Mobile takes over 75% of UK ad spend on under half of media time - EMARKETER

Mobile accounts for more than three-quarters of UK advertising spend while holding less than half of the time people spend with media, according to Bill Fisher, principal analyst at EMARKETER, who set out the figures in a special episode of ExchangeWire's MadTech Podcast published on November 13, 2026.

The conversation with ExchangeWire chief operating officer Lindsay Rowntree, titled "More Data & AI, Less Clarity: EMARKETER's Bill Fisher on Advertising Effectiveness", runs to about 36 minutes and ranges well beyond devices. Its organising question is one Fisher says he put to a conference audience recently: whether "advertising is getting more intelligent but is it getting more effective". His answer, delivered across eight chapters, is that the industry has grown more sophisticated without settling the questions clients have asked for years, that marketing mix modelling is the best tool available until something better arrives, and that generative AI, for all its back-office value, threatens to push creative work towards what he called a "sea of sameness".

In Short

An EMARKETER analyst told ExchangeWire's podcast that UK advertisers put more than 75% of their spend into mobile even though mobile gets less than half of people's media time, a gap he says social media and Meta are driving. It matters because the same conversation argued that the industry still measures what is easy, such as views and clicks, rather than whether ads produce sales, while AI-made creative risks looking and sounding alike. Fisher expects the next contest to be over trust, with clear AI labelling and human review as the differentiators, and he hopes AI use settles into productive territory in 2027.

Who is talking

Fisher describes himself as an analyst rather than a practitioner. "I'm one of these annoying people that just looks at what everyone's doing, tells them what they're doing wrong, and hopefully helps them do it better," he said in the episode's introduction. He has spent "nearly 30 years" in the industry, 15 of them in media and marketing research, according to his own account. Before EMARKETER he worked at Gartner on the editorial side rather than as an analyst, and before that at Ovum, which he described as a forerunner of Omdia.

Rowntree hosts the MadTech Podcast for ExchangeWire, the ad tech publisher and events company. The YouTube description says the episode builds on Fisher's ATS London panel on trust, standards and regulation. ATS London 2026 is ExchangeWire's own event; PPC Land covered an interview recorded there with Rakuten TV's Stuart Keith on streamer-driven broadcaster consolidation, which ExchangeWireTV published on October 14, 2026.

EMARKETER itself has been active this year. In June, its US AI Advertising Forecast 2026, written by Fisher's colleague Nate Elliott, projected US AI ad spend rising from $32.03 billion in 2026 to $68.25 billion by 2030, with chatbots taking only a small slice. In September the firm bought the research arm of RetailX, adding The UK500 and The CustomerX Index, a deal that deepened its UK coverage.

Measurement is not effectiveness

Asked what mistakes the industry makes with its surplus of data, Fisher went straight to a confusion of terms. The biggest error, in his view, is treating measurement as if it were effectiveness. "The stuff that we can measure ... easily is the stuff that we tend to measure because it's easy," he said, adding that "we don't know if it's effective".

He framed the problem with a slide from his recent presentation. It was not quite a timeline, he said, but it walked through the waves that were each supposed to fix advertising accountability: search, then social, then retail media as a third wave, alongside programmatic buying and attribution. "And every time something comes along, it's like great, let's jump on that, this shiny new thing," Fisher said. The same questions follow each wave. Clients still ask whether their advertising works and how to measure that, and whether the right ad reaches the right person at the right time. "Digital was supposed to resolve this, right?" he said. Yet EMARKETER, he noted with some irony, is still in business answering them.

Rowntree pushed on the premise. Perhaps, she suggested, advertising is not getting more intelligent at all, only more sophisticated. Fisher accepted the distinction. "I'm not saying it's easy, by the way. It's not. It's very difficult because the more sophisticated something gets, the more difficult it is to traverse," he said, before adding: "we're probably not doing it effectively enough, I would argue."

Rowntree added her own complaint about fragmentation. Each new channel, she said, brings its own metric, and each new lens on media, whether effectiveness, media quality or attention, adds more. The "shiny new thing" problem has been raised elsewhere: IAB Tech Lab chief executive Anthony Katsur warned in December 2025 that the industry was chasing "shiny pennies" in agentic AI before fixing transparency, privacy and measurement.

How Fisher's effectiveness chain works

According to Fisher, effectiveness rests on three linked stages, and the difficulty rises at each one.

