ChatGPT Ads in France: The Real CPM and CPC Numbers B2B SEA Teams Should Check Before Testing

On August 24, 2026, OpenAI opened ChatGPT Ads to businesses in France and 30 other European markets. MarketEnterprises covered that self-serve launch and what the first attribution tools showed for B2B accounts. Two months on, the number that piece left open is in: what a campaign actually costs. Agencies running real budgets, CyberCité and Gamned! among them, are reporting a cost per thousand impressions (CPM) and a cost per click (CPC) well below OpenAI’s own launch estimates, and a specific kind of advertiser is already acting on the gap.

Two months of real campaigns beat the launch estimates

OpenAI’s own pilot set the tone for how expensive this channel was expected to be: a U.S. launch on February 9, 2026 at $60 CPM, with a $200,000 minimum commitment and no self-serve access at all. Cost-per-click bidding only arrived in May, opening the door to smaller budgets, before the program expanded to France and 30 other European markets on August 24. The commercial material for the French launch pointed to a CPM between €15 and €50 and a CPC between €2 and €12, a range steep enough to make most SEA budgets wait and watch.

The agency Gamned! ran six real campaigns after the French launch and published its findings on September 11. CPM landed at €10 to €12, and effective CPC at €1.40 to €2.00, roughly a third to half of what the forecasts had suggested. One test campaign running a €500 budget saw its CPC drop 75% and its CPM drop 33% over ten days, while click-through rate rose 20%, as the platform’s bidding algorithm sharpened its contextual targeting.

Marjorie Bicaïs, Head of SEA/Média at CyberCité, puts the current market rate at a CPM “around 11 euros” and a CPC “comparable to Google Ads,” with campaigns startable from €15 a day. Three independent reads, CyberCité’s, Gamned!’s, and Bouygues Telecom’s own test below, land in the same narrow band. That convergence across three unrelated agencies is what makes the price credible: a market-clearing figure, not one agency’s best case.

The early advertisers are SEA specialists, not generalists

Bouygues Telecom ran one of the first large French tests, with StratNXT and the Cosmo5 group managing a roughly €40,000 budget across six ad groups, three for its fiber offer and three for the iPhone 18 launch, starting two days after the French opening. Rim El Bekkaoui, media performance director at Bouygues Telecom, said the observed CPM came in well under the $45 to $60 figures that had been circulating, and that the brand’s traffic coming from AI assistants had multiplied sevenfold since the test launched.

Gamned!’s own early advertisers point the same way: specialized recruitment, aerospace engineering roles specifically, and insurance, sectors where a narrow audience and a qualified lead matter more than reach. Florian Jeanin, Head of Buying & Client Solutions at Biggie, Gamned!’s parent group, credits OpenAI’s DSP and data partnerships for opening up a thematic targeting that a broad brand campaign would not use. The accounts seeing a return today share one trait: they already run tight SEA targeting and are hunting a qualified audience that costs more to reach on Google. If your own B2B segment depends on a narrow, intent-driven buyer, that fit matters more than the headline CPM when you size a first test.

Nothing in the data so far restricts the channel to recruitment and insurance by rule. The two sectors stand out because the buyer already searches with precision before contacting a vendor, the same behavior that makes a Google Ads campaign with tight keyword match types work. An SEA team running the same pattern, long, specific, pre-qualified queries, has a closer read on what to expect than a team used to broad-match, high-volume campaigns.

Hands typing on a laptop next to a tea cup and a handwritten notebook, reviewing a test ad campaign budget

The audience is large, but today’s testers are narrow by design

The reach is not the constraint. Bicaïs cites a Médiamétrie figure of 7.8 million daily unique visitors to ChatGPT in France, inside a country where 59% of people reported using generative AI in early 2026, up from 29% a year earlier. Of that group, 55% say they made a purchase after asking an AI for advice first, a habit that barely existed at this scale twelve months ago.

That scale lowers the audience-reach risk. It does nothing yet for the sector risk: the data backing a safe first budget currently exists only for the niches already proving it, recruitment and insurance so far. A large, AI-literate population is the backdrop, not the business case; the business case still has to come from your own sector matching the pattern above.

What the native reporting still will not show you

Baptiste Pechery, Head of Programmatic at Gamned!, is blunt about where the platform still falls short: native reporting stays “rudimentary, limited to clicks, impressions, conversions and costs, without detailing conversations or exact prompts.” Bicaïs adds a second constraint: few ChatGPT users accept the cookie banner, which limits the data flowing into analytics tools and weakens attribution once a click leaves the platform.

Agencies are already layering third-party measurement on top to patch the gap, itself a cost and a setup step most in-house SEA teams have not budgeted for. The same conversion-tracking discipline MarketEnterprises has flagged for Google Ads’ own B2B lead-gen push applies here, arguably more, given how little the native dashboard exposes. A test budget that does not plan for the measurement gap before launch will discover it mid-campaign, once the invoice is already in, which is close to the sequence both Gamned! and Bouygues Telecom had to work around in their first month.

The practical fix sits outside ChatGPT Ads itself: a dedicated landing page, a call-tracking number, or CRM-side tagging that lets you match a lead back to the campaign without relying on the platform’s own cookie. SEA teams already apply that discipline to any platform with thin native attribution. The only change here is timing: it needs to be in place before the first euro spends, not added once the reporting gap shows up in a client review.

A closed laptop next to a blank open notebook and pen on a wooden desk, a workspace set up to plan a test budget

None of this makes a €10 CPM a fixed price. It is the price of a channel still short on advertisers, and every account that follows CyberCité and Gamned! into the auction pushes it up. That shrinking-window pattern is not new: MarketEnterprises saw the same early-adopter compression when Google tightened its Display Network allocation deadline, and it closed faster than most budgets had planned for.

For an SEA team already working a narrow, intent-driven B2B segment, the window to test ChatGPT Ads at today’s cost is open. It is worth entering with a measurement plan already in place, rather than as an afterthought once the native dashboard runs out of answers.

If this changes how you are weighing your next SEA budget, tell us what you are seeing on your own accounts. We are curious.


Photos: Pixabay/StartupStockPhotos.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top