bring these important campaign messages directly into trusted, real-world environments like universities and colleges.”

The choice of physical environments as the delivery vehicle for a digital literacy message has a certain logic. Universities and colleges carry institutional credibility that social platforms do not, and the audiences in those spaces are, in many cases, engaging with an election for the first time.

The Advertising Association’s involvement reflects a broader industry position that has been developing over the past few years. The trade body has consistently argued that advertising self-regulation — the system it helped establish more than 50 years ago — works precisely because it builds audience trust. Political advertising sits outside that system by design, which creates a structural tension: the same industry that promotes responsible commercial communication has no formal mechanism to apply those standards to political content. Supporting media literacy education is, in part, how the sector responds to that gap without overstepping into territory that belongs to electoral regulators.

For Media Smart, the campaign extends the work the programme has been doing since it was incorporated into the Advertising Association in 2023. The organisation’s resources — which cover everything from influencer marketing to scam advertising — have been downloaded nearly 200,000 times over the past decade, and the programme claims to have reached 12 million young people. It is funded by a cross-industry coalition of advertisers, agencies, media owners, trade organisations and technology companies.

The political advertising guide is available to download via Media Smart’s website. The out-of-home campaign is now running across the Next-Gen Media network, ahead of the 7 May polling dates.

The underlying numbers from the Youth Panel survey speak to a broader erosion of trust in information environments that the advertising and media industries will need to keep engaging with well beyond this election cycle. A generation that enters voting age already sceptical of the political content it encounters — and specifically worried about AI manipulation — is not a problem that a single educational guide can resolve. But the industry’s capacity to reach young people at scale, with resources grounded in actual platform mechanics rather than generalised media warnings, is a more substantive intervention than the format suggests.

Marketing’s reputation as an AI-forward function is not bad. Adoption rates across the sector outpace most other departments. But the tools being used are not the same as the workflows being changed, and the data is already reflecting that distinction.

McKinsey’s research into European marketing organisations found that 94% have not moved beyond the pilot phase in their generative AI maturity. Alongside that sits a finding that should give AI vendors pause: generative AI ranked 17th out of 20 on the average CMO’s priority list. The gap between how much the industry talks about AI and how senior marketers actually prioritise it is considerable, and the pilot phase figure helps explain why.

Pilots, by design, are bounded. They run in controlled conditions, with defined success criteria, a small team, and an exit point. What most marketing organisations have not worked out is what comes after a successful pilot, how to take something that worked in a limited context and make it a reliable part of how work actually gets done. That transition requires process design, training, governance, and someone with clear accountability for the outcome. Most marketing teams have none of those things in place, which is why 94% are still running variations of the same pilot rather than building on it.

More tools, same problems

Zilvinas Girenas, head of product at nexos.ai — a Lithuanian-founded AI platform backed by Index Ventures — describes what fragmented adoption looks like in practice. “In larger companies, there is typically a dedicated team handling operations, analytics, or martech to support the adoption of AI,” he says. “In SMBs, that responsibility often falls on one marketing lead’s shoulders. This person is left to piece together separate tools with no clear system behind them.”

The pattern is recognisable to anyone who works with mid-sized marketing teams. Copy drafted in one AI tool, visuals in another, performance data pulled from a third, manually assembled for a leadership update. Each individual step may be faster. The process as a whole has not become more manageable, and in some cases it has become less so, because the number of tools requiring oversight has increased without a corresponding increase in the capacity to oversee them.

For enterprise marketing teams, that problem is at least partially absorbed by dedicated operations and analytics functions. For SMBs, it lands directly on whoever is running the day-to-day. The result is not failed adoption so much as adoption that never quite coheres into something that can be measured, repeated, or improved.

The ROI problem

HubSpot’s research found that 52% of marketers say AI has made their content less effective. For a capability whose primary pitch in the content space is volume and efficiency, that outcome warrants examination rather than dismissal.

The most plausible explanation is not that the tools are failing, but that volume without process produces content that is harder to evaluate and harder to stand behind. When a team is producing more material across more channels using more tools, and no one has a complete picture of what was made or why, quality control becomes an editorial judgment call made under pressure rather than a systematic check. The output reflects that.

Girenas puts the problem on visibility rather than capability. “Teams are producing more content than ever, but when you ask how that actually affected the numbers last quarter, no one can clearly identify the process that made a difference. When each team member uses AI in their own way, through different tools and without a cohesive workflow, it creates confusion.”

There is also a brand dimension that tends to get less attention in coverage focused on productivity metrics. Gartner found that half of US consumers would prefer to do business with brands that do not use generative AI in consumer-facing content. Consumer attitudes toward AI-generated material are still forming, and that preference may not persist as familiarity increases. But it represents a meaningful portion of the market right now — and for brands publishing at volume without a shared review process or clear brand guidelines, the risk of erosion is real and not easily reversed once it takes hold.

Who owns the process?

What nexos.ai is selling, a unified platform that brings model access, agent automation, and governance together, is a direct commercial response to the fragmentation problem. Their interest in that narrative is transparent. But the underlying issue does not depend on a vendor to substantiate it. Fragmented AI tooling across marketing functions is a documented pattern, and the consequences difficulty attributing results, inconsistent brand output, limited accountability follow from fragmentation regardless of which solution a team reaches for.

The more durable problem is ownership. Most marketing teams have someone responsible for the content calendar. Very few have someone responsible for how content moves through the process — which tool produced which asset, where approval sat, what changed between draft and publication, and whether any of it connected to a measurable outcome. In a manual workflow, that gap rarely becomes critical because the pace of output keeps it manageable. In an AI-augmented one, where output volume increases and steps multiply, the absence of process ownership surfaces quickly and compounds.

That points to something the pilot-phase data reflects but does not state directly. The organisations stuck at pilot are not necessarily stuck because the technology disappointed them. Many are stuck because moving beyond a pilot requires organisational decisions about ownership, accountability, and infrastructure investment, which are harder and slower than buying a new tool. Until those decisions get made, the pilot continues, the tools proliferate, and the gap between what marketing teams are doing with AI and what they can demonstrate from it keeps widening.

WARC published its Q1 2026 global ad forecast on Thursday. The headline figure — 10.4% global ad market growth to $1.32trn this year — looks solid. Then you get to the scenarios.

Up to $94bn at risk

A prolonged Middle East conflict could strip $49.9bn from ad market growth in 2026 and $93.9bn across the next two years. WARC’s worst case, Strait of Hormuz closed, oil at $150 a barrel, is modelled on the 1973 oil crisis: stagflation, aggressive monetary tightening, consumer confidence in free fall.

The most exposed categories are the ones publishers depend on. Travel and transport ad spend is already in negative territory in WARC’s baseline, down 3.5% as airlines and tourism operators hold budgets back. Food, technology and leisure either flatline or fall in the severe scenario. Brands under margin pressure cut media first. Publishers feel that early.

Reddit’s two-sided trade

While the ad market faces a potential $94bn hole, Reddit is building a business on the same forces that are pulling traffic away from publishers.

Reddit’s ad revenue is forecast to double from $2.1bn in 2025 to $4.1bn by 2027. WARC points to two drivers: the platform’s role in product research and the use of Reddit conversations to train large language models. The AI systems that redirect audiences away from publishers are partly built on Reddit’s content. Reddit gets paid on both sides of that.

The tech category alone accounts for more than a quarter of Reddit’s projected worldwide ad revenue by 2027. Technology advertisers valued Reddit because consumers use it to research purchases — product threads, comparison discussions, real user experience. That behaviour made Reddit’s data commercially useful to advertisers long before it became valuable to AI developers. When the LLM gold rush started, Reddit already had exactly what the market needed.

Publishers had archives AI companies wanted. Reddit had conversations AI companies needed. The difference is that Reddit’s content lived inside a platform with the commercial infrastructure to licence it. Publishers’ content sat behind editorial systems built for distribution, not negotiation. By the time the AI licensing conversation started, Reddit had signed deals. Most publishers were still deciding whether to object.

Publishers built archives, expertise and journalism. AI companies took it. Reddit built the communities AI companies needed and charged accordingly. That difference is showing up in the numbers, and in Reddit’s UK ad spend figures, which are forecast to rise 86.9% this year as the platform accelerates internationally.

Where the brand safety money goes

X returned to ad revenue growth in 2025, up 1.6% to $1.8bn, but remains less than half its 2021 peak. Brand safety concerns persist, sharpened by the platform’s use of its Grok AI to alter images of individuals without consent. Advertisers wanting brand-safe alternatives should, in theory, be heading toward quality publishers. There’s limited evidence they are.

Meta is the more likely destination. Instagram ad revenue hits $101.6bn this year, up 26.9%. Facebook adds $137.8bn, up 19.2%.

The Gulf crisis scenarios affect everyone, but platforms with deep advertiser relationships and sophisticated targeting absorb downturns better than publishers still rebuilding direct commercial relationships after the death of third-party cookies.

Squeezed from both ends

WARC’s James McDonald describes even a contained oil shock as a consumer tax, prices up, spending power down, brands cutting media to protect margins. A prolonged crisis tips into stagflation. Travel, automotive, food and consumer electronics, categories publishers have spent years cultivating — take the direct hits.

Publishers are dealing with an AI audience problem and a potential advertising recession simultaneously. Reddit’s numbers show what it looks like when a platform ends up on the right side of that transition. Most publishers are still waiting for someone else to answer that question.

A Los Angeles jury found Meta and Google liable for $6 million in damages on Wednesday. The numbers barely register for companies worth trillions. The legal theory behind the verdict is another matter.

This wasn’t a case about what people saw on social media. It was about how the platforms were built. The jury concluded that Instagram and YouTube were designed as defective products, deliberately engineered to exploit the developing brains of young users. That distinction matters more than the damages figure. And it matters particularly for an industry that spent the last decade building products using exactly the same methods.

