What broke the open web in digital advertising, and how to fix it

France's annual digital ad market study, the Observatoire de l'e-pub, has released its 2025 figures. It was a very good year: advertisers spent €12,442 million on digital in France.

The pie has never been bigger. The market has nearly quadrupled in 10 years and is still growing 11% a year.

But the pie has never been split so unevenly:

Still, the open web has real strengths that could help it win back share from the walled gardens (see below). First, though, it has problems to solve.

This piece looks at what has held the open web back, then lays out concrete proposals to revive it.

Advertisers decide where to spend with an ROI-driven approach built on three criteria:

On scale, the open web sits on a gold mine. France's top 500 sites generate about 11 billion page views a month, according to Similarweb. Most of that inventory is on mobile, the personal device people wake up and go to bed with.

Standardization already rules the open web. More than 80% of programmatic display spend runs through the open auction, according to the barometer published by French digital marketing trade body Alliance Digitale.

The bad news is formats and targeting. On both, the open web has fallen far behind the walled gardens.

1. Advertisers are tired of the open web's excesses

The open web has a structural problem. Ad spend tracks the density of bid requests a site sends more closely than its actual traffic.

Some publishers built their business model around advertising. Call them "ad-first publishers." They turned to aggressive tactics to lift revenue per page view:

As they kept losing share to those tactics, other publishers ended up stuck in a prisoner's dilemma.

To limit the damage, many took a step in the same direction. They added ad slots per page and plugged more bidders into their ad servers.

That puts the whole open web at risk of:

In the end, everyone loses. Ad-first publishers come out ahead in the short term, but the strategy is risky over the medium term. Premium publishers are still the festival's headliners. If they ever have to throw in the towel, advertisers will probably shut off the open web tap for good.

Each publisher is free to do what it wants with its inventory, after all. It's the buy side that should worry about practices that degrade the ad experience for brands.

But buyers can't easily audit their open web spend, because few DSPs share placement IDs in their interface.

2. Standard formats are outdated

On the walled garden side, these are the formats that won advertisers over:

That's where the gap with the open web shows.

On display, spend is still concentrated on classic IAB banners, such as 300x250 and 320x50, that haven't changed since the 1990s. Engagement and attention rates, excluding retargeting, are close to zero. CTRs range from 0.03% to 0.25% depending on the device.

On video, open web formats have their strengths. Some publishers worked on share of screen and sticky placements, giving brands a big slice of the screen. But they have one big flaw: few players turn the sound on.

Put yourself in an advertiser's shoes. You want to tell a story. You paid your creative agency in full for a high-quality 30-second video. Would you be happy to learn you bought ads that played your story to your audience... with the sound off? Probably not.

Nobody wins advertisers back with a value proposition that far behind on formats. And when inventory is abundant, buyers overwhelmingly favor the open auction, which takes 80% of programmatic display spend according to Alliance Digitale's barometer. To compete in digital advertising, the open web has to overhaul its standard formats.

3. First-party audiences are high quality but hard to buy

Publishers focus on selling inventory and on fill rates. For them, audiences mainly exist to sell inventory better. Yet the two are different assets that could be monetized separately.

On audiences, there is currently a structural mismatch between what traders want and what publishers sell:

So in the open auction, traders logically turn to intermediaries: the data providers in DSP data marketplaces. They offer massive audience segments usable across most inventory, with drawbacks:

1. Bet on a full-screen standard

Publishers have a simple fix for their format problem. They can sell five-second full-screen placements, the mobile interstitial being the typical example.

The benefits stack up.

For brands:

For users:

So if it's that simple, why isn't anyone doing it?

Because of Google. Several years ago, Google joined the Coalition for Better Ads and decided to penalize sites that use the format in search rankings, just as it does pop-ups.

The logic doesn't hold up:

But Google traffic to publishers is set to fall sharply in the coming months because of LLMs. It may be time for publishers to break free of Google's rule.

2. Standardize ambitious formats and build them into OpenRTB

On their own, IAB banners have little impact. Unified and placed smartly on the page (so no sticky banners everywhere), they improve the ad experience for brands and readers and lift recall.

