ChatGPT ads: why we advise against them
Partner · Strategy, advertising and measurement
We advise our clients against ChatGPT ads. The problem starts before the campaign: an assistant that answers your question and sells the space under its answer has two customers to satisfy. The product looks like a banner, targets by conversation topic and only reaches the free plans, where your business buyers are rare. The same budget would likely produce more sales elsewhere, starting with your presence in unpaid answers.
- An assistant paid by advertisers serves two customers: the person asking the question and the company paying for the space. That business model is what we question.
- The ad is a sponsored block below the end of the answer: a banner, in a tool people open to get an answer.
- Targeting starts from the conversation topic, with no exact keyword, no customer list and no targeting by company or job title.
- Only the Free and Go plans see ads. Plus, Pro, Business, Enterprise and Edu see none: business buyers are largely out of reach.
- In our B2B test in Canada, ChatGPT ads converted nearly six times less than Google Ads for the same client.
On this page
Sponsored ad in ChatGPT
A sponsored ad in ChatGPT is a paid placement that appears below the end of an answer, clearly marked as sponsored and visually separated from the content produced by the conversation model. OpenAI documents that these ads run on separate systems and that an advertiser has no ability to modify, rank or influence the answers. Selection starts from the topic of the current conversation, and can extend to earlier exchanges if the person has turned on personalization.
Why we advise against them
A business model that leaves us skeptical
A search engine that shows ads is still a results page: the person chooses among links. An assistant gives an answer. When the same company sells the space under that answer, it has two customers to satisfy: the person asking the question and the advertiser paying.
OpenAI documents that ads do not influence the answers and run on separate systems. We have no reason to doubt that today. The problem is elsewhere: the whole product rests on the trust people place in the answer. Every ad placed under it chips away at that trust, and the brand buying the space is tied to it.
A banner format
The ad is a block marked "sponsored" below the end of the answer, with a title, a text and a link. It is a banner, placed in a tool people open to get an answer, not an offer. By the time the ad appears, the person already has what they came for: they read the answer and move on to their next question.
Targeting too broad to aim well
Targeting starts from the conversation topic. The advertiser describes in plain language the conversations where their offer would be useful, and the system chooses, with no guarantee of showing in a specific conversation. There is no exact keyword, no customer list and no targeting by company or job title. For a business that sells to a specific type of buyer, that is too broad.
An audience without decision-makers
Only the Free and Go plans see ads. Executives and teams paying for Plus, Pro, Business or Enterprise see none. OpenAI even presents the absence of ads as a benefit of the paid plans. Our test confirmed it: across two B2B campaigns in Canada, ChatGPT ads converted at 0.9% versus 5.95% on Google Ads, for the same client over the same period. The details are in our test of ChatGPT ads in Canada.
For a low-priced consumer offer, the math can differ: the free-plan audience looks more like the buyers. For a considered purchase, B2B or B2C, we keep the budget elsewhere.
What is actually in place
The test began in the United States on February 9, 2026, with a gradual rollout since. Ads can appear to people on the Free and Go plans. The Plus, Pro, Business, Enterprise and Edu plans are exempt, as are accounts where the person states they are under 18 or is estimated to be a minor. Temporary conversations do not show them either, and during the test, neither does the ChatGPT Atlas browser. For a management team, this line is the first to remember: a business clientele equipped with Business or Enterprise accounts will never see your ads, and the budget committed will not reach those buyers.
The format is restrained: a single block below the end of an answer, which can hold one or several items, from one or several advertisers. Selection starts from the topic of the current conversation. If the person has left personalization on, OpenAI can also use earlier exchanges, memory and past interactions with ads. Anyone can turn off this personalization, clear the advertising data, hide an ad or ask why it appears.
| What the advertiser gets | What it does not get |
|---|---|
| Views, clicks and the conversions measured by its pixel | The conversations, the history, the memory |
| A placement marked and separated from the answer | The name, the email, the IP address, the precise location |
| The messages a person sends it directly | Any sensitive data, including health and politics |
| Targeting by conversation topic | Any influence on the answers |
If you want to test anyway
A new channel attracts by novelty and is judged by measurement. Here, measurement is still thin and the audience is limited: a test gets prepared before the first dollar is spent. These are the questions to settle first.
- Can our conversion tracking cleanly isolate a visit coming from ChatGPT?
- What is our margin on what this channel would sell, and has the break-even point been calculated?
- Do we already appear in ChatGPT's unpaid answers, and have we checked?
- How much are we willing to lose to learn, and over how many weeks?
- If the channel produces nothing after that amount, what do we cut?
A solid answer starts from your conversion tracking and your margin, and treats the first spend as the purchase of information. A weak answer sells the novelty of the channel, promises a quantified return with no baseline data, or talks about privileged access to the platform.