The publishing industry has a problem with this verdict. Not a legal problem, at least not yet. A credibility problem.

The architecture of addiction

The case centred on a 20-year-old California woman known as KGM who began using YouTube at six and Instagram at nine. Her lawyers didn’t argue that the content she saw was harmful; that path would have run into Section 230 protections that shield platforms from liability for user-generated content. Instead, they targeted the design: infinite scroll, algorithmic recommendation engines, notification systems calibrated to maximise re-engagement, interfaces deliberately tuned to keep users returning regardless of whether that was good for them.

Internal Meta documents shown to the jury were explicit about the intent. One described the strategy for attracting children as early as possible:

“If we wanna win big with teens, we must bring them in as tweens.”

Another document showed that 11-year-olds were four times as likely to return to Instagram as to competing apps. The executives who wrote those documents knew precisely what they were building.

The jury’s conclusion: platforms that deliberately engineer compulsive use, particularly in minors, can be treated as defective consumer products under product liability law. Not media companies protected by editorial immunity. Products are subject to the same legal framework as a faulty appliance.

This was the first verdict of its kind. It will not be the last. There are over 2,000 pending lawsuits in the broader wave of litigation. A federal case involving multiple platforms is scheduled for Oakland in June. Another LA trial follows in July, this time involving Instagram, YouTube, TikTok and Snapchat. Legal commentators have reached for the Big Tobacco comparison: the industry laughed at those cases, too, for a while.

If those design choices constitute defective product design when Meta implements them, what are they when publishers implement them?

Where publishers sit in this

The publishing industry’s relationship with engagement design is complicated. And almost entirely unexamined.

For fifteen years, publishers watched as Meta and Google captured advertising revenue and redistributed their audiences. The collective grievance is real. But the response to that disruption, namely competing for attention on the same terms, brought publishers deep into the same design playbook.

Infinite scroll arrived on publisher sites. Push notifications proliferated. Recommendation carousels appeared below articles, optimised for clicks rather than editorial relevance. On-site behaviour tracking fed algorithmic systems designed to surface whatever kept people reading, regardless of whether it was what they had come for. Mobile apps introduced the same compulsive loop mechanics that are now at the centre of litigation.

None of this was accidental or obscure. It was the subject of product roadmaps, conference presentations and trade press coverage. ‘Time spent’ replaced page views as the core engagement metric. Publishers hired product teams from social platforms. The tools were different, but the underlying logic was the same: keep people on the platform as long as possible, reduce friction at every potential exit point, and use behavioural data to predict and pre-empt the moment when attention might drift.

The question Wednesday’s verdict raises is straightforward. If those design choices constitute a defective product when Meta implements them, what are they when publishers implement them?

The distinction publishers will reach for

There’s an obvious counter-argument, and publishers will make it. Their audiences are predominantly adults. The KGM case was specifically about minors; the verdict hinged on evidence that Meta and Google knew they were targeting children and engineering their products to exploit developmental vulnerabilities. Publishers serving professional or general adult audiences aren’t building products for six-year-olds.

That holds, for now. It doesn’t fully close the argument.

The design theory at the centre of this case doesn’t ultimately depend on the user’s age. It depends on whether a product was deliberately designed to override users’ own judgment about how much time to spend with it — to substitute the platform’s commercial interest for the user’s actual preferences. The argument is about autonomy and informed consent as much as it is about children. Age-gating creates a legal distinction. It doesn’t create a moral one.

There’s also the question of what happens outside the US. The UK’s Online Safety Act and the EU’s Digital Services Act are both working through their implementation phases. Neither is focused exclusively on social media. Both are concerned with the design of digital services and their effects on users. A verdict that establishes design-as-defect in US courts gives regulators in both jurisdictions additional language and framing to work with. UK publishers in particular should be paying attention to how that argument develops.

The credibility question

Publishers spent years arguing that what Meta and Google built was fundamentally different from journalism — that the aggregation of attention through algorithmic feeds was corrosive to public discourse in ways that editorial media was not. That argument has genuine force. Editorial judgment, professional standards, accountability to sources and readers: these aren’t nothing.

But publishers undermined their own case every time they bolted social-platform mechanics onto editorial products. The recommendation engines are optimised for engagement rather than editorial value. The notification cadences are engineered to create habitual daily check-ins. The product decisions prioritised session length over whether the session was useful to the reader.

If the distinction between journalism and social media rests on something real, on genuine concern for what is good for the audience rather than what maximises engagement time, then publishers need to be able to demonstrate that in how their products are built. Not just in what they publish.

The Meta verdict is, in financial terms, a rounding error. As a statement about how the law is beginning to think about designed compulsion, it is significant. The 2,000 cases in the pipeline are about Meta and Google. But the legal theory that produced Wednesday’s verdict that deliberately engineering compulsive use is a product defect, not a feature, does not have natural limits that stop at the edge of social media.

Publishing has been pointing at that boundary for years. It’s time to check whether its products actually sit on the right side of it.

Steve Grove, CEO of The Minnesota Star Tribune spoke at a WAN-IFRA Innovate Local webinar hosted by Cecilia Campbell, programme editor, Malmö, Sweden.

In December last year, US Immigration and Customs Enforcement launched Operation Metro Search in Minneapolis. Within weeks, Minnesota became the most-watched place in America. The Minnesota Star Tribune was already there.

“We realised pretty quickly it’s a very visual story,” says Steve Grove, CEO and publisher of the Star Tribune, speaking at a WAN-IFRA webinar hosted by Cecilia Campbell. “What took place in the streets of our Twin Cities was probably the most documented crisis in American history up to this point.”

Grove, a former Google and YouTube executive who returned to Minnesota eight years ago, took over the Star Tribune four years earlier. The newsroom had already covered George Floyd’s murder and the civil unrest that followed. The Floyd coverage won a Pulitzer.

Safety First

Before the ICE raids escalated, editor Kathleen Hennessey had commissioned crisis training. A national safety firm ran a two-day workshop covering field protocol, navigating tear gas, PPE, and evacuation procedures. Basic hygiene for reporting in hostile environments.

“Journalism today doesn’t just mean in the streets, it means online,” Grove says. The newsroom ran a parallel upgrade of its cybersecurity. Doxxing and online harassment are not just risks for national political journalists. They arrive at local newsrooms too, particularly in a polarised coverage environment.

When Alex Pratt was shot on a Saturday, Grove says, “you just had instantly a dozen reporters volunteering to come in.” Outside normal working hours. No hesitation.

The Misinformation Problem

Citizen documentation drove the story. Thousands of bystanders filmed events in real time.

In one case, footage showed a masked ICE agent involved in a shooting. Someone took a still from the video, uploaded it to Grok, the AI tool on X, and asked it to digitally reconstruct the face behind the mask. The AI complied. The generated image was then run through a reverse image search, which returned a match: a gun shop owner in Missouri named Steve Grove.

Within hours, misinformation had linked the Star Tribune’s own publisher to the shooting.

“Suddenly, there was this rumour that Steve Grove had shot Renée Good,” he says. “Which is preposterous, of course.”

The Star Tribune’s reporters eventually identified the real shooter and published the story. By then, the misinformation had spread further than the correction.

The newsroom also used AI verification tools to identify fake footage. When a genuine video of Pratt emerged weeks after his shooting, someone created an AI-generated version and circulated both simultaneously. The product and news teams worked together to debunk it. “We’ve invested in the newsroom,” Grove says. “That’s what it takes.”

The Commercial Response

The Star Tribune made all ICE-related coverage free. Subscriptions went up anyway.

“People would land on a page that they didn’t have to pay to get to, and then they would subscribe,” Grove says. They did it because they believed in journalism. Donations arrived from 45 US states. The newsroom was already 40 weeks into a consecutive streak of digital subscription growth.

The Star Tribune launched a campaign titled “Because the World is Watching.” They lifted the cap on gift articles, launched a family subscription allowing four email logins per account, and wrapped the set under a brand message that answered the obvious question: Why pay attention to a paper in Minnesota? “The answer is, well, the world’s watching Minnesota,” Grove says.

Print subscribers are declining at around 20 per cent year-on-year. Digital is growing, but the gap is real. Philanthropy is now a formal channel, startribune.com/donate has been running for some time, and the crisis has significantly accelerated inbound giving.

The Investigative Follow-up

Once the ICE operation scaled back, the Star Tribune launched a standalone investigative team, a structural change Grove had been working toward. Their first major project is the Data Detention Project, a collaboration with multiple local newsrooms attempting to identify who was among the 4,000 people ICE claimed to have arrested.

That means court filings. Data journalism. Shoe leather. The kind of reporting that takes months and doesn’t trend on social media.

“A big story happens, then how do you dig into precisely what took place and make sense of it for your readers?” Grove says. “That’s been the journey we’ve been on.”

What Other Publishers Should Take From This

Grove’s message to local publishers was direct: you never know when a crisis will hit your community, so preparation has to happen in the quiet moments. Training, infrastructure, philanthropy mechanisms, these need to be in place before the story breaks, not after.

He also challenged the instinct to simply report and move on. “When these crises hit, you can’t be shy about asking for help. It’s okay to campaign for local news when a crisis hits.”

Minnesota has structural advantages, great statewide pride, a dense local news ecosystem, and an owner willing to invest. Not every local newsroom has those conditions. But across all of them, the same question applies: when the big story breaks, is the infrastructure there to meet it?

The 2026 FIFA World Cup is the biggest men’s tournament in history. More matches, three host nations, record audiences. Yet according to new research from WARC Media, its impact on global advertising is heading in the wrong direction.

WARC forecasts a $10.5 billion uplift to the global ad market in the quarter the tournament runs. Set it against the 2018 Russia World Cup — which generated a $12.6 billion boost, a 2.8% incremental gain — and the comparison is uncomfortable. The 2026 figure represents a 1.1 per cent increment on 2022’s Qatar tournament. The world’s biggest sporting event is delivering diminishing returns for advertisers even as it expands.

