“If ChatGPT Has Ads Now, Can’t We Just Pay To Get Recommended?”
ChatGPT getting an ad slot came up at last week’s regular meeting.
The first remark that reached me was this one: put some budget in, and we’ll start showing up in the AI’s recommendations.
The opposite view came up in the same room. If ads are coming in, the AI’s answers can’t be trusted anymore, so why invest in visibility at all?
The provider itself had already answered both.
Can You Buy AI Recommendations With Ad Spend?
No, according to the provider. ChatGPT’s ads sit below the end of the answer, and they come from a separate system from the one that writes the answer.
OpenAI started testing these ads in the US on February 9, 2026.
The Spec, In Short
As OpenAI described it at launch, the test runs in the US and started on February 9, 2026. It covers logged-in adult users, and only on two plans: Free and Go.
On Plus, Pro, Business, Enterprise and Edu, no ads appear at all.
Only one ad unit shows at a time. It always carries a sponsored label and is visually set apart from the organic answer (the part the model wrote). Ads don’t appear in temporary chats, when you’re logged out, or right after an image is generated.
They apparently stay away from sensitive topics too — health, mental health, politics and the like. On the advertiser side, some categories, such as dating, health, financial services and politics, are excluded for now.
All of that is the launch-day description, though. On August 11, OpenAI added that ads had launched in the UK, Mexico, Brazil, Japan and South Korea.
By September 2026 the Help Center described a different setup, too. Ads can appear when you’re logged out, you may see more than one ad unit below a response, and advertisers in health and financial services can qualify if they meet strict criteria (political ads still aren’t allowed).
Where The Line Gets Drawn
Here’s how the provider puts it, in its own words: “Ads run on separate systems from our chat model, and advertisers have no ability to shape, rank, or alter ChatGPT’s responses.”
And one more line: “Ads are paid placements, and seeing an ad doesn’t mean OpenAI endorses or recommends the advertiser or its products or services.”
So being named inside the answer and appearing below it as an ad are, by the provider’s own description, two different mechanisms. And for the first one, it says outright that advertising is not a way to move it.
The ad itself gets picked mainly by matching against the topic of the thread you have open. When several ads are eligible, OpenAI says it weighs relevance together with advertiser bids to decide which comes first. If you’ve turned personalization on, past conversations feed in as well. Advertisers, for their part, can’t see the conversations or anyone’s personal information; according to OpenAI, all they get back are aggregate figures like impressions and clicks.
This Is The Provider’s Own Account
Everything above is what the provider says about its own system. It’s not the result of an outside party checking it.
There’s no published way, at least for now, for anyone outside OpenAI to check the claim at the center of it: “Ads do not influence the answers ChatGPT gives you.”
I still think it matters that this got written into the spec. It makes the boundary explicit: whatever route there is into the answer, advertising isn’t one the provider offers.
So Count Your Appearances In Two Buckets
When your name turns up on an AI screen, you now need to record which kind of appearance it was: named inside the answer, or shown in the ad slot.
If you lump them together, you lose the ability to read your own numbers. Say visibility looks like it dropped in the month you paused ads. Was something really lost, or was the brand never in the answers to begin with? A single combined count can’t tell you.
The Side You Can’t Buy Is The One Worth Investing In
Ads run tomorrow if you pay for them. Your name inside the answer can’t be bought that way. Going by the provider’s explanation, it’s territory that only reflects what you’ve measured and built up.
That asymmetry is, I think, the whole case for investing in AI visibility. Anything you can buy, you solve with money. Anything you can’t buy leaves you with one route: measure where you stand, make yourself easier to cite, and keep building.
Five Caveats On All Of This
None of the above comes without strings, and five of them are worth stating plainly:
- This is the provider’s self-report, not a third-party audit. There’s no public way to check from the outside that ads don’t influence answers.
- It’s a test that started in the US and has since spread to other markets, Japan among them; the presentation and the exclusions have already changed since launch.
- Numbers like how often ads appear, what they cost, or any track record of ads placed so far aren’t on the official pages.
- Ads stay away from sensitive topics like health, mental health and politics. The limits on advertiser categories have been loosened since launch.
- This is a description of the spec, not data on how users behave. No measured visibility figures are in here.
An Extra Ad Slot Leaves The Measuring Work Where It Was
Which ad you see depends on matching against the topic of the thread you have open, plus what advertisers bid. Whether your name comes up inside the answer isn’t tied to that mechanism.
So whatever you decide about running or pausing ads, the levers for visibility inside the answer keep working. Some are things you control on your own side, like the finding that seven to nine in ten sites that blocked AI crawlers still turned up in AI citations. Others are about how you write, like rewriting product descriptions, which moved those products’ share of AI shopping agents’ picks.
With no shortcut through the ad budget, I’d start small: once a month, write down whether your name appeared inside the answer or in the ad slot, and keep the two in separate columns.
Sources
- OpenAI, “Testing ads in ChatGPT,” February 9, 2026, https://openai.com/index/testing-ads-in-chatgpt/
- OpenAI Help Center, “Ads in ChatGPT,” https://help.openai.com/en/articles/20001047-ads-in-chatgpt