A French brand can test ChatGPT ads now, provided the budget is capped, sits outside its profitable lines, and the rule for continuing is written first. It has no reason yet to move an established budget: as of 7 November 2026, no French campaign cost or result was found for this article, and OpenAI says it has no cross-advertiser benchmarks.
The change of frame matters as much as the decision. A ChatGPT ad sits under an answer. It does not buy a place in that answer, and nothing shows it helps get one. These are two investments, to be run and read separately.
The February analysis of ads and trust in AI described a US pilot and a strategic choice. The channel is now open in France, so the question has become one of budget.
What has been open in France since 24 August 2026?
A contextual ad format, self-serve, shown to Free and Go users. OpenAI announced 31 European markets on 18 August, France among them. Serving began on 24 August, first through its own teams and partner agencies, and self-serve followed on 31 August.
| Item | Status as of 7 November 2026 |
|---|---|
| Who sees ads | Free and Go users. Not Plus, Pro or Business, nor accounts identified as minors |
| Who can buy | Advertisers in eligible categories, after company verification, through Ads Manager |
| Selection | Current conversation, approximate location. No past conversations, no memory |
| Targeting | Natural-language context hints, no keywords. Custom audiences blocked for the European Economic Area and Switzerland |
| Buying | CPM, CPC, bidding optimized toward an objective |
| Measurement | Pixel, Conversions API. One-day view-through conversions, reported separately |
| Minimum budget | Sources disagree: none per Blog du Modérateur, €15 a day required in one euro account per Turbine |
Sources: OpenAI, as reported by Blog du Modérateur, E-Commerce Nation, Turbine and TourMaG, September 2026. OpenAI's own pages could not be opened directly for this article.
A contextual ad in ChatGPT is an ad selected from the current conversation, without using the person's history. According to Turbine, context hints are not exact-match controls, not audience rules and not a guarantee of appearing against a topic: you describe a need, and the platform decides where it shows up.
This format is distinct from ChatGPT's product results, which LLM Pulse leaves out of its ad count. Those belong to ChatGPT Shopping.
What does the lack of personalization change for a French advertiser?
It makes results published elsewhere hard to transfer. Outside Europe, an advertiser can upload an audience. In France, that lever is gone, along with retargeting and user memory. A US case study therefore does not quite measure the product you are buying.
This limit may also expire. Sources describe personalization as absent at launch, not ruled out for good. A test run in 2026 describes the 2026 channel.
Audience scope weighs more in B2B. Ads exclude Plus, Pro and Business plans. A logical deduction, to check against your own audience: a decision-maker using ChatGPT on a Business license will not see your ad, whatever answer they read.
Geographic targeting needs checking in your account. TourMaG mentions country targeting. On 1 September, Turbine found only country-level targeting for Lithuania in one European account, and calls OpenAI's documentation inconsistent on finer levels. Do not build a test plan on regional granularity before seeing it in the interface.
What is known about costs and results?
Almost nothing French, and little that is solid elsewhere. Search Engine Journal wrote on 8 September 2026 that OpenAI does not yet have benchmarks across advertisers, industries or campaign types. What circulates comes down to individual reports, published by advertisers or their agencies.
| Advertiser | Audience | CPC | CTR | Reported outcome |
|---|---|---|---|---|
| Common Thread Collective, e-commerce client | High average order value | $4.41 | 0.94% | Estimated ROAS of 3.3 to 6.8 depending on attribution model |
| Workshop Digital, agency | B2B | $3.14 | 1.1% | 739 clicks, no qualified lead |
| Floyd Blaikie | B2B | $9.29 | 0.7% | 336 clicks, 146 organizations identified, 5 in the target |
Sources: Search Engine Journal, 8 September 2026, from LinkedIn posts; Workshop Digital, 26 August 2026. None of these tests is presented as French. Currencies as reported: the third test's budget was in Canadian dollars.
Three readings follow. The first ROAS doubles without a single campaign changing, through the attribution model alone. The second pays 54% less per click than on Google Search ($6.76), yet Google Search brought the only two qualified leads. The agency itself says it is far too early to write off the channel for B2B, and notes that the comparison involves different intent. The third is a reminder that an identified B2B click is not a useful account.
Search Engine Journal sums up CPCs from $3 to $22 from one test to another, the top of the range coming from a test in New Zealand. OpenAI recommends, per E-Commerce Nation and Turbine, starting bids of $3 to $5 per click. A bid recommended by the seller does not tell you what you will pay.
Ad density is just as uncertain. Evertune puts it at 13% across 401,000 prompts (21 May to 8 June 2026), Promptwatch at 32.4% over the last seven days of a period ending 17 August (read through E-Commerce Nation), LLM Pulse at 71.9% across 36,153 US answers (25 July to 4 August), Seer at 77% on about two weeks of prompts tracked for its clients (July). These panels differ in dates and questions, and LLM Pulse acknowledges that its own leans commercial. The same sample shows 7 ads in 8,436 UK answers, where the format had been open since June. An open market is not a full one. No French figure was found.