Why the Numbers Keep Shrinking

The World Cup used to mean a single commercial surface: broadcast television, a captive audience, a relatively straightforward media buy. That model is fracturing.

WARC’s Global Ad Trends report documents the shift. Linear TV audiences for Qatar 2022 fell by 11.9 per cent compared with 2018, even as total viewership reached 2.87 billion people. Audiences aren’t disappearing — they’re dispersing. Digital viewing expanded significantly, particularly in China and India, pulling consumption away from the channels that traditionally generated the most advertising revenue.

The 2026 tournament will accelerate this. TikTok has become a FIFA partner and will show behind-the-scenes footage. YouTube will stream live matches from rights holders. Netflix is positioning itself to monetise the conversation around games through video podcasts. None of these platforms acquired broadcast rights in the conventional sense. All of them will capture audience attention, and advertising spend that would previously have flowed to linear broadcasters.

Premium tournament pricing increasingly reflects redistribution rather than market expansion. Advertisers aren’t spending more overall; they’re moving existing budgets around.

Host Markets Aren’t a Sure Bet

The conventional assumption holds that hosting the World Cup is an advertising windfall for the countries involved. WARC’s data doesn’t support it. Annual ad spend growth in Mexico shows no consistent pattern of acceleration in World Cup years. Canada follows a similar trajectory. WARC’s forecast of around four per cent growth for Mexico is positive, but not exceptional for a host market.

The USA, where co-hosting duties come alongside the fact that soccer competes with multiple established domestic sports, tells its own story. In the most positive years on record, the World Cup’s effect on US ad investment has been between 0.4 and 1% of total spend. Broader economic cycles matter more than tournament cycles.

Where the Real Opportunity Sits

WARC’s analysis is pointed in terms of where brands should actually be looking. Alex Brownsell, Head of Content at WARC Media, said: “This World Cup is no longer just about live matches — brands will engage with fans across touch-points before, during and after matches have concluded. Media plans will include platforms that benefit from the conversation about the World Cup without the burden of bidding for rights — from creator content to podcasts, turning conversations around the games into powerful opportunities for connection and impact.”

Rights are expensive and increasingly contested. The conversation around rights — creator content, podcasts, social media commentary, publishing — is not. Brands that treat the tournament as a content ecosystem rather than a broadcast slot may find better value than those competing for diminishing linear inventory.

The late-night scheduling for European audiences further sharpens this. Fewer than half of 2026 matches, 42.3%, will air during daytime hours in Western Europe. In China, that figure drops to just over a third. Live broadcast advertising opportunities during those windows will be limited.

But the scheduling creates category-specific openings. In the UK, high-fat, sugar and salt products are subject to pre-9 pm advertising restrictions. A World Cup running into the early hours effectively removes that constraint for several weeks. For brands in restricted categories, quick-service food delivery being the obvious example, the timing is commercially useful in a way that wouldn’t apply to a European or Middle Eastern tournament.

The Audience Is Still There

Football remains the world’s most popular sport, with 51% of global respondents identifying as fans. Viewing levels during Qatar 2022 were highest in Africa, Latin America, and the MENA region. In the UK and Germany, European audiences remain concentrated and substantial.

In the US, 37% of Americans expect their interest in football to increase over the next 18 months — meaningful momentum heading into a tournament on home soil.

Reach is growing. The ability to monetise it through traditional channels is not keeping pace. Rights holders charge premium rates for access to mass audiences that are no longer quite as mass as they were. Non-rights holders capture attention without the infrastructure costs.

For advertisers planning 2026 campaigns, WARC’s data points in one direction: build media plans around the conversations that surround the games as much as the games themselves.

WARC Media’s Global Ad Trends: FIFA World Cup 2026 report is available to WARC subscribers. A podcast on the findings will be available from 31 March.

The GWI founder and CEO on supercharging the business around AI agents, why most LLMs produce “friendly rubbish” without proper data, and what going agentic first actually requires.

When Tom Smith and I spoke at Cannes Lions last year, GWI’s AI ambitions were already clear enough. The company had built foundational infrastructure to run natural-language queries across its dataset, and Smith was carefully considering how consumer insights could flow into the tools marketing teams use. That conversation was barely nine months ago. A lot has changed.

Agent Spark launched in January 2026, integrating GWI’s proprietary dataset, 35 billion data points drawn from 1.4 million annual surveys across 52 countries, directly into ChatGPT, Claude, and Microsoft Copilot. The idea is to put verified, human-grounded consumer insights into the AI tools marketing teams are already using, without the detour through a separate analytics platform or a request to a research team that’s already at capacity. More on how it works at Agent Spark.

Smith is direct about what prompted the move. “We could see that what people wanted was to connect insights to proper output and answers,” he says. “And one of the technologies that transformed the ability to do that was MCP — Model Context Protocol — which wasn’t really around even a year ago.”

MCP is what makes Agent Spark work at this level. It allows GWI’s agent to communicate directly with other agents, such as Claude, ChatGPT, and Copilot, so that insights are automatically pulled into the workflow a marketer or sales team is already running. What was previously a separate step, requiring a trip to the research department, becomes part of the same conversation.

The lean insight team’s problem

Smith’s argument for Agent Spark rests on something that rarely gets discussed openly: insight teams inside large organisations are extraordinarily lean. “It’s not untypical that you might have two researchers supporting a thousand salespeople,” he says. “You might have a hundred thousand employees and a hundred people working in consumer insights. But the demand for that data just can’t be satisfied at that scale.”

Rather than positioning Agent Spark as a cost-reduction tool, Smith frames it as a growth play. More people getting insights means more decisions grounded in data. Sales teams pitch faster. Marketing strategies get stress-tested before they go live. The research function’s reach expands without requiring additional headcount.

One of his early case studies is from a major electronics goods company. The shift from a handful of expert users of GWI’s products to a full marketing organisation illustrates the three things Smith says Agent Spark is actually solving:

• Speed: more than a hundred marketers now get answers directly, without routing through the research team

• Language: the agent works in Cantonese or whichever language the team is operating in, removing a significant barrier for globally distributed organisations

• Scale: what previously required specialist knowledge to extract and interpret is now available to anyone with a business question

“That’s speeding up workflows by a thousand times,” Smith says.

Omnicom is another reference point. Bharat Khatri, Chief Digital Officer APAC at Omnicom Media Group, has described testing Agent Spark in the agency’s autonomous-agent workflows, with plans to integrate GWI audience taxonomies into end-to-end media activation and creative iteration. Smith explains what that looks like in practice: when a platform needs to tailor a global campaign across 25 markets, it can now automatically pull audience context from GWI, feeding the data into creative models as input for localisation. A process that previously required considerable human coordination can now occur in near real time.

Why generic AI gives you “friendly rubbish”

Accuracy is where the conversation gets interesting. GWI’s own research suggests 85 per cent of marketers use AI to support their work — but most of that is generic AI, without access to properly structured audience data.

“If the data isn’t there, ChatGPT and Claude produce friendly rubbish,” Smith says. “They produce really generic, inaccurate output that is difficult to act on, and can be quite dangerous if you use it. Most consumers aren’t represented in the content that trains large language models.”

I pushed him on the hallucination risk, the concern that the AI presentation layer might introduce inaccuracies even when the underlying GWI data is correct. His answer comes down to the difference between open-ended and concrete questions. Ask Agent Spark a broad question, such as the ten most unexpected things about Xbox gamers in Germany, and then interrogate the output. Ask how many people in Germany use Xbox, and the answer will be correct every time because the model pulls from a deterministic data layer rather than generating text via pattern matching.

“As long as the question is right, the answers will come back accurate,” he says. “We’ve spent a lot of time testing and calibrating. If you bring an agent to market, you have to validate this.”

Data company, AI company, or both?

I asked him the obvious question: Is GWI still a data business, or has the product evolved into something else? He doesn’t think the distinction holds any more.

“We can’t operate like a software business from ten years ago,” he says. “Our whole strategy is about being an agentic-first business. Even if you come into our platform now, the dashboards are still there, but over time, everything will evolve into an environment where you use the agent to do any task.”

What Smith is describing positions GWI not as a competitor to the LLMs, but as the data infrastructure they need to be useful in a marketing context. The LLMs handle the interface and reasoning; GWI provides the human truth beneath the surface. That’s a bet that proprietary, first-party data becomes more valuable as AI spreads, not less, because without it, the outputs are, as he puts it, friendly rubbish.

He points to companies building similar models in adjacent sectors — structured, proprietary data used to ground AI outputs for specific professional contexts — as evidence that the model works. “LLMs and agents are incredibly accurate if the data and context are well-defined and accurate,” he says. “The biggest challenge is having clean data in the right format. That’s our focus.”

2026: the agentic pivot year

Smith is confident this is the year the agentic model breaks through in agencies. The past twelve months have been build-out — platforms constructed in the background, partnerships established, pipelines developed. This year, those systems will go public.

“Every agency will need that kind of system in order to be relevant and competitive,” he says. “I think this is the big year where it breaks out.”

On the industry’s shape by 2027, he is more cautious. The mega-mergers — Omnicom’s acquisition of IPG being the most significant — are partly about building the scale to invest in these platforms. Smaller boutique agencies may find that AI gives them the reach that their size wouldn’t previously have allowed. The squeeze, he thinks, will fall hardest on mid-tier groups: large enough to need the infrastructure, not always large enough to fund it.

There’s also a counterintuitive point he makes about workloads. In creative production, the ability to generate infinite variants of an ad doesn’t reduce demand — it tends to increase it. Clients who once wanted three ideas now expect ten. “AI can make more work, not less,” he says.

GWI is reorganising its entire business around agents rather than attaching them to existing structures. Platform, product, commercial model, user experience — all of it. Smith says companies that approach this by “just building a couple of chatbots” won’t cut it. Whether that assessment ages well will depend on how quickly the rest of the market catches up.