Publishers could create a new page takeover standard built into the OpenRTB protocol. It would let them offer brands an immersive ad experience, and even win back luxury advertisers.

Another idea: people go to news sites to READ. So why not go all the way on native and offer sponsored articles, with a five-second skip, for instance?

That would give the open web a format the walled gardens don't have. In theory, OpenRTB's native ads module can support this kind of format.

3. The video battle is also a battle for sound

If open web publishers want more video share, they should offer formats that make sense for advertisers. That means good share of screen and sound on. Two examples:

Real mobile-first video players, vertical, with sound on and good share of screen, placed at the heart of the content. But selling pre-rolls or mid-rolls means producing quality video, and that costs real money.

Video interstitials with a five-second skip and sound on by default. They're the easiest to implement, and they would let publishers compete in video with social platforms and YouTube's short formats.

1. Make publisher and advertiser audience segments easier to buy

Publishers should embrace being first-party data providers too. They could push their audience segments transparently into DSP marketplaces, for a percentage fee.

Segments from French sports daily L'Équipe and French tech site 01net might look like this:

"équipe.fr audience, rugby interest – 20% fees"

"01net audience, Samsung smartphone interest – 20% fees"

The benefits would be many. More transparency and fewer intermediaries mean more competitive segments. Publishers could also pool costs through their alliances.

Down the road, AI agents and lookalike algorithms are likely to set up and run campaigns. The Trade Desk's Audience Unlimited initiative already picks segments from the marketplace automatically to build its lookalikes. Transparent, high-quality, competitively priced first-party segments could come out ahead when AI agents choose.

A few guardrails would be needed, starting with the ability to block brands or verticals directly in DSP marketplaces. SSPs already do this for ad inventory, so the model should be easy to copy for audience segments.

Advertisers could do the same. They may have no inventory to sell, but they could still set up their own ad sales arm. It would sell their first-party audiences through curator seats (quick to set up) or data marketplaces (compatible with the open auction). The money could go back into their own digital campaigns. Agencies could sell advertisers support on these projects.

2. Get users to log in

One of Big Tech's great strengths is that users are logged in when they use its services. Logins would be a big plus for publishers. They would make audiences interoperable in an open ecosystem such as programmatic. Imagine being able to tap every open web audience on CTV, for example. That's one more way for publishers to boost audience monetization.

Publishers have so far been reluctant to force logins, because doing so could drive away visitors arriving from search engines or referral links. But AI answer engines are gradually replacing search engines, so that traffic is set to drop sharply in the coming months anyway. Some publishers, including L'Équipe and French tech news site JDN, have already put up a login wall (only in its app, in L'Équipe's case).

1. Win back lost ground: no-consent traffic

Under the GDPR, publishers must get users' consent before tracking their online activity for ad resale. When users are given a choice, consent rates plateau at 60%, according to consent management platform Didomi. Without identifying people, meeting advertisers' needs on targeting, frequency capping and measurement gets much harder.

The good news for publishers: France's data protection regulator, the CNIL, softened its stance and approved the consent wall. Publishers can now ask users either to accept tracking on the site or to pay a reasonable price. Few users pay, and consent rates jump above 85%, according to Didomi. Major French publishers and platforms such as Le Monde, Le Figaro, Les Echos, Le Parisien, Webedia, Prisma and Leboncoin have adopted the consent wall. They could pull the rest of the market along.

2. Win back lost ground: Safari

About 27% of French mobile users browse on Safari, according to Statcounter. Yet Safari accounts for only 10% of mobile web ad spend, according to Alliance Digitale. That's a big gap between where mobile audiences actually are and where the money goes.

You know the reason. Safari is at war with cookies. And whatever people say, cookies remain in 2026 the simplest, cheapest way to target audiences in programmatic.

Yet the industry has been told for years that targeting without cookies is possible, through alternative IDs in particular, such as EUID, First-ID, ID5, Pair ID, Panorama ID, RampID and UtiqID.

Publishers have played along and integrated IDs on their sites. Some IDs appear in 50% of bid requests. The underlying problem is that an ID only really works if it can be activated across the whole programmatic chain:

If publishers sold their audiences directly in marketplaces, as suggested above, those audiences would work with every ID the publisher has deployed. That would strengthen targeting on Safari.