Where do Gemini, AI Overviews, AI Mode, Copilot and Perplexity stand in Europe?
Among these engines, ChatGPT is the only one for which the sources consulted confirm, in 2026, a purchase possible in France and run separately. The others are closed, inseparable from the rest of the campaign, or unconfirmed.
| Engine | Status as of 7 November 2026 | France and European Union |
|---|---|---|
| Google AI Mode | Ads in a list of countries, in English | No EU country listed |
| Google AI Overviews, inside | Same list, plus Indonesia | No EU country listed |
| Google AI Overviews, above or below | All countries where the feature exists | France not named, covered by the general wording |
| Gemini app | No ads reported in the sources consulted | Nothing specific |
| Microsoft Copilot | Ads through Microsoft Advertising campaigns | French-speaking markets named in March 2025, not reconfirmed in 2026 |
| Perplexity | Dropped advertising in late 2025, saying it undermined trust in its answers | Not applicable |
Sources: Google Ads Help, consulted November 2026; PPC Land, November and January 2026; Shopifreaks, April 2026; Microsoft Advertising, March 2025 and April 2026; TechCrunch, February 2026.
For Google, the useful point is about budget. Google Ads Help says an ad cannot target AI Overviews or AI Mode exclusively. PPC Land adds that existing Search, Shopping and Performance Max campaigns are automatically eligible, with no separate opt-out. So there is no "AI" line to open at Google. A deduction to check in your accounts: your current campaigns may already run around these blocks in France. ADS-UP reported on 30 June that the French launch would have no ad integration. Google launched both features in France on 22 July, and its help page does not detail the French case.
How the two coexist remains poorly known. Xu, Iqbal and Montgomery (Washington University in St. Louis, May 2026) find an ad on 2.16% of US pages carrying an AI Overview, on trending queries. They note that their mostly informational sample underweights commercial queries, where ads are most common.
On Gemini, the message has moved. Dan Taylor, Google's VP of global ads, told Business Insider in January 2026 that there were no plans for ads in the app (reported by PPC Land). Philipp Schindler, chief business officer, said on the late-April earnings call (Shopifreaks) that a format working well in AI Mode could transfer to it, with no timeline.
On Copilot, Microsoft wrote in March 2025 that its ads had fully ramped in English, French and German speaking markets. Its April 2026 post describes AI Max, in open pilot from May across Copilot Search, Copilot Answers and Bing, without mentioning Europe. Check availability in your account before any trade-off.
Should a French brand test now?
Yes, if three conditions are met. Otherwise, waiting has no demonstrated cost: no public data found proves an early-mover advantage, and the $1 billion annualized revenue run rate OpenAI announced on 31 August (per Turbine, unaudited) tells you what the platform collects, not what advertisers earn.
- A decision made in conversation. Considered purchases, comparisons, travel, equipment: the need comes out in sentences. An impulse buy or a B2B niche of a few hundred accounts gives less to work with, given the Free and Go audience.
- A budget you accept losing. TourMaG advises travel agencies to take nothing from their Google Ads budget to fund a test. The principle travels: do not fund uncertainty with what already works.
- Measurement that reaches the sale. Without reliable server-side or CRM conversions, you will only read clicks.
How do you set up and read a small test?
By fixing the decision rule before the first euro, then checking what the volume allows you to conclude.
- Ask one question. "Does ChatGPT bring new customers at an acceptable cost?" Awareness, organic visibility and creative learning are other tests.
- Work out volume before budget. Arithmetic illustration, not a benchmark: €3,000 at €4 a click gives 750 clicks, about 22 conversions at 3%. Chance alone puts that result roughly between 13 and 32, so a cost per conversion between €94 and €231 for identical performance. A test this size rules out a very bad channel or spots a very good one. It cannot separate two close channels.
- Limit contexts. Three to five purchase situations, each with its own ad and a landing page that answers the conversation. More contexts spread already thin volume.
- Reconcile the numbers. Compare reported clicks, sessions in your analytics tool (see measuring AI traffic in GA4) and sales in the CRM. Install the pixel and the Conversions API with a shared event ID, and keep the
{oppref}parameter and UTMs. Keep view-through conversions separate, as the platform does. - Look for what would not have happened without the ad. If regional targeting is unavailable, alternate weeks with and without serving, and track branded search, direct traffic and share of new customers. This design is weaker than a control group, and should be described as such. The logic is set out in proving GEO incrementality.
- Write the verdict in advance. Continue, extend or stop, with a threshold on cost per new customer, compared with your marginal channel rather than your average.
Does paid presence help organic visibility?
Nothing shows it, in either direction. OpenAI says ads do not influence answers, and AirOps notes, on 30 October 2026, that no independent audit has verified this and that no public study measures an effect on citations.
The only data describes overlap. LLM Pulse finds that 2.8% of ads point to a domain cited in the same answer, and that 73% of the 3,033 advertisers observed are never cited in the sample. Seer, across 3,697 ads found on its clients' prompts, sees the advertiser named in the answer 5.4% of the time and cited as a source 3.3% of the time. Mostly US or English-language samples, publishers that sell tracking, and no causality established.