SAP Engagement Cloud is six years in the making. The solution, until recently known as SAP Emarsys, is now fully embedded within SAP’s operational stack, sitting on the same data that runs supply chains, tracks inventory and powers ERP systems across some of the world’s largest enterprises. Joanna Milliken, its CEO, is clear that the name change is the least interesting part of the announcement.

“From my experience seeing competing systems internally over who’s the primary marketing solution is just a waste of time and energy,” she told The Media Stack.

What the move represents, in Milliken’s account, is six years of integration finally delivering something structural. SAP Engagement Cloud now sits directly on SAP’s operational data, the same ERP data that tracks inventory, manages supply chains and runs procurement across some of the world’s largest enterprises. This is not an API connection or a data feed. The product and the operational layer are, in her words, the same core.

“Engagement decisions are increasingly dependent on business context, service issues, inventory levels, supply chain breakdowns happening locally. What SAP Engagement Cloud represents is our ability to combine those proven omnichannel capabilities with SAP operational data and the AI platform. That’s what’s different.”

What existing customers actually see

For day-to-day campaign users, the change is incremental for now. The segmentation tools, channel capabilities and automation are the same.

“They’re going to see their job as generally the same — they’ll still segment, they can still hit the send button,” Milliken said.

“But that’s going to change dramatically over time.”

At the CMO level, the proposition is more immediate. Milliken’s argument is that most large brands currently run separate engagement systems for consumers, distributors, retailers and partners. Those systems share almost nothing despite drawing on the same product data, the same compliance requirements, and the same brand guardrails. She used Wella as an example. A product launch would typically run a consumer campaign in parallel with entirely separate communications to retail and distributor networks: launch kits, promotional calendars, and demand alerts.

“Right now, that’s coordinated across different departments, different tools. If there’s a common view across all of those, the data is the same, the tools carry the same thread, you get a much more holistic analysis of your business. You can bring it all together in one solution.”

For brands already running SAP’s ERP platform, the integration is direct: product IDs, inventory positions and supply chain signals connect without heavy configuration. For those outside the SAP ecosystem, Milliken’s case rests on the same underlying problem.

“Being able to communicate to all of your audiences with the same toolset, the same data, with the same IT teams doing the same product catalogue ingestion, that reduces complexity. And brands need to flex with their business. You shouldn’t have to go and buy a new piece of technology every time your model changes.”

The gap SAP’s own research exposes

The announcement arrived alongside new research that gives Milliken’s positioning some external grounding. Conducted among 10,000 consumers and 4,800 executive leaders globally, the study found that only 22% of brands believe they have a problem with delivering a seamless customer experience, while 82% of consumers report dissatisfaction. SAP CMO Sara Richter, presenting the findings at the company’s Engage online event, called it the engagement divide.

Mark Ritson, renowned marketing professor and founder of the Mini MBA in Marketing, spoke at the Engage event and put the gap in context. He cited Bain research showing that 80% of brands believed they were delivering superior customer experience, while only 8% of customers agreed.

“The SAP report is the update a decade on,” he said. “We’re not getting any better.”

Ritson’s diagnosis maps directly onto what Milliken describes as the problem her solution addresses. The gap, he argued, is not primarily a technology failure. It is organisational: siloed teams, fragmented data, and a marketing function that cannot see beyond its own systems. The SAP research found that only 40% of decision-makers believe their departments are truly coordinated on customer experience. Data compounds the problem: 60% of companies have what the report terms “dark data” they cannot use, and two-thirds remain dependent on third-party rather than first-party data pools.

“You need to engage with your consumers, your distributors, your retailers, your partners, your employees. Right now, that engagement is very fragmented, in different systems, becoming unnecessarily complex when all of it should sit on top of the same business data, the same product data, the same compliance and guardrails,” Milliken said.

Loyalty: why discounts are no longer enough

SAP also published an updated Omnichannel Guide to Retention and Loyalty alongside the Engage event, drawing on four years of Customer Loyalty Index data from surveys of more than 10,000 consumers across the US, UK, Australia and Germany. The numbers sit awkwardly alongside the way most brands still spend their retention budgets.

The report identifies five distinct types of customer loyalty — from incentivised loyalty built on discounts and points, through inherited and silent loyalty, to ethical and true loyalty at the far end. The further along that spectrum a brand can move its customers, the better the commercial outcomes.

The trend data across 2021 to 2024 makes the direction clear. Incentive-driven loyalty has remained broadly flat. Experience-led true loyalty has grown by six percentage points; ethical loyalty, driven by a brand’s stance on social and environmental issues, by seven percentage points. The conclusion is straightforward: discount and points-based retention programmes are delivering diminishing returns, while the investment case for experience-led engagement is getting stronger.

The research on what drives and damages loyalty is equally pointed. High-quality products top the list of positives at 57%. On the negative side, irresponsible data handling is cited by 32% of consumers as damaging loyalty, and making it harder to reach a human by 24%, both live risks for brands deploying AI-driven automation without the operational data infrastructure to make it feel personal.

It is a tension Milliken acknowledges directly in her AI roadmap: the engagement agent she is building is designed to scale personalisation, not replace the human judgement that sets the guardrails around it. The loyalty data suggests that distinction will matter to customers.

The AI agent Milliken is building for 2026

Through 2025, SAP built out AI capabilities across segmentation, analytics and content generation within the platform. Milliken’s roadmap for the first half of 2026 focuses on bringing those pieces together as an engagement agent: a system that works alongside the marketer to produce audience selections, channel mixes, and content approaches from a stated goal, with the marketer setting the business rules and guardrails.

“You put in your goal at the top, it generates the right content, the right audience, the right business logic, the right channels — and then you can manipulate it, and it continuously does that. We have all those pieces and parts. You’ll start to see that come together over the summer.”

Ritson’s argument at the Engage event ran in the same direction. The greater value of AI in customer engagement, he said, lies not in front-end features consumers can see but in the back end, predictive segmentation, operational signals that feed real-time decisions, and churn detection before it becomes churn.

“The less we talk about AI and the more we do AI, the better our engagement becomes.”

Six years after SAP acquired Emarsys, Milliken’s case is that the platform is now genuinely different from the one it acquired: not a mid-market direct-to-consumer tool sitting within a larger organisation, but the engagement layer for the enterprise data beneath it. Whether that framing lands with marketing teams used to working in their own systems is the next test.

Stuart Forrest is not panicking. Given the scale of disruption currently hitting digital publishing, declining organic referrals, AI overviews cannibalising click traffic, and Google Discover delivering less traffic, that is worth paying attention to.

Forrest joined Bauer Media in 2023 as Head of Audience Development, a role that spans the publisher’s UK and German portfolios. Bauer operates across automotive, women’s lifestyle, entertainment, food, golf, motorcycling and fishing in the UK, with parallel categories in Germany. It is a business built on specialist depth, and that matters more than most people currently give it credit for.

The team of nearly 30 people across both territories spends much of its time not doing things directly, but influencing others to do things differently. “A lot of it is coaching,” he says. “We’re constantly trying to influence editorial teams and product teams. How we communicate internally is something I put a lot of importance on.”

LLM Optimisation Is Not a New Discipline

When Forrest delivered a masterclass to industry peers earlier this month, one framing cut through: LLM optimisation isn’t a new discipline. It’s SEO with a different shape.

“Whatever surface we want our content to land on, the algorithm has to decide whether it’s interesting and valuable enough to show the user. That’s job one. Job two is whether it’s presented in a way that makes them engage with it. Those two principles remain the same.”

The signals that determine whether content surfaces in an LLM — authority, clarity, structured data, topical expertise — have significant overlap with organic search fundamentals. Where it breaks down is economics. Traditional search is built on links and clicks. LLMs synthesise across multiple sources and produce a single answer.

“They might read ten articles and produce one answer. The value exchange for publishers, compared to a referral-based model, is very different.”

Since late January, Google’s AI Overview interface has included a ‘View More’ link that pushes users into full AI mode rather than returning them to blue links. Forrest ran experiments the morning of our conversation. His reading: the AI overview is being deliberately truncated, not complete enough to satisfy the query, but just enough to prompt a click deeper into AI mode.

“If that’s correct, it’s obviously a real concern.” He is straight about the limits of his certainty. “I don’t have the data on this.” What he does note is that citation cards within AI mode have a more prominent format — logo, image, meta description — and whether that translates into meaningful publisher traffic is something he’s actively trying to measure.

The EU press’s view, he says, is that the direction of travel is bad for publishers. “They’re probably right.”

Google Still Needs Publishers. Probably.

Whether Google will continue sending traffic to publishers is, for Forrest, part strategy and part informed bet.

“It knows it needs to find some form of value exchange with publishers so it can have access to that content. We don’t know what that looks like.”

The regulatory context helps. Mounting pressure across Europe and the US over how AI companies compensate content creators is making the asymmetry increasingly visible to policymakers. The argument Forrest finds compelling is that LLMs need computing power, energy, and content. They’re only paying for two.

Google Discover, he suggests, may have functioned as a quiet concession, a way of maintaining referral traffic while traditional search clicks declined. But 2025 has been difficult there, too. News Reach reported traffic down 50% year-on-year. Not everyone is seeing that, he says, but it’s not an isolated case.

Bauer’s organic search traffic is down in the high single to low double digits, depending on where you look in the business. Some brands are still growing in organic year on year, but the composition is shifting. “We’re seeing a change in the types of content that drive those clicks compared to where they were last year.”

His longer view draws on print. People were writing off print a decade ago. It found its level. Search, he argues, has moved from being the great hope for publishers into late-life-cycle territory, requiring active triage rather than denial. Some content types are declining faster than others. The job is to work out which is which and allocate resources accordingly.