1. Interaction. The first stage confirms that someone engaged with an ad. The test is simple: "has there been a view or a click, you know, that's easy stuff," he said. This is the part the industry is good at.
2. Incremental impact. The second stage asks what happened because of that exposure. "So, what happened after that click or that view?" Fisher said. He described this incrementality as measurable and the first two stages as linkable, though not easily.
3. Business impact. The third stage asks whether the ad produced a sale or revenue.

Joining all three, Fisher said, is "the golden goose, right? But it is incredibly difficult to do."

That difficulty, he argued, explains the strength of the large platforms. Walled gardens do well, in his account, because "they are good at stitching those three things together. Certainly someone like Amazon which is now all along that journey." The point is consistent with research PPC Land has covered. A study by the vendor Incremental, reported in June, estimated that siloed attribution misses 36% to 53% of total retail media impact. And in July, a joint AppsFlyer and EMARKETER survey of 157 US agency and brand leaders found that 58.6% believed their organisations were underinvesting in channels they cannot measure, with social media and connected TV the most-named blind spots. Measuring what is easy, in other words, shapes where money goes.

The UK numbers: money follows mobile, time does not

The episode's hardest figures came from EMARKETER's newly published UK ad spending benchmarks. "We split it by all manner of things," Fisher said, and one split is by device. On that cut, mobile accounts for what he called "like 75 plus%" of UK ad spend.

Time is another matter. Fisher first put mobile's share of time at "less than 40%", then corrected himself: "Less than 50%. It's in the 40s." Later in the same section he offered a more precise figure: "I think it's 47% or something like that." Either way, mobile's share of spend runs well ahead of its share of attention. "And it's driven by social media," he said.

Rowntree found the direction of the gap striking. A decade or more ago, she recalled, people spent far more time on mobile than advertisers were willing to fund. Fisher agreed that the relationship has reversed, and he offered an explanation for why mobile time falls short of half the day. "So mobile continues to be fairly snackable stuff," he said. Long-form viewing is where the minutes accumulate: television and digital video, the "lean back" experience that the industry once contrasted with "lean forward" phones. Rowntree remarked that she had heard the "lean back" phrase three times in a week after never having heard it before. Fisher's reply was dry. People, he said, "still watch TV despite what people say."

Rowntree then asked about the split between mobile web and in-app inventory, a recurring argument in programmatic circles. Fisher said EMARKETER does not have much data on it for Europe, since it covers the topic more in the US than in Europe, but that clients ask. His characterisation of the problem was blunt. In-app environments, he said, are "like thousands of wall gardens" (walled gardens), while the open web is easier for the ad industry to manage. "We need more insight into that," he said. Rowntree asked whether that could be arranged. "I'll try," Fisher said.

MMM: "the best we've got"

Rowntree observed a sharp rise in interest in marketing mix modelling and asked whether that reflects real suitability or a lack of alternatives. "Probably a bit of both," Fisher said. The method is required, in his view, because of the number of touchpoints and stages in a purchase journey that need to be stitched together. MMM, he said, "is ... the best we've got, I'd say, at the moment until something better comes along."

What would that be? Fisher did not know. "If you know, let me know," he told Rowntree, who described the trend as "a modern take on an old school practice" she had used in her media-buying days. The resurgence is visible in PPC Land's coverage: Google made its open-source Meridian model available to all marketers on January 29, 2025, backed by more than 20 certified measurement partners. In September 2025, IAB and IAB Europe published a one-page framework, "Demystifying Incrementality in Commerce Media", which separates incrementality from attribution and media-attributed return on ad spend.

Rowntree suggested AI could do the stitching. Fisher agreed that this is where the technology fits. "For this aspect, AI is a good thing, right? Because this is what AI is good at," he said, pointing to back-end work: interrogating large volumes of data, finding patterns and connecting them.

An AI sceptic who uses AI every day

Fisher was candid about his own position. "I'm not the biggest fan of AI. I'll put that on record," he said. "However, I love what it does for me." His objection is to the way it is sold. "What I don't like about AI is how it's seen as a catch all for everything and just does everything because I think we need to be really careful about that because it absolutely doesn't."

At EMARKETER, he said, staff are encouraged to use it, and the company holds enterprise licences for "a couple of different platforms". He did not name them. In his own work it functions partly as a search tool. "We do use it a little bit as a glorified search, right?" he said. Thirty years ago, his "deskbased research for a report would have taken an inordinate amount of time compared to what it takes now." The result is more output. "So I produce maybe not twice as many reports as I used to, but more than I ever could."