3. Draw a line with LLMs

LLMs set off a technological revolution, and ChatGPT.com is now the fifth most-visited site in France. Publishers are torn between two radically opposite choices:

In our view, AI is no friend of publishers. AI companies enrich their models with publishers' content without paying for it, and they threaten to dry up search traffic. In the U.S., the rollout of AI Overviews cut publisher traffic by 35%.

Publishers should draw a line with AI: entry isn't free. Quality comes at a price.

Then they should focus on growing direct traffic, because recurring audiences are the most valuable to monetize.

France is one of the last countries where Google hasn't rolled out AI Overviews yet. The reason is the legal shield of neighboring rights, the copyright-related rights that entitle French publishers to payment for reuse of their content. Google would face heavy compensation claims if it launched the service. French publishers have a card up their sleeve.

4. Limit social publishing

To offset falling revenue, more and more publishers are betting on social publishing: sponsored posts and stories distributed through their accounts on Meta, YouTube, X and TikTok.

Social publishing has its upsides for publishers. It extends their reach. It also improves media performance, thanks to social platforms' native formats and the credibility a media brand lends to the message.

But the risk is real. What happens the day the walled gardens decide to deprioritize content from professional profiles? It already happened in January 2018. Meta changed its recommendation algorithm to give individuals more visibility, and traffic to company pages fell 28% on average, with drops of up to 70% on some pages, according to Marketing Charts.

Social publishing should be treated as an acquisition play to reach younger audiences who are heavy users of these apps. But publishers should always aim to bring readers back to their own sites and apps.

And why not make publisher sites more community-driven by adding comment sections?

5. Limit reselling

Not all resellers are equal. Some bring real value, such as an exclusive format or exclusive demand. But in most cases, resellers get called on the same impressions as the publisher's direct SSPs and take a commission along the way.

Publishers could:

6. What to do about Google's tools?

Google is an unusual partner for publishers. On one hand, it lets them monetize their inventory with small and midsize businesses. Without Google, much of SMBs' Google Ads spend would probably end up on Meta, which is just as easy to access. But publishers also have a long list of grievances against Google (we'll spare you the full list).

One point gets badly underestimated. By plugging Google's ad tools into their properties, publishers give Google free access to all their audiences. You can test this easily and for free on publishers' in-app inventory, using the PCSA (App Tracking Protection) tool in DuckDuckGo's Android app. Install it and browse your favorite French apps. You'll see Google is very often plugged in and collects plenty of user data along the way. Handy for feeding its own audience segments.

Google's audiences also draw on other sources: Search, YouTube, all of Google's apps and more. But publishers will struggle to monetize their audiences if they also hand them to Google on a silver platter. Google then:

Some publishers, broadcaster video-on-demand (BVOD) and subscription video-on-demand (SVOD) services in particular, chose more specialized ad servers and SSPs that don't play several sides at once. If publishers decide to build a new open web with a strong value proposition, sovereignty should be central to the discussion, and with it the choice of tools.

1. Rethink bid caching

Bid caching is simple. The SSP keeps the last bid response it received from a DSP in cache for a set period, until it finds a similar bid request to serve the impression.

The practice drew heavy criticism at the time because:

Yet bid caching is a clever concept. Programmatic advertising pollutes and costs SSPs and DSPs a lot in infrastructure, which they offset with throttling.

Bid caching could sharply cut the number of calls between SSPs and DSPs and absorb infrastructure costs. It could also boost publishers' SSPs if they get exclusive use of the practice.

That would require changing the OpenRTB protocol to add new fields to bid responses, such as:

2. Get ready to put a value on "human" targeting

Bot traffic is ad tech's elephant in the room. Cybersecurity companies such as Imperva warn that 50% of internet traffic comes from bots. With the boom in AI use, the bot share of traffic is set to soar in the coming months.

Ad tech's message is mostly reassuring, with SSPs and DSPs running strict checks. Yet last year a documentary made waves in the U.S.: Unclickable, by Babis Makridis. A journalist fed up with fraudsters stealing her articles put together a team of experts to find out how hard it was to get into ad fraud. Spoiler: they diverted a fair amount of money over the course of the experiment.