The 47-Step Briefing Tool

The most concrete thing Forrest discussed is an internal content briefing tool now in production across Bauer’s UK and German portfolios. It runs 47 steps: extracting data, pulling HTML from competitor pages, analysing ranking positions, and mapping internal link structures. It produces an editorial brief.

It does not write articles. It creates the brief an editor uses to write one. That distinction matters to Forrest.

“We’re leaving all the room for tone, context and expertise. We’re using it to shortcut a lot of the manual research process.”

Where it’s most powerful is in competitor analysis: what are the minimum requirements to compete for a given keyword, and where has a competitor recently outranked Bauer, and why? There are two versions: one for net-new articles, and one triggered by a ranking drop, which reviews what’s now outperforming the existing piece and makes suggestions.

The smallest brands in the portfolio are using it on a self-service basis. Angling Times, which doesn’t have a dedicated SEO resource, is using it to improve buying guides. It tells the editor not which products are best — that knowledge belongs to the journalist — but what the content framework needs to look like to be competitive. “Very often it’s reinforcing their instinct anyway. It’s shortcutting the research process, not replacing the expertise.”

Forrest is direct about the risks of AI adoption in editorial. “We’re in a world where a premium on just creating content no longer exists. The premium is on expertise and authority. We wouldn’t ever want to do anything which dilutes that.” That’s the hill I’d die on in my career.

“The premium is on expertise and authority. We wouldn’t ever want to do anything which dilutes that.”

Newsletters: What Job Is This Doing?

Forrest is wary of the word newsletter. Not because email is unimportant — Motorcycle News has a healthy database built on deep audience loyalty — but because the term slides too easily into something indiscriminate.

“You have to be really, really thoughtful about what’s going into that newsletter and why it’s there.”

Empire, Bauer’s film title, should be built around a specific weekly problem: what should I watch this weekend? Today’s Golfer should adjust its email rhythm around tournament schedules. Parkers, the automotive valuations platform, is better characterised as a transactional brand in finance. The expectation isn’t regular weekly engagement. It’s warming up users throughout the purchase cycle so they’re active when they’re next in-market for a car.

“Think about your newsletter subscribers as your most loyal audience. Because if they are, they’ve already read the biggest story from Wednesday. There’s no point putting that in the newsletter.”

Blocking, Measuring and the Value Exchange

Bauer is currently aggressive on AI crawler management. Robots.txt blocking is in place, along with CDN-level restrictions. “You can’t create a market for a good unless you’re controlling access to that good.”

Forrest is tracking four metrics for AI visibility: referral traffic from LLMs (small but monitored), bot crawl data via Tollbit, AIO citation rates, specifically what percentage of ranking queries show an AI overview, and of those, how many cite Bauer, and share-of-voice monitoring, which he treats as directional rather than definitive.

On share-of-voice specifically, he’s sceptical of the tools on the market. Most are sampling data rather than direct measurements. “SEOs are quite comfortable with discomfort around the certainty of data. If there’s an SEO who didn’t do anything until they were absolutely certain, they’d never get anything done.” His approach is to measure the same thing consistently over time and use it as an index rather than a count.

In twelve months, he expects the landscape of content value exchange to have moved significantly. Microsoft’s content marketplace is emerging. ChatGPT has hired a UK-based partnerships lead. The CMA is active. IAB Tech Lab is establishing standards. “We didn’t learn how to trade advertising out of nowhere. Somebody had to write the standards.”

The advertising industry has spent the last two years restructuring, merging and cutting. Holding companies have shed thousands of jobs. Agencies have asked remaining staff to do more with less. And through it all, a small charity has been quietly handling the fallout.

NABS — which describes itself as the charity helping everyone in advertising, media and marketing move forward — recorded over 14,000 engagement points in 2024, a 20% increase on the previous year. Its advice line took more than 5,300 calls. Mental wellness was the number one reason people contacted, followed by redundancy and financial support.

Katrina Urban, Head of Learning and Development at NABS, has been at the organisation for 12 years. She joined as the office manager. She stayed because of the work. Speaking to The Media Stack, she was characteristically direct about what the industry is going through — and what it keeps failing to address.

The People Who Stayed

The conversation around redundancy tends to focus on those who left. Urban thinks that’s missing half the picture.

“The impact doesn’t stop with the people who leave. When teams shrink, people are expected to do more with less. That can stay there for months, even years; it’s like a barrier of trust that gets damaged.” - Katrina Urban

She describes what she sees as a ripple effect: uncertainty about further rounds of cuts, survivor’s guilt when colleagues disappear overnight, and a cultural shift within teams that leadership often underestimates. The Omnicom-IPG merger, WPP’s restructuring, and broader consolidation across the sector have amplified it all.

NABS has a redundancy guide on its website that has racked up thousands of downloads. Urban is careful to distinguish between engagement points and individual contacts; one person might access the advice line, attend a workshop, and go through therapy. That’s three touch points, one person. “We never turn anyone away,” she says. Given the demand, that’s not a small commitment.

What Managers Aren’t Being Taught

NABS runs a suite of training programmes, all free to access. The most attended is Managers Mindsets, a programme that grew out of the organisation’s All Ears community consultation, which made it clear that managers were being failed by their employers.

“They might be trained on how to run a PIP. But are they being trained on how to have mental wellness conversations with their teams? How to give feedback constructively? You get put straight into a one-to-one and expected to run the process.” - Katrina Urban

The programme covers managing pressure, leading through change, difficult conversations and giving feedback. Urban is clear that attending a session doesn’t automatically change behaviour — the sessions are designed to be practical and scenario-based, with people practising the skills in the room rather than absorbing theory they’ll never apply.

Organisations can buy in the same sessions available to individuals, sometimes with bespoke elements added. The uptake, she says, has been high and the informal feedback from managers who’ve attended tells its own story.

Cannes Lions: A Magnifying Glass

Urban wrote about the ongoing sexual harassment at Cannes Lions. The article published in Campaign, the advertising industry’s trade publication, was direct enough to generate attention in an industry that usually prefers to address these things quietly, if at all.

Her view is unambiguous. What happens at Cannes isn’t unique to Cannes — it’s a compression of what happens across the industry year-round. Working away from home and alcohol are the two factors most associated with increased harassment risk. Cannes provides both at full intensity for five days.

“People act like they’re on holiday. They’re actually working. Boundaries massively slip. And the people who speak openly about it are a very small portion of those who experience it.” - Katrina Urban

She credits the Cannes Lions team for taking it seriously. NABS worked with them last year to run training sessions in the weeks before the festival, but it is clear that the responsibility can’t rest solely with the festival organisers. Advertising, Media agencies, Hold Co’s and technology companies booking large spaces at Cannes need to take ownership of those environments. Zero tolerance can’t be a policy that applies in some places and not others.

NABS's bystander intervention training is built around five methods: Defuse, Distract, Challenge, Document, Check-In and Report. Each offers a different way to respond to harassment from creating a diversion to confronting the harasser outright. When asked which element people find hardest to use in the moment, Urban doesn't hesitate: Challenge. The misconception that confronting inappropriate behaviour means being aggressive runs deep. The training works through multiple scenarios drawn from real industry situations, showing that direct intervention doesn't have to mean a confrontation; it can be as simple as saying "that's not okay" if it feels safe to do so. The goal, she says, is getting people comfortable with being uncomfortable.

AI: Anxiety in the Margins

AI isn’t yet the primary reason people contact NABS. But it’s increasingly part of the wider context. Urban is careful to distinguish between the genuine opportunities and the anxiety that’s showing up in conversations with the support team.

Junior staff are worried their tasks will be automated first. Creatives are uncertain about what their work will look like in an industry that keeps telling them to embrace tools that can replicate their work. The pressure to upskill is real — but, as Urban notes, “where do you upskill when AI is moving at a pace that nothing can keep up with?”

She’s firm on one point: AI shouldn’t replace human insight, especially not management skills. When it does, the work becomes inauthentic and generic. Urban sees the downstream consequence clearly: you can’t become a senior if you were never allowed to be a junior.

Ageism: The Least-Discussed Problem

IPA Census data show that just 8% of agency staff are over 51, compared with 33% in the general working population. Urban doesn’t dress this up.

“The numbers speak for themselves. People tell NABS they feel like they’ve been put on the scrap heap. They’ve got decades of experience, they struggle to get callbacks, and finding new roles can take much longer. Many eventually leave the industry altogether.” - Katrina Urban

She acknowledges that it's sometimes framed as a cost issue; older, more experienced people are more expensive to hire. But the correlation with age is clear, and NABS's support team is having these conversations regularly. Coaching by referral can help people re-enter the industry, but Urban is under no illusions: coaching people back in is not the same as stopping them being pushed out.

The issue compounds. Women who took time out to care for children or elderly parents, and workers from minorities communities, face a harder road back. NABS works with organisations, including MEFA and Outvertising, to ensure its offer reaches all of them.

The Common Thread

Across mental health, ageism, harassment, AI anxiety and merger fallout, Urban sees one through line: people’s sense of security, confidence and their ability to feel they can thrive at work.

NABS has also just launched its Next Gen Board, for people aged 30 and under, with experienced mentors involved, designed to create the kind of career longevity support that’s becoming harder to find inside agencies themselves.

For an industry that has spent two years telling itself that disruption is just transformation in disguise, the demand for what NABS provides suggests something simpler: people are struggling, and the organisations employing them aren’t always equipped to help. “We’re in people’s corner,” Urban says. For a growing number of people in advertising, media and marketing, that corner is getting more crowded.

Katrina Urban is Head of L&D at NABS. nabs.org.uk

In June 2025, at the Palais des Festivals at the Cannes Lions Festival, the Titanium Lion, the festival’s most prestigious award, given only to work that shifts the direction of the industry, went to a campaign built around three words added to an insurance policy. This writer was in the front row when it was announced.