EMARKETER also uses AI to automate some back-end processes. The firm produces "loads of forecasts across the gamut of media, marketing, e-commerce, right?" and the spreadsheets behind them are unwieldy. Analysts "dump it all into AI, and then we can spot the patterns" that once had to be found by hand.

The point of removing that friction, Fisher argued, is to add a different kind back. "What it allows us to do is put friction back in," he said. AI will find a pattern and explain why it matters, but it is "a people pleaser", so analysts push back and question it. That, he suggested, is how the lesson translates to advertising.

Rowntree raised a cautionary story she had seen in the news "only I think last week". She was unsure of the firm at first, naming PwC before settling on Goldman Sachs, and said headlines had reported that non-technical traders were banned from using AI, citing studies, in her telling, that showed AI had contributed to "brain rot". On closer reading, in her account, the restriction covered AI coding tools such as GitHub and Claude Code, with only IT teams permitted to use them for coding. She did not name the publication that reported it. She also said she knew of companies that had turned off staff access to Claude "because they've just given them access with no mandate about how to use it, what to use it for", and warned that staff using a paid enterprise tool as a search box "makes it a very expensive search engine". That view has some support in the wider data: McKinsey's survey of 1,993 respondents, released in November 2025, found that 88% of organisations used AI regularly but only about one-third had begun scaling it.

Seven, 94% and the sea of sameness

The sharpest warning in the episode concerned creative work, where Fisher said the friction humans supply is needed again. "AI is probabilistic. It gives us the answers that we want and it gives us the creative that it thinks we want," he said.

He offered two data points, both second-hand and identified as such. The first came from a report on AI and creativity by Elliott, EMARKETER's AI specialist, which cited research Fisher thought was by YouGov. Asked to name a number between 1 and 10, people pick seven most often. "33% of the time, people will say seven," Fisher said. But "you ask an AI platform and 94% of the time it will tell you seven because probabilistically that's what humans say."

The second was, in Fisher's own words, "third hand". He attributed it to someone he named as Paul Evans, who had discussed it on another podcast while citing other people's research. Researchers gave two AI engines the same prompt to write a product description for an iPhone case "or something like that". Fisher thought one was DeepSeek and the other OpenAI's. "So, Chinese, American, completely different worlds, right?" he said. "They came back with answers that were 80% the same." Two people in a room would produce similar ideas, he conceded, but not the same words, and "not 80% the same anyway".

"What you risk with creativity is just coming up with this sea of sameness. Vanilla middle black hole of mediocrity. Call it what you will," Fisher said. Rowntree's concern ran in the same direction: AI draws on what already exists. "It's all been created already," she said.

The worry has been circulating. A WARC study commissioned by TikTok, covered in July, found that 88% of 400 marketers reported higher creative volume from generative AI, but only 45% a significant quality gain. In August, PPC Land asked whether Google's automatic move of eligible Search campaigns to AI Max from September 1 meant ads would all look alike. The performance evidence is more mixed than the aesthetic complaint. A Taboola-backed academic study of more than 500 million impressions found AI-made ads performed on a par with human-made ones, but ads that looked AI-made underperformed.

Christmas, Coca-Cola and the suspicion problem

The holiday season gave both speakers their examples. For the past couple of years, Fisher said, Coca-Cola's Christmas advert drew news coverage for its use of AI. "And how boring it was. Still performed okay, but it was really dull." In what he called "year three", he said, "we haven't heard about that yet." The talk instead was of human-made campaigns from John Lewis, Marks & Spencer and Waitrose, the last of which he said he liked. "Humans cause friction and friction is good sometimes," he said. In pitching, "it tends to be the bold ideas that get picked up."

An AI version of Coca-Cola's campaign is still part of this season's mix, however. System1's early 2026 UK Christmas leaderboard, covered by PPC Land on November 12, placed an entry titled "Holidays Are Coming (AI)", described as an AI-created rendition, ninth with 5.7 Stars. M&S ranked third and Waitrose 12th; John Lewis did not appear in the top 12.

Rowntree described the opposite problem: suspicion. She said she saw Amazon's Christmas advert on television a couple of days before recording and "it looked like it was completely AI". She was angry, she said, because brands know audiences react badly to AI-generated people, citing a McDonald's advert in the Netherlands that she recalled was pulled last year after a backlash. She then searched and found the lead actor was a real person, the spot was filmed in New Zealand and it had a director. It might contain AI elements, she said, but was probably not AI-generated. She also objected that it used a song John Lewis had used about ten years earlier. "I should probably get back in my lane," she said. Her conclusion was that the next danger is people "accusing everything of being ... AI slop when actually people have put effort into it."

