Summer hits 2026: five ad tech stories that shape the fall agenda

It was a busy summer for ad tech. Consolidation, valuations under pressure, new ad formats built for AI agents, and a direct challenge to the agency business model. Several moves from the past few weeks already point to the topics that will drive the fall. Open Garden picked the stories worth keeping from summer 2026, each with our take on what it says about where the market is going.

The deal brings together two heavyweights of ad measurement working in different fields. Nielsen handles audience measurement and cross-media deduplication, mostly in TV and CTV. DoubleVerify covers media quality: fraud, brand safety, viewability and, increasingly, attention measurement.

The plan is to combine both layers and give advertisers a single view of who was actually reached and how well the impressions were delivered. As Adweek reports, the deal should also help Nielsen move faster in digital and cross-media measurement. Fragmentation across linear TV, CTV and digital platforms keeps making the advertiser's job harder.

Not everyone is convinced by the industrial logic. Aaron Fetters, a former Nielsen executive, said a similar combination of audience measurement and verification had already been tried when comScore bought AdXpose in 2011, without winning over the market. More to the point, adding DoubleVerify does not solve one of the core problems of cross-media measurement: access to walled garden data and deduplication of audiences across platforms.

The integration looks complicated too. Bringing together audience measurement, cross-media deduplication and ad verification means aligning very different methodologies, data sets and technologies. French audience measurement body Médiamétrie has had trouble upgrading its own systems. That shows how long and how technically delicate a truly unified TV and digital currency can be.

The deal also says something about the finances of independent ad tech. DoubleVerify is the third listed company in the sector to leave Wall Street after Integral Ad Science and LiveRamp, and the trend may only be starting. Digiday notes that lower valuations have made several ad tech companies far more affordable for private equity firms and strategic buyers.

Nielsen is paying $13.60 a share, close to half the $27 of its 2021 IPO. And the pressure is not limited to DoubleVerify: Q2 results cost The Trade Desk and Criteo about a quarter of their value, and AppLovin nearly 20%.

Growth is slowing, but many of these companies remain profitable and cash generative. That profile makes the sector increasingly attractive for more acquisitions and take-private deals.

Our take. The deal is above all a signal about where measurement is heading. Nielsen no longer wants to say only how many people saw a campaign. With DoubleVerify, it can also qualify the conditions in which they saw it: viewability, fraud, brand safety or attention. On paper, that fits neatly with TV/CTV/digital convergence and advertiser demand for cross-media measurement.

The logic carries a risk for DoubleVerify: losing part of its specialization and becoming one component of a much broader measurement offering. Moat shows how a respected specialist can end up as a feature inside a larger whole. Oracle bought it in 2017 and gradually folded it into its ad products. For DoubleVerify, whose value rests on being a trusted third party focused on media quality, that change of status matters.

It also raises the question of Nielsen's culture and its ability to preserve DoubleVerify's ad tech DNA. Nielsen remains closely associated with audience measurement and television. How it absorbs a digital company running on much faster technology and commercial cycles is worth watching. The question is sharper in France, where Nielsen is far less visible in measurement than Médiamétrie, Kantar or Ipsos.

Time is testing a new ad format aimed not at readers but at AI agents and engines. The publisher embeds brand messages directly in the markdown versions of some of its pages, structured files designed to be read by crawlers and AI systems.

These "markdown ads" often take the form of FAQs carrying the advertiser's message and labeled as sponsored. The goal is to let brands buy a presence inside content that AI systems ingest, and potentially influence how they show up in answers. Time told Digiday it is a first for a publisher.

The product arrives as advertisers look for new ways to improve their visibility in ChatGPT, Perplexity and other conversational interfaces. Buyers Digiday spoke with are split. Some see an easy way to answer that demand quickly. Others point to the lack of evidence linking these formats to a better presence in AI answers.

The test has already run into opposition. Less than two weeks after launch, Perplexity told Digiday it had excluded the ads from its index, calling the practice "deceptive." The engine also said publishers using this kind of content risk a lower trust score.

Our take. The problem with Time's approach is less about effectiveness than about its nature. Time embeds sponsored messages in a markdown version built for crawlers and AI agents, and those messages are not necessarily shown the same way to readers. That puts the publisher close to cloaking: serving different content depending on whether the visitor is a human or a machine. That is risky, since search engines have historically penalized this kind of mechanism when it is used to influence their results.