The campaign was Three Words, created by Publicis Conseil Paris for AXA. The idea was precise: add the phrase ‘and domestic violence’ to AXA’s standard home insurance terms in France. Those three words reclassified domestic violence as an insurable risk, giving women in abusive relationships immediate access to emergency accommodation under a policy they already held. No new product. No new cost to the customer. Three words in a legal document that millions of people already owned.

Nine months later, Three Words tops the WARC Creative 100 rankings, published on 10 March, accumulating more points across global and regional awards shows than any other piece of work entered in 2025. The Titanium was its most prominent win, but far from its only one.

“I believe the reason this idea resonated with so many people is the contrast between its simplicity and its exponential impact. With just three words, we are pivoting millions of home insurance policies to make them more inclusive of a risk that affects women more than fires or floods. It serves as a reminder of the power of creativity. It proves that even the most common product – like a home insurance that hasn’t evolved in centuries – can be reinvented to make a difference.”

Marco Venturelli, CEO and Chief Creative Officer of Publicis Conseil and Global CEO of Leo

The WARC Creative 100 aggregates results from the most significant global and regional creative awards shows each year, determined by an industry advisory board and annual global panel survey. It is one of the few genuinely independent benchmarks in the industry — no single show or geography dominates the methodology.

B2B in the Top Three

The second-placed campaign is also worth attention. Spreadbeats, by FCB New York for Spotify, repackaged the platform’s advertising performance data inside a functional spreadsheet — a piece of work aimed squarely at media planners, designed to shift their perception of Spotify’s video ad capabilities and ROI. It is a B2B campaign, and it finished second in the world.

Third place went to Michelob Ultra’s Formula 1 partnership Lap of Legends, which staged a race between real and virtual drivers. Two of the top three campaigns this year operate in the B2B space, a shift WARC has identified as a broader theme, with creative work increasingly used as a differentiator in business-to-business marketing rather than reserved for consumer-facing brands.

LePub and the Independent Story

At agency level, LePub Milan takes the top spot for the first time, rising from second place last year. Two Heineken campaigns in the top 100 anchor its points tally, alongside contributions from across its portfolio. CEO Bruno Bertelli said:

“LePub Milan is honoured to be ranked #1 in the WARC Creative 100. This achievement celebrates our visionary clients and extraordinary team, who inspire bold, culturally resonant ideas. At the intersection of creativity, strategy, tech, and insight, we craft work that resonates in people’s lives, builds brands, and delivers measurable impact. This recognition reinforces our commitment to innovation, rigor, and fearless creativity, helping clients move faster, smarter, and with confidence worldwide.”

The independent agency story belongs to Serviceplan. Serviceplan Munich climbed from seventh to second in the overall agency ranking and simultaneously took the top independent agency position. The Munich office also leads the independent network ranking for the second consecutive year — making Serviceplan the top-ranked independent operation at both agency and network level.

Ogilvy’s Sixth Consecutive Year at the Top

Ogilvy leads the network ranking for the sixth year running — a run that began in 2020 and shows no signs of ending. Fifteen campaigns in the top 100 this year, contributed by offices across New York, Singapore, Mumbai, Madrid and Lima. Global Chief Creative Officer Liz Taylor:

“Six years at the top of WARC’s Creative 100 is more than a milestone — it’s a testament to how our network shows up. The best work, on the biggest brands, in every market around the world. I am endlessly proud of our teams and our clients who, every day, prove that creativity isn’t just what we do together, it’s the most powerful competitive advantage out there.”

WPP holds the number one holding company position for the fourth consecutive year, with Ogilvy and VML ranked first and third respectively among networks, and AKQA at twenty-eighth.

“Our continued recognition in the WARC Creative 100 is a testament to WPP’s incredible talent and the hard work of our dedicated teams globally. While awards represent the visible tip of the iceberg, we fundamentally believe in the power of creativity to shape culture and drive profound, sustainable growth for our client partners. This commitment to delivering real-world impact for the world’s most iconic brands is what truly defines our success.”

WPP’s Global Chief Creative Officer Rob Reilly

Omnicom Group and Interpublic Group are ranked second and third as holding companies, assessed on their respective performances prior to the completion of their merger.

Heineken and Unilever’s First-Time Wins

Two brand-level firsts stand out. Heineken reaches number one brand for the first time in the ranking’s history, having appeared in the top ten for four consecutive years. Its position is built on genuine global reach — two campaigns in the top 100, nineteen in the full database, across twelve countries. Nabil Nasser, Head of Global Heineken Brand:

Unilever takes the top advertiser position for the first time, represented by fifteen brands in the full Creative 100 database. Vaseline and Dove both reach the top 50 brands — Vaseline, notably, as a new entrant driven by its Vaseline Verified and Transition Body Lotion campaigns across Asia.

The United States retains the top country position — a rank it has held since 2006. Thirty of the top 100 campaigns originated from US agencies. The UK is second with eleven, Brazil third.

The WARC Media 100 publishes on 13 March and the Effective 100 on 18 March.

The full WARC Creative 100, including campaign case studies, is available to WARC members at warc.com

The IAB Tech Lab has released the Content Monetisation Protocols (CoMP) specification for public comment, establishing, it says, a standardised framework for AI systems to agree commercial terms with publishers before crawling or using their content. The specification is open for industry feedback until April 2, 2026.

The premise is straightforward. AI companies need content. Publishers produce it. Right now, there’s no consistent commercial infrastructure governing that relationship. CoMP is designed to fix that by creating a standard protocol that content owners and AI systems can use to signal permissions and agree terms before any content is accessed.

AI systems require chips, power, and information. Information is the only input in that equation that does not yet have a consistent commercial infrastructure around it.”

Anthony Katsur, CEO of IAB Tech Lab

Katsur is correct. The chip manufacturers get paid. The energy companies get paid. Publishers, whose content trains the models and populates the answers, largely do not.

What CoMP Actually Does

The framework is designed to work across both direct licensing arrangements and third-party marketplaces. The idea is that instead of every publisher building bespoke integrations with every AI platform, a technically and commercially unworkable situation, there’s a single standardised handshake.

Before content is accessed, CoMP creates a mechanism for signalling what’s available, on what terms, and at what price. It’s not an access control system in itself. IAB Tech Lab is explicit that publishers need robust blocking at the CDN or Edge Compute level first. CoMP then provides the commercial pathway from those restrictions to a structured market.

Think of it less as a lock and more as a till. You still need the lock. But without a till, anyone who does want to pay has nowhere to go.

For publishers who’ve watched search referral traffic fall by more than 50% in some cases, a figure IAB Tech Lab cites directly, the promise of a new revenue stream tied to AI usage is significant. The question is whether AI companies will actually adopt it, or whether CoMP joins the growing list of industry initiatives that get nodded along at conferences and quietly ignored in procurement decisions.

Industry Voices

Early support has come from several directions. Rob Beeler of Beeler.Tech, a publisher-focused ad ops community, welcomed the framework, arguing that while much remains unresolved about how LLMs will ultimately work with publishers, CoMP at least moves the conversation from theory to practice.

Jennifer Bas, Chief of Staff at Mobian, made a broader point about the nature of the relationship between AI and content: “As AI systems and agents become the front door to information, they inherit the authority, credibility, and trust of the publishers and brands whose content they learn from and surface. CoMP creates a mechanism for that quality and trust to be valued in the AI supply chain.”

That’s a more interesting argument than it might first appear. AI companies don’t just want any content — they want authoritative, trusted content. The FT, The Guardian, Reuters. If that quality is what gives AI answers their credibility, there’s a reasonable case that it should carry a price tag commensurate with its value to the product.

The media and data giant Bertelsmann has been an early contributor to the CoMP working group. Achim Schlosser, VP Global Data Standards at Bertelsmann, set out the company’s position clearly: “The first release of the CoMP API marks an important step toward establishing interoperable, transparent standards for fair value exchange in the AI ecosystem, recognising that AI systems depend on high-quality, trusted content. As an early contributor, we believe scalable, robust compensation frameworks — alongside visibility and attribution for content usage are essential to sustaining high-quality journalism and premium content in the AI era.”

The visibility and attribution point is worth noting. Compensation matters, but so does knowing when and how your content is being used. That’s something publishers have never had from current-generation AI systems.

The Harder Questions

CoMP is a technical standard, not a legal requirement. Its effectiveness depends entirely on adoption by AI companies, and the major players — OpenAI, Google, Anthropic — haven’t publicly committed to it. There’s a pattern here worth watching. Industry bodies create frameworks. The platforms wait to see what regulatory pressure materialises. Then, and often only then, do they move.

That’s not necessarily cynicism. Standards like OpenRTB took time to achieve critical mass. The IAB Tech Lab has a strong track record of building infrastructure that eventually becomes the industry default. But “eventually” is cold comfort for publishers who are losing traffic now.

There’s also a structural question about who benefits most from a standardised marketplace. Large publishers with legal teams and negotiating leverage will be positioned to do well. Smaller, independent publishers, the ones who arguably need the revenue most, may find themselves accepting whatever terms the marketplace defaults to, simply because they lack the resources to negotiate anything better.

What Happens Next

The public comment period runs until April 2, 2026. IAB Tech Lab is explicitly seeking input from publishers, marketplaces, technology providers, and AI developers before finalising the specification. That’s the right approach — a framework built without meaningful publisher input would likely serve the AI platforms more than the content creators it’s meant to protect.

For publishers, the immediate action is to engage with the comment process. Those who shaped OpenRTB had more influence over how programmatic advertising developed than those who simply showed up to implement whatever the spec said.

The content monetisation problem isn’t going away. AI companies are not going to stop needing publisher content. If CoMP can establish a working commercial infrastructure before national legislations create a patchwork of conflicting rules, it will have done something genuinely useful for the industry. Whether it gets there depends on who turns up to the table and who decides to stay home.