Deciding whether this channel deserves a test budget, and at what amount, is settled on your measurement capacity and your margin. A 90-minute consultation settles it, with a written summary your team can execute.
What you can measure, and what is missing
At launch, advertisers only received an aggregated report of views and clicks. OpenAI now offers a measurement pixel and a Conversions API to track sales and leads after the click. Conversion-optimized campaigns arrived in June 2026. Advertisers still never get the conversations, the history, the memory, the name, the email, the IP address or the precise location.
What Google and Meta have offered for years is still missing: reliable measurement after a simple view, a study of the ads' real contribution (what would have sold without them) and independent measurement. Without clean conversion tracking and a distinct landing page, you will know how many people clicked and nothing more. That is the situation that produces, six months later, an internal debate on the channel's value, then a renewal decision made without a figure. The return logic is the same as everywhere else, and it is detailed in how to calculate real campaign profit.
No ad is eligible near topics of personal health, mental health or politics, and OpenAI does not allow political advertising. Some advertisers in regulated sectors, such as health and financial services, may be eligible if they meet strict criteria. A company in these sectors must check its eligibility before planning a budget, not after approving it.
What we recommend instead
There is a logical order, and it is almost always reversed. Rather than buying a placement in ChatGPT, check whether you already appear in its unpaid answers on the questions that precede a purchase in your sector. The reverse order costs more to learn the same thing.
A company already cited gains consistency: the person sees a recommendation, then a paid placement from the same brand. A company absent from the answers buys visibility that contradicts what the assistant just said. This is not an absolute rule, but it is the most profitable priority call, because the organic work also serves every other surface and is paid for only once.
What stays in-house: the list of questions your customers ask before buying, in their real wording, the margin on what the channel would sell, and confirmation of the sales closed. What can be delegated: the survey of unpaid answers on those questions, the setup of conversion tracking, the creation of dedicated landing pages, the production of the ads and the reading of the results. A company that checks its organic presence before buying reduces the test budget it has to commit by the same amount.
The general framework for visibility in generated answers is laid out in from SEO to GEO, and the content structure that gets a company cited in writing answers engines pick up.
What each platform makes measurable is compared in who measures what, and who measures nothing.
Taking a position in answer engines before your competitors is at the heart of the Strengthen your visibility in AI answers goal.
Already running a marketing team? See how we plug in as reinforcement on paid advertising.
Frequently asked questions about ChatGPT advertising
Are ChatGPT ads worth it for a business?
For most businesses, we advise against them. Targeting is broad, the audience is limited to the free plans and executives are largely absent from it. In our B2B test in Canada, ChatGPT ads converted nearly six times less than Google Ads for the same client. A low-priced consumer offer can be an exception.
Can you buy ads in ChatGPT from Canada?
Yes. OpenAI states that its ads manager is available in Australia, Canada, New Zealand, the United Kingdom and the United States, with Japan and Korea announced. Availability may change as the test expands.
Do ads influence ChatGPT's answers?
No, according to OpenAI's documentation. The ads run on systems separate from the conversation model, and an advertiser has no ability to shape, rank or modify the answers. They appear below the end of an answer, marked as sponsored and visually separated.
Who sees these ads?
People on the Free and Go plans. The Plus, Pro, Business, Enterprise and Edu plans show none, nor do the accounts of minors, whether declared or estimated as such, nor temporary conversations. A significant share of a business clientele is therefore out of reach of this channel, and that is the first criterion to check before committing a spend.
What data does an advertiser receive?
Views, clicks and, if you install its pixel or its Conversions API, the conversions measured on your site. OpenAI specifies that advertisers never obtain the conversations, the history, the memory, the name, the email, the IP address or the precise location.
- OpenAI, help center, Ads in ChatGPT, official documentation, accessed July 2026.
- OpenAI, help center, Ads Manager Availability, official documentation, accessed July 2026.
- OpenAI, Testing ads in ChatGPT, official announcement, updated March 26, 2026.
- OpenAI, developer documentation, Measurement Pixel, official documentation, accessed September 2026.
Gabriel almost always takes your first call and carries out your audit. He builds the strategy starting from your growth goal: where to put your budget, which market to test and how to connect each lead to a real sale in your CRM. He mainly leads engagements for three goals: Optimize the profitability of your digital campaigns, Develop a new market, and Generate demand and growth. With Geneviève, he also works on organic search (SEO), AI visibility (GEO) and conversion rate optimization (CRO). The sales a Google Ads or Meta Ads campaign brings in depend on the page that receives the click. He writes mainly about marketing strategy, paid advertising and measurement.
About Falia →
SEO
Paid advertising
Conversion
AI visibility
Web design