Skip to content
Strategy

Ads in AI answers sit beside the recommendation, so they cannot buy it

ChatGPT and Google now sell ads next to AI answers. What their own documentation says an ad can change, what it cannot, and which buyers ever see one.

Obility Editorial · · 4 min read

The ad and the answer come from different systems

There is now a direct way to put a brand next to a ChatGPT answer: buy an ad. OpenAI’s help center says ad testing started in the United States on 9 February 2026 and is rolling out to eligible Free and Go users in more regions. Google has sold ads around AI Overviews for longer. So founders ask a fair question. If we can pay to appear in the conversation, can we pay to be the recommendation?

Both companies say no in writing, and the way they describe the plumbing explains why. OpenAI’s help center says ads run on separate systems from the chat model and that advertisers have no ability to shape, rank or alter ChatGPT’s responses. Ads appear below the end of a response, labelled as sponsored and visually separated from it. The same page adds a detail most commentary skips: by default, ChatGPT cannot see the ads shown to you. The model writes its shortlist without knowing which sponsored card will sit underneath.

Google matches its ads to an overview that already exists

Google’s arrangement is different in form. Its Google Ads help page on AI Overviews says ads can appear above, below or within an overview, and that ads within one are currently offered in English in twelve countries, including the United States, Canada, Australia and India. To show inside an overview, an ad has to win the normal auction and be relevant both to the search and to the content of the overview itself. Google’s own example is a question about a green swimming pool: the overview suggests testing the water and clearing debris, and an ad for pool vacuums is matched to that.

Notice the order. The overview says something, and the ad is chosen to fit it. Google also says an advertiser cannot target that placement directly and cannot opt out of it, and that Google Ads offers no separate reporting for ads shown inside an overview. Text and Shopping ads from existing Search, Shopping and Performance Max campaigns are eligible. Nothing on the page describes a way to pay for a place in the overview’s own text.

Many B2B buyers will never see the ad

The audience is narrower than the headlines suggest. OpenAI says ads may appear on the Free and Go plans, and that Plus, Pro, Business, Enterprise and Edu accounts will not have ads. Temporary chats show none, and accounts identified as belonging to someone under 18 are excluded. A buyer researching software from a company ChatGPT Business or Enterprise workspace sees the answer and nothing beside it.

For a consumer brand selling to people on free plans, the ad reaches a real audience. For a vendor whose buyers research at work on a paid plan, those buyers are out of reach of the ad entirely, and the only thing they see is what the model chose to say. Category rules narrow it further. Google’s page says ads within AI Overviews do not currently show for sensitive verticals such as finance, healthcare, alcohol, gambling and politics, and OpenAI says its ads are not eligible to appear near conversations about personal health, mental health or politics.

Buy the ad for a different job

None of this makes AI ads pointless. An ad is a labelled offer placed at a moment of intent. OpenAI says it ranks eligible ads on factors that include relevance and advertiser bids, and selects them using signals such as the context of the current conversation, the landing page and the ad copy. That is useful when the answer already names you and you want the next click, or when the answer leaves an obvious gap that a well-matched offer can fill.

What an ad cannot do is fix the reason the model left you out. If the shortlist names three competitors because the pages it searched describe them more clearly, a sponsored card beneath that shortlist can read as the brand the assistant did not pick. OpenAI itself says that seeing an ad does not mean it endorses or recommends the advertiser. The work that changes the answer is the same as before ads existed: pages that answer the buying question, accurate facts about your product, and coverage on the sites the engines cite.

Two limits on all of the above. These are the vendors’ own commitments, and we know of no independent audit that tests them. And OpenAI describes its programme as a test, so formats and rules may change; check the help pages again before planning spend around them.

Count sponsored and organic appearances separately

Ads also add a measurement trap. A session on a free ChatGPT account can now show your brand twice, once in the answer and once in a sponsored unit, or only in the sponsored unit. If a team counts both as visibility, a paid campaign will look like progress in AI search. Google makes the trap harder to spot, since its help page says ads inside an overview are reported as Top Ads, with no line of their own.

Keep the two apart from the first observation. In Obility’s Answer Engine Insights, each observation keeps the prompt wording and its context together with the surrounding answer, the competitors it recommends and the sources it cites. That is the record of organic presence, the number your content work actually moves. Ad delivery belongs in your ad platform’s reporting, judged as paid media on its own terms.

Share this article

Put the workflow into practice.

Bring your questions and content to an Obility demo.

Book a demo