The CoMP specification is available for public comment at iabtechlab.com/comp until April 2, 2026. You can also read the IAB’s full press release HERE

About IAB Technology Laboratory

Established in 2014, the IAB Technology Laboratory (Tech Lab) is a non-profit consortium that engages a member community globally to develop foundational technology and standards that enable growth and trust in the digital media ecosystem. Comprised of digital publishers, ad technology firms, agencies, brands, and other member companies, IAB Tech Lab focuses on solutions for brand safety and ad fraud; identity, data, and consumer privacy; cross-channel programmatic effectiveness and ad measurement; and the impact of LLMs and AI agents on advertising. Its work includes the OpenRTB real-time bidding protocol, ads.txt anti-fraud specification, Open Measurement SDK for viewability and verification, VAST video specification, Trusted Server, and the Agentic Roadmap initiative for agentic advertising. Board members and companies are listed at https://iabtechlab.com/about-the-iab-tech-lab/tech-lab-leadership/. For more information, please visit https://iabtechlab.com.

A Pentagon contract, a ‘Quit GPT’ campaign with 1.5 million sign-ups, and a coding tool that went from zero to $2.5 billion in nine months. What actually happened — and what it means for anyone buying enterprise AI.

On the morning of 28 February 2026, Anthropic walked away from a contract with the US Department of Defence. The DoD had been negotiating with the company for months over the use of its AI models in classified operations. Anthropic’s position was straightforward: it would not allow its technology to be used for domestic surveillance of American citizens or in autonomous weapons systems capable of making lethal decisions without human oversight. The Pentagon wouldn’t accept those terms. Anthropic walked.

Hours later, OpenAI stepped in.

What followed was one of the most damaging weeks in OpenAI’s history and one of the most commercially spectacular in Anthropic’s.

The deal that backfired

OpenAI CEO Sam Altman announced on Friday evening that his company had agreed to deploy its models on the Pentagon’s classified networks. The timing, within hours of Anthropic’s public refusal, struck many observers as opportunistic. It struck some of OpenAI’s own employees the same way.

Internal frustration surfaced publicly within 48 hours. Research scientist Aidan McLaughlin posted on X that he personally did not think the deal was worth it. Others spoke to CNN anonymously, saying they ‘really respect’ Anthropic for standing its ground. The sidewalk outside OpenAI’s San Francisco offices was covered in chalk graffiti. Protesters gathered. A digital campaign called QuitGPT launched and, according to The Times (UK), attracted around 1.5 million sign-ups within days. [Note: some sources reported a higher figure of 2.5 million the lower figure is used here as the more conservative estimate.]

“We were genuinely trying to de-escalate things and avoid a much worse outcome, but I think it just looked opportunistic and sloppy. Good learning experience for me as we face higher-stakes decisions in the future.”

Sam Altman, internal memo to OpenAI staff, 3 March 2026

Altman moved quickly. By Monday, he had reworked the contract language to explicitly prohibit the use of OpenAI’s technology for domestic surveillance, including via commercially acquired data such as location information or fitness app records — the grey area Anthropic had specifically flagged as a concern. He also publicly stated that Anthropic should not have been designated a supply chain risk and urged the Pentagon to reverse the decision.

The revisions satisfied a few critics. Legal experts noted that the contract’s protections were anchored in current law, which the government could alter through executive orders. Jessica Tillipman of George Washington University’s law school noted that the published excerpt did not grant OpenAI a free-standing contractual right to prohibit otherwise lawful government use.

The ambiguity was precisely what Anthropic had refused to accept.

Blacklisted — and booming

The Pentagon’s response to Anthropic’s refusal was punitive. Defence Secretary Pete Hegseth designated Anthropic a supply chain risk to national security — a designation, according to MIT Technology Review, that has never previously been publicly applied to an American company. Under the designation, no Pentagon contractor or supplier can conduct commercial activity with Anthropic. President Trump separately directed federal agencies to immediately cease using Anthropic tools. The US Department of State, the Treasury, and the Federal Housing Finance Agency all announced they were ending their use of Anthropic’s services.

Claude surged to the top of the Apple App Store’s free app rankings over the weekend, displacing ChatGPT. The QuitGPT campaign explicitly recommended Claude as an alternative. By Monday, Anthropic’s run-rate revenue had jumped to approximately $19 billion — up from $14 billion just weeks earlier, and from $9 billion at the end of 2025.

Anthropic’s annualised revenue stood at around $1 billion at the end of 2024. The New York Times technology podcast Hard Fork cited roughly 20x growth in just over a year — a trajectory with no real precedent in enterprise software. [Note: The Media Stack has not independently verified the exact multiplier from the Hard Fork episode; the underlying ARR figures — $1B end of 2024 to $19B March 2026 — are sourced separately from Bloomberg and Sacra.] Bloomberg confirmed the $19 billion run-rate figure, citing people familiar with the matter, attributing the growth to the adoption of Anthropic’s AI models and specifically its coding tool, Claude Code.

Claude surged to number one on the Apple App Store. Anthropic’s annualised revenue had reached $19 billion. The Pentagon had just blacklisted them.

The product nobody saw coming

Claude Code launched as a standalone product in February 2025. By February 2026, it was generating $2.5 billion in annualised revenue — more than double its level at the start of the year. According to the analysis platform Sacra, it is the fastest product ramp in the history of enterprise software.

4% of all public commits on GitHub are now authored by Claude Code, according to analysis by SaaStr, with projections suggesting this could reach 20% by the end of 2026. Boris Cherny, head of product at Anthropic for Claude Code, has said he now writes 100% of his daily code through the tool. Across Anthropic’s engineering teams, between 70% and 90% of all code is reportedly produced by it.

In January 2026, Anthropic launched Cowork, described internally as ‘Claude Code for general computing’to automate white-collar tasks including spreadsheet work, report drafting, and file management. According to SaaStr, the launch triggered a significant sell-off in global software stocks, with the sector losing roughly $2 trillion in market capitalisation as investors repriced the disruption risk to traditional enterprise software vendors.

Anthropic’s enterprise model has always been different to OpenAI’s. Claude has approximately 5% of ChatGPT’s consumer user base but accounts for around 40% of OpenAI’s revenue. The company monetises at roughly $211 per monthly user, compared with OpenAI’s $25 per weekly user, an 8x difference in revenue efficiency. Eight of the Fortune 10 are now Claude customers. Over 500 companies spend more than £1 million annually on Anthropic products.

A paradox, not a clean win

Anthropic is simultaneously the fastest-growing enterprise software business in history and a company politically blacklisted by its largest potential customer. The supply chain risk designation, if it survives legal challenge, could force disentanglement of Claude across Pentagon contractor networks, despite the fact that, as MIT Technology Review noted, Claude is currently the only AI model actively deployed in the military’s classified systems, including in operations in Venezuela. Anthropic has said it will sue if the designation is pursued.

The Pentagon is also in a strange position: it has escalated against the only AI company whose product it was actually using in live classified operations, while simultaneously pressuring that company to accept terms that its own employees and independent legal experts say are weaker than what Anthropic was seeking.

Enterprise AI is no longer a back-office experiment. It is core infrastructure embedded in classified government networks, driving product development timelines, and generating the commercial numbers that are reshaping how investors value the entire software sector. Companies selecting AI vendors now are making decisions with consequences that extend well beyond the immediate use case.

The OpenAI-Anthropic divergence — one company built on consumer scale, the other on enterprise trust and safety positioning — is becoming a genuine differentiator in buying decisions. The QuitGPT campaign is consumer sentiment, not B2B procurement. But it points in the same direction. Brand trust in AI now moves commercial decisions.

Anthropic’s growth was already extraordinary before this week. The Pentagon’s intervention may have accelerated it.

The Next Gen News 2 study, researched and produced by FT Strategies and Knight Lab at Northwestern University and supported by Google News Initiative, was presented at a live panel this week. Its central finding is blunt: the next generation hasn’t disengaged from news. They’ve stopped waiting for publishers to catch up.

More than half of 18 to 25-year-olds across five countries say they consume news at least once a day. Around a third do it more than once. Nearly two-thirds believe news has direct value for their personal and professional lives. These are not the numbers of a generation that has switched off.

That was the central challenge Jeremy Gilbert laid down at the opening of the Next Gen News 2 webinar. The research, which surveyed 5,000 consumers across the United States, United Kingdom, Brazil, Nigeria, and India, is the follow-up to the landmark first study, and it largely confirms that the disengaged-youth narrative is one the industry has found convenient.

Gilbert is direct about what the data shows: “The sense that young news consumers are avoiding news is wrong. We really didn’t see it.” In four of the five markets studied, young consumers were no more likely than older consumers to avoid news. In the United States, the one exception, the gap was a single percentage point, within the margin of error.

The platform shift is real, but incomplete.

Where young audiences consume news has changed dramatically. Instagram, TikTok, and YouTube now far outpace television for 18 to 25-year-olds. YouTube is the third most prominent primary news source overall. But the picture is not entirely binary. Around 40% of young news consumers still access national newspapers and magazines, higher than many publishers might assume.

Search is also strikingly important as a discovery mechanism, particularly when a young consumer wants to go beyond the surface. And perhaps the most counterintuitive finding: human conversation still beats artificial intelligence. Co-workers, friends, and family remain a more significant primary news channel than chatbots, for now.

Scroll, seek, subscribe: the new sifting framework.

Lamberto Lambertini, presenting the updated Modes of Engagement framework, drew a distinction that goes to the heart of how publishers need to rethink content strategy. The research identifies three distinct discovery behaviours:

Scroll — incidental discovery while browsing. The default mode for most young people, low effort, driven by entertainment or fear of missing out.

Seek — intentional departure from passive browsing to actively hunt for information.

Subscribe — direct receipt from trusted, opted-in sources.

Lambertini argues that sifting dominates because young consumers are drowning in content rather than starved for it. They’ve become active engineers of their own information environment, turning on notifications, following specific accounts, and reading comment sections to gauge whether a story is worth pursuing.