Survey evidence suggests the industry may overestimate that hostility. FreeWheel's report, covered in September, found that 48% of buyers assumed viewers find AI-made creative off-putting, while only 10% of 2,496 US adults said they disliked the AI ads they had seen.

Labels as the new basis for trust

For Fisher, mandatory disclosure is part of the answer. Requiring brands to "state whether AI has been used", as under the EU's AI Act, is "a good idea possibly", he said. On holiday in France this year, he saw AI disclosures on television advertising throughout. The UK has no such requirement yet, he noted.

The European rules are now in force. The European Commission published the finalised Code of Practice on Transparency of AI-Generated Content on July 20, 2026, and the AI Act's Article 50 duties became binding on August 2, 2026, with fines of up to EUR 15 million or 3% of worldwide turnover. Meta signed the code in late July, and Google added an AI label setting across five advertising platforms, with visible overlays for campaigns aimed at the EU, India or New York.

Fisher tied the issue to news publishing and to his own employer. "We are owned by Axel Springer," he said, which has "just taken over the Telegraph". Axel Springer agreed to buy Telegraph Media Group for £575 million in March, and the UK government cleared the deal on April 14, 2026. EMARKETER shares an office with Business Insider, another Axel Springer title, where Fisher said AI "is being used sensibly and carefully". Business Insider said in May 2025, when it cut about 21% of its staff, that more than 70% of its employees used Enterprise ChatGPT.

The risk, Fisher said, is perception. Readers who see news stories that the Telegraph "is laying off however many hundred people" may start to doubt what it produces. He gave no figure for the reported cuts. Stating that content was produced without AI or by a human, he argued, "is going to be the next differentiating factor in promoting trust".

He drew a line between media. "Trust in digital advertising is declining but trust in TV for example has never been higher I don't think," he said, because television advertising must meet standards and carries legal consequences. He cited no survey for this; the closest recent data PPC Land has covered measured news rather than ads, with VAB finding US households earning $100,000 or more nearly 13 times more likely to trust TV news than social media. "This is going to be the next battleground," Fisher said, and "digital needs to get its head around it quite quickly."

How EMARKETER says it discloses AI use

According to Fisher, EMARKETER's reports carry a disclosure with four elements.

1. A statement that AI was used. "We say AI was used in the production of this report," he said.
2. A description of what it was used for. In his account, "all we use it for is for background research".
3. A statement that a human was involved in the review and editorial stages.
4. A statement of responsibility for the content. "And so we stand by everything in this report."

Self-regulation and fines "not worth the paper"

Rowntree suggested EMARKETER's labelling was an example of the self-regulation the UK advertising industry prides itself on. Fisher was wary. "Self-regulation is dangerous. We know that, don't we?" he said, "especially when we look at certain platforms and the amount of AI slop that we see on them and they're not regulating what's on there."

Europe is "quite good at this", he said, citing the AI Act, the Digital Markets Act and the UK's Competition and Markets Authority. Yet "there are some platforms that just think they sit outside this," and until they are properly regulated, "they will continue to do what they're doing because they make money. Why would they not?" Where penalties do apply, he argued, "often it is cheaper for them to continue to rake in the money from this disingenuous type of advertising and then pay the fines because the fines aren't really worth the paper they're written on." Rowntree added that platforms could pay them off with about a day's worth of advertising revenue.

She described trying to act on that frustration herself. Having read that Instagram allowed users to report a post as undisclosed AI, she came across a post showing new SUVs being crushed by a 10-ton steel coil, including a Tesla Cybertruck and a Volvo, in which a disembodied arm emerged from one vehicle and the Volvo emerged unscathed. Commenters praised Volvo's safety. When she went to report it, she said, there was no option to flag it as AI, only general categories. The exchange closed on a point both speakers accepted: "You'd like to think that there are enough people annoyed by this that it would force the platforms to act, but there's not enough compared to the number of people that buy into it." Detection data points the same way: Resemble AI's 2026 Synthetic Advertising Report, published in October, flagged 3,379 Facebook and Instagram ad files as AI-generated or manipulated and found that only 2 of them mentioned AI in the ad copy.

Past the peak, perhaps

The final chapter turned to 2027. "Everything we write about at the moment has an AI element it seems because it is so transformational," Fisher said. In the past couple of years he had seen "everybody rush to use AI", often without knowing why. "They're just using it because they feel they have to."