These figures suggest a cautious reading: the brands that pay and the brands that answers cite are, for now, largely different. An ad can provide presence where you are not cited, and nothing more can be expected of it so far. It can also work against you. Across the 2,611 LLM Pulse answers that name a tracked brand and carry an ad, AirOps counts 5.1% with a competitor's ad against 1.7% with the brand's own, on a US sample.
The budget consequence fits in a sentence: keep the two lines separate, with two indicators (see the trade-off between GEO, SEO and paid). An ad test does not justify cutting work on sources, and strong organic visibility says nothing about an ad's return.
What remains to be established as of 7 November 2026?
Five points, where a firm claim should raise suspicion.
- Prices and ad density in France. No public figure found.
- Incrementality. No public controlled test found, in France or elsewhere.
- How long the contextual regime lasts. No date announced for personalization in Europe in the sources consulted.
- The effect on answers. Only OpenAI's word exists so far.
- Transparency. The European Commission designated ChatGPT a very large online search engine on 31 August 2026, with obligations applying from January 2027. Tech Policy Press sees a public ad repository obligation in this, without settling whether the chat itself is covered. If the repository covers the format, you will be able to read your competitors' ads there.
Of the two lines to keep separate, the organic one is the heavier to track: the same questions have to be put to the engines week after week, and your brand read against your competitors. That is the work of Nanga's GEO module: share of voice, position and sentiment for your brand, engine by engine, against the competitors you track. You know where your brand is already recommended before buying a place under the answer.
To see where your brand already appears in AI answers: request a demo.
Frequently asked questions
What is a contextual ad in ChatGPT?
An ad selected from the current conversation and an approximate location, without past conversations or the user's memory. According to Blog du Modérateur and TourMaG, that is the only selection mode available in Europe as of 7 November 2026.
What budget should a first test have?
No French reference exists. Search Engine Journal reports cost per click from $3 to $22 from one test to another, outside France. Set the cap you accept losing instead, then check that the expected conversion volume allows a reading.
Do Plus, Pro or Business subscribers see your ads?
No. According to OpenAI, as reported by E-Commerce Nation, ads go to Free and Go users. Plus, Pro, Business and accounts identified as minors are excluded. A B2B audience on professional licenses is therefore partly out of reach.
Can you buy ads inside AI Overviews in France?
Not inside them. Google Ads Help lists no European Union country for integrated ads. Ads above or below apply to every covered country, without France being named, and cannot be targeted separately.
Does paying improve your citations in answers?
Nothing shows it. OpenAI says ads do not influence answers, with no independent audit according to AirOps. LLM Pulse finds that 2.8% of ads point to a domain cited in the same answer, which describes overlap, not an effect.
Sources:
- Blog du Modérateur, ChatGPT : l'Ads Manager est disponible en libre-service en France, September 2026.
- E-Commerce Nation, ChatGPT Ads en France : ce que les premiers tests révèlent sur la diffusion et le ciblage, September 2026 (reports OpenAI and the Promptwatch study of 18 August 2026).
- TourMaG, ChatGPT Ads est disponible en France : faut-il se lancer ?, September 2026.
- Turbine, a publisher of AI visibility tools, ChatGPT Ads launched in Europe. Here is what actually shipped, September 2026 (reports OpenAI and one observed European account).
- Search Engine Journal, 6 Months Into ChatGPT Ads, Advertisers Still Don't Know What Good Looks Like, September 2026.
- Workshop Digital, ChatGPT Ads: What We Learned From Our First B2B Test, August 2026.
- LLM Pulse, 72% of US ChatGPT Answers Now Carry an Ad, August 2026, updated October 2026.
- Seer Interactive, Our Early View of ChatGPT Ads: 5 Initial Observations, July 2026.
- Evertune, ChatGPT Shows an Ad in 1 in 8 Responses, June 2026.
- AirOps, Do ChatGPT ads hurt organic citations? Early data, October 2026.
- Google, About ads in AI-powered experiences in Google Search, consulted November 2026.
- PPC Land, Advertisers in 10 more countries can't opt out of Google's AI Mode ads, November 2026.
- PPC Land, Why Google won't put ads in Gemini: VP explains the strategy gap, January 2026.
- Shopifreaks, Google says it is open to putting ads in standalone Gemini app after focusing on AI Mode first, April 2026.
- ADS-UP, AI Overviews and AI Mode in France, June 2026.
- Google France, launch post for AI Overviews and AI Mode, July 2026.
- Xu, Iqbal and Montgomery, Measuring Google AI Overviews: Activation, Source Quality, Claim Fidelity, and Publisher Impact, May 2026.
- Microsoft Advertising, Transforming the future of audience engagement, March 2025.
- Microsoft Advertising, Win across all three eras of the web, April 2026.
- TechCrunch, Perplexity's new Computer is another bet that users need many AI models, February 2026.
- European Commission, Commission designates ChatGPT, Reddit, Roblox under Digital Services Act, August 2026.
- Tech Policy Press, What ChatGPT's DSA Designation Means for OpenAI and the EU, September 2026.