Others are trying a different route. Startup Smalk AI, which we covered here, wants to bring to AI search a logic closer to ad networks such as Teads, Outbrain or Taboola. The model: a network of clearly labeled ad placements, visible to humans and accessible to AI agents, with dedicated infrastructure on the buy and sell sides.

We find that approach healthier, because it keeps advertising and editorial content apart. Rather than quietly altering the material agents come to read in order to influence their answers, it builds real ad inventory around AI search. One big unknown remains: will agents and engines keep recognizing, reading and using those placements?

WPP wants to shift its compensation model toward the results it generates for clients rather than the resources it commits. CEO Cindy Rose said the change is necessary as AI scales up. The old model based on time spent is "probably not sustainable long term," she told Digiday, since AI will let agencies do their work faster and with fewer people.

WPP has already applied the logic internally. It now ties pay for the executives running its largest accounts directly to those clients' growth. The shift is far less advanced on the advertiser side. Jaguar Land Rover is so far the only WPP client on an outcome-based compensation model. Rose said it will take a few more years for that kind of contract to become common, and expects several models to coexist in the meantime.

The topic ties directly into WPP's turnaround plan. The group keeps cutting costs and headcount: 97,400 employees in the first half, down from 105,900 a year earlier, an 8.1% drop. WPP is targeting £500 million in savings over three years. At the same time it keeps investing in WPP Open and its AI tools, with the stated goal of having humans and agents work together to lower the cost of producing its work.

First-half revenue less pass-through costs fell 4.7% to £5 billion, and WPP Media's fell 5.4%. Early signs of stabilization are showing, though WPP is still shrinking. The group now targets a return to organic growth during 2027.

Our take. AI creates a concrete problem for agencies. The more productive they get, the more a model based on hours and headcount hurts their revenue. WPP is trying to decouple its pay from its staffing and tie it to the value it produces. Moving to outcome-based pricing stays hard: the agency and the advertiser have to agree on the result to measure, isolate the agency's contribution, and settle how much risk the agency takes on. With a single client across the whole group, the gap between the economic logic and real adoption is plain.

Starting Aug. 24, the platform counts a view as soon as a video starts playing, whether the user clicked or it autoplayed. Until now, a viewer had to watch for a set number of seconds for it to register.

YouTube had already applied that method to Shorts since 2025. Extending it to long-form video brings its metrics closer to TikTok and Instagram, which also count a view at start. The old metric does not disappear. It becomes the "engaged view" and stays available in YouTube Analytics.

For YouTube, the change makes performance easier to compare across formats and platforms. It will also mechanically inflate the counters under videos, without attention rising in the same proportion. For creators, higher numbers can make a channel look more attractive to brands. For advertisers, they make engaged views, watch time and other attention signals essential.

The change lands as YouTube prepares to tighten access to its revenue-sharing program. Starting February 2027, new channels will need 8,000 watch hours on long-form video, up from 4,000, or 20 million Shorts views, up from 10 million. Shorts creators will then have to keep at least 10 million views every 90 days to keep earning ad revenue.

As Digiday reports, the tightening does not mean ad budgets for Shorts are shrinking. Agencies say investment is growing. But the format still trails TikTok and Reels, which can account for roughly 60% of a short-form social video buy, against close to 10% for Shorts.

Part of the problem is how the inventory is sold. Advertisers rarely buy Shorts on its own. The format is usually bundled into campaigns that also include in-stream and in-feed. Because that traditional inventory is so plentiful, algorithms can push most of the delivery there and leave Shorts with a minority share.

Agencies are starting to isolate Shorts in dedicated line items. They are also applying TikTok and Reels creative codes to it (vertical video, creator content, a more native look) rather than cut-downs of TV spots. That raises an org chart question: YouTube is often handled by video, search or programmatic teams, while TikTok and Reels sit with social media teams.

Our take. By lowering the bar for counting a view while sharply raising the bar for monetization, YouTube is pursuing two separate goals. It makes its public numbers more flattering and easier to compare with TikTok and Instagram, and it reserves revenue sharing for creators who can generate a large and steady audience.