If bots aren't clearly identified, advertisers will have a hard time making sure their spend reaches real, human prospects.

Get ready: "human attention" will be the buzzword KPI of 2027. ;-)

Publishers have a card to play on quality. Much deeper checks would let them assure advertisers that their sites offer the best shot at "human attention." Some walled gardens, such as LinkedIn, already pursue this strategy with ID verification.

Short of that, publishers could take a cue from Leboncoin, which already runs minimum checks on new users to limit fraud on listings.

3. Go further than the TID

The TID (transaction ID) helps fight bid repeating. Duplicated bid requests share a common ID, which makes supply path optimization easier for DSPs. But it does nothing against the second excess of aggressive ad strategies: piling more ad slots onto each page, or ad cluttering.

One fix would be to add a temporary unique ID for each loaded page, alongside the TID. DSPs could then easily calculate each site's level of ad clutter and judge the quality of the opportunity.

4. What about a Spotify for news?

This is our last proposal, and the most ambitious, complex and strategic one.

Here's the starting point.

You may be among the people who pay for Spotify, Deezer, Apple Music or a similar service to listen to your favorite music. If so, would you pay one subscription per record label? Sounds absurd, right?

Yet that's what happens in news. To read Le Monde, Le Figaro, Le Parisien, Les Echos and L'Équipe without limits, you have to pay in each app. People are used to giant audio and video catalogs for €10 to €15 a month, so that fragmentation stands out.

So the question is fair: why not a Spotify for news?

It could come with Netflix-style pricing, for example €6 with a little advertising or €15 ad-free.

A shared app would have several major advantages.

Imagine an app where not just anyone can publish anything. Only articles written by journalists could be published. In a near future that looks very "bots and fake news friendly," that offer would matter a lot to users and brands.

It could become a habit app. Only a handful of the apps installed on smartphones get used every day. One mega content app has a better chance of breaking through with the general public than if every publisher builds its own. It could also create synergies between content creators.

An app is an ecosystem, and you can do pretty much what you want in your own ecosystem. GPT now lets outside developers build mini apps inside it, and WeChat is an all-in-one super app in China.

Articles and ads can be preloaded, which keeps the connection with users when they're offline. Handy on subway and train rides.

Costs get pooled. Dependence on Google Discover drops sharply. Logins are easier to collect on a single app. And since it would be a new app, the publisher shareholders would be completely free to choose the technology. It could be the right time to stop plugging in tech that feeds rival walled gardens' databases directly.

Publishers could use their unsold inventory to promote the super app. France's competition authority would probably have a say, but the hostile climate for news publishers could work in their favor.

There have been shared apps before, such as French digital newsstand Cafeyn. But they mostly offer PDFs rather than every article on site, have no free ad-supported version, and get no self-promotion from the publishers.

That wraps up our proposals. They may not be perfect, and some will likely spark debate over their technical or political feasibility. But at least they're on the table.

Figures from French digital ad sales trade body SRI and from Alliance Digitale put the French open web in a tough spot. But it isn't doomed (and neither are our jobs).

The good news is that despite falling traffic, it still has real potential: billions of page views a month, all connected to programmatic pipes.

Advertisers are savvy investors first. Pragmatism will lead them to back an ambitious project with upside rather than an option meant to limit the damage. That's why rebuilding a supercharged French open web matters so much, taking cues from what worked for our walled garden counterparts.

The music industry made bold choices after seven years of decline caused by streaming and piracy. The open web now needs the same boldness, and the proposals above point that way:

Rethink standard formats, make publisher and advertiser first-party segments easier to buy, win back lost ground, protect against LLMs and walled gardens, innovate, build a shared app. The levers exist. There's plenty of work to do.

But none of it will really work if each player pulls its lever alone. French publishers that don't want to pursue aggressive monetization should collectively standardize immersive formats and make their audiences easier to activate.

The French open web can become a major force in French digital advertising again by 2030, provided the work starts in 2026. It's up to us, the French open web's players, to turn things around and go after those thousands of millions together. ;-)