What he now sees, across advertising and other industries, is more deliberate use: "not just using it because they can, but looking for a problem first and can AI solve that problem." Sometimes it cannot. Often, with iteration, it can, and then "AI becomes this force for good."

Rowntree asked whether things must peak before people reassess. Fisher reached for the Gartner hype cycle, his former employer's framework. "I think we're probably past the peak inflated expectations," he said. "Hopefully in 2027 we'll be along that ... plateau of productivity perhaps."

His hope for that period returns to the effectiveness chain. He wants "incredibly clever people, much smarter than me" to work out how to stitch together impression, incremental effect and business outcome, with AI doing the joining. Then, he said, some of the questions clients have asked for years might stop arriving, though "not too positive because I don't want to go out of business." The goal is to close the loop on his opening question: "So, we can say yes, advertising has got more intelligent and it's got more effective."

Why this matters for marketers

Three threads in the conversation bear directly on budgets. The first is the UK spend gap: if more than 75% of spend chases less than half of media time, and that spend is concentrated in social platforms, then the measurement Fisher describes as easy, views and clicks, is doing much of the work of justifying it. The second is the measurement gap itself. Fisher's view that only walled gardens currently connect exposure to revenue explains both the pull of retail media and the renewed interest in marketing mix modelling as an independent check. The third is creative risk. If generative tools converge on the most probable answer, distinctiveness becomes harder to buy, at the same moment that EU labelling rules make AI use more visible to audiences.

None of these points is new on its own. The source adds weight: EMARKETER sells research and forecasts to the industry whose measurement Fisher questions, and he acknowledged that clients still ask the questions digital advertising was supposed to answer.

The episode leaves several things open. Fisher did not name the edition or period of the UK benchmarks, and his mobile time figure moved from under 40% to about 47% within a few minutes. The YouGov attribution and the DeepSeek and OpenAI comparison were relayed from memory without study names, sample sizes or methods. Rowntree's Goldman Sachs account named two firms and no publication, and Fisher gave no number for the Telegraph job cuts he mentioned. EMARKETER's AI platforms were not named, and the recording date was not stated.

Key facts

• Episode title: More Data & AI, Less Clarity: EMARKETER's Bill Fisher on Advertising Effectiveness
• Series: The MadTech Podcast, special episode, from ExchangeWire (playlist of 76 episodes on ExchangeWireTV)
• Published: November 13, 2026, on the ExchangeWireTV YouTube channel
• URL: https://www.youtube.com/watch?v=u_Hm_3ErFL0
• Length: about 36 minutes, eight chapters
• Guest: Bill Fisher, principal analyst, EMARKETER; previously Gartner (editorial) and Ovum
• Host: Lindsay Rowntree, chief operating officer, ExchangeWire
• EMARKETER ownership: Axel Springer, according to Fisher
• Related session: the description says the episode builds on Fisher's ATS London panel on trust, standards and regulation; ExchangeWireTV lists an ATS London session titled "The Cost of Playing Safe: Unlocking Revenue Through Trust, Standards and Regulation"
• Core figures cited: mobile "like 75 plus%" of UK ad spend; mobile time "in the 40s", "47% or something like that"
• Second-hand figures cited: seven chosen by 33% of people and by AI 94% of the time; two AI engines' outputs "80% the same"