The standardization is good news for advertisers who want to compare platform reach. It does not make views genuinely comparable. A split-second autoplay, an engaged view and several minutes of watch time reflect very different levels of attention. The measurement fight moves from the public counter to deeper retention and engagement data.

The change could also reshape influencer economics. Contracts indexed on raw view counts get less meaningful, since counters can rise without the attentive audience following. Brands have an interest in bringing engaged views into their measurement, capping variable fees, or favoring longer partnerships that reveal whether a creator actually holds an audience.

The new Shorts threshold also favors creators who can publish at industrial pace. More expensive or slower formats may be penalized, while automated channels can flood the zone to hit the volumes. YouTube says it wants to reward engagement better, but its system could end up pushing harder for quantity, including AI-generated content.

For YouTube, the stakes go beyond creator pay. Shorts still has to convince advertisers to move part of their social budgets over from TikTok and Reels. Lower CPMs will not be enough. Until the format proves better effectiveness and can be bought as simply as real social inventory, it stays an extension of YouTube rather than a full competitor to the other two.

Since Aug. 24, OpenAI has been serving ads in ChatGPT in France and 30 other European markets. The ads are limited to the free and Go tiers. Plus and Pro subscribers, along with Enterprise customers, keep an ad-free experience.

ChatGPT says it has close to 900 million weekly active users and estimates that about 20% of interactions carry commercial intent, whether searching for a product, comparing offers or preparing a purchase.

OpenAI is adapting the setup to European rules. Personalized advertising runs on user consent. A user who declines still sees ads, but selection relies only on the current conversation and approximate location, with no use of memory or history. The platform can therefore monetize a large share of its audience through contextual targeting alone.

The offering has changed a lot since the U.S. pilot launched in February. Back then ChatGPT Ads sold at around $60 CPM with a $200,000 minimum commitment. In six months, OpenAI added cost-per-click buying, an Ads Manager, conversion optimization, geotargeting, custom audiences and measurement tools including a pixel and a Conversions API. The platform now says it has onboarded tens of thousands of advertisers.

French access is still gated. Campaigns can be bought through OpenAI's sales teams, the major media agencies, or technology partners such as Adobe, Criteo, Kargo, Pacvue and StackAdapt. Broader self-service access to the Ads Manager comes later. Bouygues Telecom, Carrefour, Cultura and TotalEnergies are among the first French advertisers testing the channel.

Formats are expanding too. The first Chat Cards paired a headline, text, an image and a link. Ecommerce carousels have since been added, showing several products with prices and reviews. The advertiser sends a catalog, and the model picks the product it judges most relevant to the conversation. Feed quality now matters as much as creative or bid: titles, categories, attributes, images, price and availability.

Criteo says it works with more than 2,000 brands on ChatGPT and sees click-through rates two to three times higher than comparable formats. More than 80% of the traffic from those campaigns comes from new customers, according to the company. Those figures come from a partner with a direct interest in the channel's growth, and they say nothing about actual volumes or incremental sales.

Our take. ChatGPT Ads landing in France is not just another inventory source opening up. OpenAI is building a hybrid platform, somewhere between search, commerce media and conversational recommendation. The ad is no longer triggered by a keyword the advertiser bought. The model reads the intent expressed in the conversation, decides whether showing an ad makes sense, and in dynamic formats picks the product to feature.

That shifts a meaningful share of control from the media buyer to OpenAI. Advertisers can give hints about the contexts they want to appear in, but those are not yet the precise targeting criteria or guaranteed negative keywords of Google Ads. They also get less visibility into the signals that lead the model to serve an ad or pick one product over another.

Measurement is the other structural limit. The pixel and the Conversions API can now link a click to a visit, a lead or an order. They establish a sequence of events, not causality. Serve an ad in the middle of a conversation already loaded with intent, and it risks claiming a sale that would have happened anyway. Until OpenAI and its partners publish solid incrementality tests, advertisers will know ChatGPT Ads produces attributed conversions, but not how many additional sales it actually drives.

The French launch opens a learning phase rather than a mature channel. The question for the coming months is whether OpenAI can turn big brands' curiosity budgets into recurring investment, then attract the long tail of small businesses that built Google's and Meta's ad businesses.