The key implication for publishers: winning the first two seconds is everything. Lambertini cited Dave Jorgenson of Local News International, formerly of the Washington Post, as a model for scroll-optimised journalism, always the lead character, always a strong hook, always a relatable anchor question built into the format.

Credibility has inverted.

Gilbert outlined four factors shaping how young consumers judge news: credibility, trust, transparency, and accessibility. On credibility, the inversion is stark. It used to be that being employed by a major masthead was sufficient proof of expertise. That logic has flipped. Today, young consumers often extend credibility to institutions through the individuals who work for them — not the other way around. If they trust the journalist, they extend that to the brand. If they’ve never encountered the journalist, the brand means little to them.

Transparency matters because young audiences have received formal media literacy training, which taught them to look for bias. They want to know how a publication makes money, how its reporting works, who gets included and why. Accessibility matters because institutional voice — once the mark of serious journalism — now reads as a barrier.

From Nigeria: chatbots, volume, and the GEO problem.

The most striking number from the panel came from Nigeria. According to the research, 33% of Nigerian news consumers now use chatbots to find and read news stories. Oluwadunsin Sanya, Editor at Bella Naija, flagged the figure as both familiar and alarming.

Her team is already wrestling with the shift from SEO to what she called generative engine optimisation — figuring out how to ensure Bella Naija surfaces when a reader asks ChatGPT or Gemini a question rather than typing into a search bar. “We were still trying to figure that out when I saw your report,” she said.

Sanya also raised the tension at the heart of most editorial operations: her audience says it feels overwhelmed by the volume of news, but her commercial imperatives push her team to publish more. No one on the panel offered a clean answer. That’s probably the honest position.

From Brazil: the Instagram brand problem.

Tai Nalon, speaking from her experience in Brazil, which the research identifies as having the highest engagement rates for youth journalism globally — described a different kind of identity problem. Her organisation’s audience no longer perceives it primarily as a news operation.

“People see our Instagram profile, and they perceive us as an Instagram profile and not as a news organisation,” she said. Built on social platforms for over a decade, the brand’s identity has fused with those platforms. That creates reach, but it also makes it harder to communicate what distinguishes journalism from content.

Nalon’s solution is transparency about methodology — making the sourcing, reasoning, and reporting authority legible to audiences who’ve grown up online and learned to fact-check themselves. Not because they distrust journalism, but because they want to understand it.

Where this leaves publishers.

The Next Gen News 2 research does not offer a simple blueprint. But it does remove one long-standing excuse. Young people consume news daily. They value context and depth. They’ll go long-form when a story earns it, nearly 70% say they want to dig deep into stories that interest them. They just won’t wait to be told when and how.

News needs to go to audiences, not sit waiting to be found. Formats need to match the platforms, not the style guide. And the journalists doing the work need to be visible, because credibility now travels through people, not brands.

Half of young consumers see value in AI-generated overviews. Only 16% find them extremely helpful. The chatbot shift is coming; in Nigeria, it is arguably already here. The real question is whether they can make themselves discoverable in the environments their audiences actually inhabit.

The Next Gen News 2 report, produced by FT Strategies and Knight Lab at Northwestern University, is available at next-gen-news.com.

The Solow Paradox, named after Nobel Laureate Robert Solow in 1987, observed that “You can see the computer age everywhere but in the productivity statistics”. Despite massive investment in computer technology during the 1970s and 1980s, worker productivity only increased by about 1%. This disconnect between technological investment and productivity offers useful insights for media and creative agencies navigating AI adoption today.

The Evolution of the Creative Industry

The advertising industry has not been slow to adopt new technology, from computers replacing foam boards to the internet enabling digital advertising and social media creating entirely new channels. However, this technological progress coincided with a fundamental shift in business models. Through the 1960s to the early 1980s, independent agencies influenced boardroom decisions, funded by 15% media commissions. By the mid-1980s, pressure from public company acquisitions, client price demands, and diversifying promotion channels forced a shift to time-and-materials models.

This transformation moved the industry from charging for commissions to billing for production time. In the process, this gradually shifted client relationships from boardrooms to operational teams. The internet further disrupted the value chain, with Google and Meta building walled gardens that tool increasing share of advertising profits. Now AI threatens to reinvent not just channels, but the core operating model and business fundamentals of the entire industry.

Understanding the Productivity Paradox

Four decades of academic research have identified micro-level effects and macro-level effects that explain why productivity gains lag technological investment. Micro-level factors include measurement issues, in which productivity metrics fail to capture intangible benefits such as improved quality and employee satisfaction. There are also significant time lags - the 1990s productivity surge reflected 1970s-80s computer investments. Organisations needed time to adapt processes, navigate regulations, and eliminate bottlenecks in non-computerised functions. Additionally, price effects created paradoxes: Increased efficiency drove price competition, which eroded productivity gains even as product quality improved.

Macro-level factors show that the benefits of technology are unevenly distributed. Early adopters capture market share without necessarily growing the overall market, concentrating benefits among a few firms at the expense of adjacent industries. Commoditisation effects mean widespread technologies like cloud computing become utilities—everyone gains access, eliminating competitive advantage. Labour displacement occurs as innovative companies eliminate jobs, pushing workers to less productive organisations and dampening overall productivity. Geographic displacement due to globalisation has shifted production to cheaper labour markets.

The Adoption Lag is Real

Technology adoption has followed predictable patterns, regardless of the technology: internet (1990s), mobile internet (early 2000s), cloud computing (late 2000s), and social media (late 2000s). Each follows an S-curve of early hype, disillusionment, then rapid growth, followed by a plateau or decline. AI will be no different.

In the case of technology adoption in the UK, economic data have mirrored US patterns, with one critical difference: our adoption tends to be slower. Both the UK and the US saw no growth in computing productivity in the 1980s, followed by surges in the 1990s. However, by the 2000s the UK began to lag for several reasons, but one of the largest factors was regulatory constraints. In the case of e-commerce, UK planning regulations aimed at preserving high streets delayed retail e-commerce adoption compared to US companies, which rapidly built integrated online-offline distribution models.

Looking more specifically at marketing technology, we have seen similar adoption lags. Despite continually increasing digital marketing investment, marketing effectiveness declined between 2005 and 2016. That’s a full decade of reduced productivity while marketing teams experimented with a constantly evolving digital landscape. Technology typically takes a decade or more to reach mainstream UK adoption, with the US taking around 5 years due to looser regulations and labour laws, and Europe lagging the UK some time later.

Implications for AI Adoption

Looking at historical patterns, we can anticipate AI’s future trajectory and prepare accordingly. There are three areas to focus on:

Measurement

Productivity shouldn’t be the sole metric of success. The industry’s focus on AI to reduce headcount through cost-cutting is short-sighted. Klarna’s failed attempt to replace customer service agents with AI demonstrated that AI still cannot handle complex, nuanced, or emotional issues. When these skills are needed, people prefer to talk to people.

While early adopters may achieve short-term cost improvements, this inevitably leads to price erosion. The opportunity lies in using AI to deliver higher-quality products and services that sustain price premiums. Agencies must leverage AI alongside humans, not as a replacement, creating space for humanity rather than imitating it. The focus should return to what made advertising valuable: Creativity, human insight, and ideas, not mechanical production.

Adoption Strategy

Despite the hype, AI adoption is struggling. One-third of UK companies claim to have adopted AI, yet MIT research shows 95% of AI projects failed to generate returns. Failures stem from unclear outcomes, poor data quality, skills gaps, overambitious scope, and critically, failure to integrate AI within business processes and systems. This creates an enormous opportunity for agencies through two approaches:

The incrementalist bottom-up approach involves eliminating small tasks or adding new features within existing processes. Success requires a thorough understanding of business processes, identifying granular tasks, and asking whether AI could improve cost, speed, quality, or value. Implementation must be controlled, manageable, and clearly measured—a multi-year continual improvement process.

The transformational top-down approach recognises AI’s broad impact shifts value sources in business processes and models. Businesses that rely on vendors to automate tasks face existential risks. Agencies must regularly assess what business they’re really in and where value originates, considering how AI changes not just their business but also how it impacts customers, suppliers, and competitors. Low regulatory pressure and high task-oriented job concentration make media and creative industries particularly exposed, but they are also positioned to adapt first and share lessons with clients facing less immediate pressure.

Pricing Models

A shift in value demands a change in pricing models, not just price points. Agencies face a double threat: Investing in AI while using it to discount services through reduced billable hours. However, AI’s infinite scalability creates an opportunity to profitably serve previously priced-out buyers. There are two strategies that can counter commoditisation:

• Change pricing models to reflect that AI decouples service charging from the effort required

• Move to different value chain positions, i.e. charge for other services which are more valuable.

The last 20 years have seen marketing increasingly shift toward operational tasks. “Marketing Operations” emerged as the fastest-growing marketing job title post-2005. Salesforce reports that marketers spend 60-70% of their time on tactical execution, while McKinsey shows that CMOs spend 70% of their time on operational management. The issue is that AI threatens the operational tasks that now make up most of marketing’s workload. This means value will shift from operational tasks such as content production, media planning, personalisation, A/B testing, and reporting, towards strategic activities such as brand positioning, customer insights, creative thinking, strategy, and decision-making.

There is time to act, but don’t waste it

Historical technology waves have shown that agencies do have time to adapt to AI, but that time must be spent actively planning for change. This means reimagining business models, identifying new sources of value, redesigning workflows, and establishing new success metrics. Agencies that learn through experimentation will become advisors to clients on similar journeys. However, simply waiting for existing vendor tools to add AI features is profoundly risky. Much of the value in new technology is captured by tech vendors that disrupt others’ business models. Expect companies like Adobe, Google and Salesforce to take value, not enable it. Now is the time to proactively rebalance services, pricing models, and operations before disruption forces reactive crisis management.

Read more