Timeline

• January 29, 2025: Google makes its open-source Meridian marketing mix model available to all marketers
• May 29, 2025: Axel Springer-owned Business Insider cuts about 21% of staff and says over 70% of employees use Enterprise ChatGPT
• September 9, 2025: IAB and IAB Europe publish "Demystifying Incrementality in Commerce Media"
• November 9, 2025: McKinsey reports 88% of organisations use AI regularly, about one-third scaling it
• 2025 holiday season: A McDonald's AI Christmas advert in the Netherlands is pulled after a backlash, as recalled by Rowntree
• December 29, 2025: IAB Tech Lab's Anthony Katsur warns the industry is chasing "shiny pennies" in agentic AI
• January 28, 2026: Taboola-backed study finds AI ads match human ads unless they look AI-made
• March 6, 2026: Axel Springer agrees to buy Telegraph Media Group for £575 million
• April 14, 2026: UK government clears Axel Springer's Telegraph takeover
• June 4, 2026: EMARKETER's US AI Advertising Forecast 2026 by Nate Elliott projects $68.25 billion in US AI ad spend by 2030
• June 15, 2026: PPC Land reports Incremental's estimate that siloed attribution misses 36% to 53% of retail media impact
• July 9, 2026: Google introduces an AI label setting across five ad platforms
• July 14, 2026: WARC and TikTok find 88% of marketers get more creative volume from AI, 45% a significant quality gain
• July 16, 2026: AppsFlyer and EMARKETER find 58.6% of US marketers underinvest in channels they cannot measure
• July 20, 2026: European Commission publishes the Code of Practice on Transparency of AI-Generated Content
• Late July 2026: Meta signs the EU AI content code
• August 2, 2026: AI Act Article 50 labelling duties become binding
• August 5, 2026: Google tells advertisers eligible campaigns move to AI Max from September 1
• September 5, 2026: PPC Land reports FreeWheel's finding that 48% of buyers assume viewers dislike AI ads, against 10% of viewers who do
• September 24, 2026: EMARKETER acquires RetailX's research arm
• October 14, 2026: ExchangeWireTV publishes an ATS London 2026 interview with Rakuten TV's Stuart Keith
• October 22, 2026: Resemble AI finds only 2 of 3,379 AI-flagged Meta ad files disclosed AI use
• November 12, 2026: System1's early UK Christmas leaderboard ranks Coca-Cola's "Holidays Are Coming (AI)" ninth
• November 13, 2026: ExchangeWire publishes the MadTech Podcast special with Bill Fisher
• 2027: Fisher's hoped-for arrival at the "plateau of productivity" for AI

Related PPC Land coverage

• EMARKETER says US AI ad spend hits $68bn by 2030 - and ChatGPT misses most of it - Nate Elliott's forecast puts most AI ad spend next to AI content rather than inside chatbots.
• EMARKETER acquires RetailX research arm, adds UK500 and CustomerX Index - The September deal that extended EMARKETER's UK research.
• 58.6% of marketers underinvest in channels AI can't measure, AppsFlyer finds - A joint AppsFlyer and EMARKETER survey on measurement blind spots.
• Retail media's hidden ROI: how siloed attribution misses half the picture - Incremental's estimate of what single-retailer attribution leaves out.
• Google opens Meridian marketing mix model - The January 2025 release of Google's open-source MMM.
• IAB unveils incrementality framework for commerce media budgets - IAB and IAB Europe's one-page guide separating incrementality from attribution.
• Only 45% of marketers gain quality from AI, WARC and TikTok find - Generative AI raised creative volume far more than quality.
• AI ads perform as well as humans - if they don't look AI-made - Academic study of 500 million impressions on AI and human creative.
• FreeWheel finds 48% of buyers overrate viewer dislike of AI ads - Buyer assumptions against viewer survey results on AI-made TV ads.
• Supermarkets take 7 of top 12 places in System1's early Christmas ad rankings - The 2026 UK Christmas leaderboard, including Coca-Cola's AI entry.
• EU AI content rules force publishers to label or risk 3% of turnover - The Article 50 guidelines and code that took effect on August 2, 2026.
• Advertisers face mandatory AI ad labels across Google's five platforms - Google's July 2026 AI label setting and regional overlays.
• Only 2 of 3,379 AI-flagged Meta ad files disclosed AI use, Resemble AI finds - Detection data on undisclosed AI in Facebook and Instagram ads.
• UK clears Axel Springer's Telegraph takeover - what happens next? - The April 2026 clearance of the £575 million deal.
• IAB Tech Lab CEO warns industry chasing 'shiny pennies' in agentic AI - Anthony Katsur on fixing transparency and measurement before the next wave.

Summary

Who: Bill Fisher, principal analyst at EMARKETER, interviewed by Lindsay Rowntree, chief operating officer of ExchangeWire, on a special episode of the MadTech Podcast.

What: Fisher said mobile takes more than 75% of UK ad spend while holding less than half of media time, argued that the industry confuses measurement with effectiveness, called marketing mix modelling the best available tool for now, described EMARKETER's own AI use and disclosure practice, and warned that AI-generated creative risks a "sea of sameness", citing second-hand research on AI choosing the number seven 94% of the time.

When: The episode was published on November 13, 2026. Fisher said he hopes AI use reaches a "plateau of productivity" in 2027.

Where: The ExchangeWireTV YouTube channel; the spend figures relate to the UK, and the regulatory discussion covered the EU AI Act, the Digital Markets Act and the UK's Competition and Markets Authority.

Why: Fisher's case is that years of new digital tools have made advertising more sophisticated without proving it more effective, that only walled gardens currently link exposure to revenue, and that trust, backed by clear AI labelling and human review, is becoming the next point of competition between media.