Reading time
6
min.

ChatGPT Ads has been available in France and 31 European markets since 24 August 2026.
The self-service Ads Manager opened a week later, on 31 August. OpenAI claims 1 billion weekly active users, with around 20% of conversations showing commercial intent (OpenAI, 2026).
Enough to catch every advertiser's attention. But between the advertised potential and the reality on the ground, the gap remains significant.
Self-service access opens the door to all advertisers, with two objectives available for now: clicks and reach. The minimum launch budget has disappeared, replaced by a floor of 25 dollars per day.
An OpenAI pixel and a conversions API finally make it possible to start measuring.
Targeting does not work as it does on traditional search engines. There are no keywords and no predefined audiences.
The advertiser describes in natural language the conversation topics in which their ad should appear, and the AI analyses the exchange as a whole to detect purchase intent.
Take a robot vacuum brand. Rather than buying keywords as on Google Ads, it describes the situations in which its product meets a need: a household with a pet looking for a way to deal with hair, or someone who lacks time to clean their home.
This description of context, called a context hint, is the basis for targeting.
Another example: a Paris hotel. A user who asks ChatGPT for a hotel recommendation in the capital may see a sponsored box appear below the answer, at the precise moment they are comparing their options.
These cases remain illustrative.
The channel's logic is still young and real-world uses are stabilising week by week.
The channel is open, but it is not yet mature. Several field reports confirm this.
Le Journal du Net tested the tool with a real advertiser: the campaign launched automatically as soon as account verification was complete, without any explicit approval being requested, and the daily budget was doubled by the platform without any action by the advertiser (Journal du Net, 2026).
No data is provided on the precise delivery area, the queries that generated impressions or performance by creative. A real black box, at this stage.
On top of this come three structural limits that currently weigh on the channel:
The channel works. Fine-grained measurement of its performance is still stabilising.
Another point to watch, which goes beyond targeting alone: OpenAI recently restricted access to its ad platform for some competing AI tools, on the basis of its own commercial interests.
A practice that also exists at Meta and Google, but takes on another dimension at OpenAI, which controls both the advertising inventory and the way a brand can appear organically in the chatbot's answers.
This signal is worth keeping in mind regarding the platform's governance, beyond performance alone.
The question is not whether to go for it, but for whom and under what conditions.
ChatGPT Ads is not recommended in the same way to every client.
Many agencies today position themselves on volume and pure operational delivery of ChatGPT Ads.
At Agence Mantu, a different first step is essential: the Research team carries out a GEO audit. It checks whether a brand is already cited, how it is described and for which queries in ChatGPT, Gemini, Perplexity or Claude.
This snapshot of organic presence makes it possible to distinguish what is already acquired from what is a real advertising opportunity, before spending the first euro. It is the first step of the method, systematic before any activation.
Agence Mantu's legitimacy on B2B and institutional accounts is another advantage.
This is precisely where ChatGPT Ads is least mature today, even though visibility and trust issues remain high.
Finally, a paid campaign does not mechanically improve a brand's organic citation. They are two complementary but independent levers.
Want to know whether ChatGPT Ads makes sense for your brand?
Contact the Agence Mantu media team.