ChatGPT ads in 2026: pay for the answer, or earn it?

The tempting reading of ChatGPT ads for a crypto marketing team is: finally, a way to buy the answer. That reading is wrong, and OpenAI says so itself. What is on sale is a labelled slot under the answer. The sentence that names three exchanges, or three wallets, or three cards is not for sale, and on OpenAI’s own account it is not touched by what you spend.

That does not make the ads useless. It makes them a different product from the one most teams think they are buying, and it moves the real question back to where it was before ads existed: what has the model read about you?

What are ChatGPT ads?

ChatGPT ads are sponsored placements matched to the topic of a conversation and shown to logged-in users on ChatGPT’s Free and Go plans. OpenAI announced plans to test them on 16 January 2026 and began the test in the United States on 9 February 2026.

OpenAI sets out the programme in its post Our approach to advertising and expanding access to ChatGPT and in its help centre article Ads in ChatGPT. Checked on 3 November 2026, they describe five things that matter to a marketer:

The rules are young and OpenAI still describes parts of the programme as a test. Check the pages above on the day you plan a budget, not on the day you read this.

Where does a ChatGPT ad sit relative to the answer?

Below it, labelled, and outside the answer text. The buyer reads the generated answer first, with its named brands and cited sources, and reaches the ad after that.

Can a ChatGPT ad change what ChatGPT says about your brand?

Not according to OpenAI. It states that ads run on separate systems and that advertisers have no ability to shape, rank or alter ChatGPT’s responses. If that holds, the brands named in the answer are decided exactly as they were before ads existed.

Take OpenAI at its word for planning purposes, because the alternative is worse for you. If ads did leak into answers, users would learn to distrust the answers, and the earned placement every serious brand is working for would lose value. Either way, the asset worth building is the one an ad cannot buy.

You can rent the space under the answer. You cannot rent the answer.

What builds the answer next to the ad?

Other people’s pages, in proportion to how many of them discuss you. Across 88,975 citation events from 81 audited crypto brands, citation volume tracks the number of pages that mention a brand at approximately 1.48 citations per page, with r = 0.80.

And those pages are mostly not yours. In our citation study, 7.78% of citations pointed at the audited brand’s own domain. The rest went to review sites, comparison pages, YouTube at 4.49%, Reddit at 2.19%, mid-tier crypto outlets and a long tail of other domains. Tier-2 crypto press, a basket of 21 outlets, was cited 6.34 times more often than the four Tier-1 names (907 events against 143).

So the honest answer to “pay or earn?” starts with what the ad cannot reach. Every one of those pages sits outside the ad system. The full breakdown is in the AI citation study, and the practical route onto those pages is in how to get cited by ChatGPT.

Who actually sees ChatGPT ads?

Adult users on the Free and Go plans, in the markets where the test runs, starting with the United States. OpenAI says Plus, Pro, Business, Enterprise and Education accounts do not see ads.

That cut matters more for crypto B2B than for consumer brands. If you sell custody, payment rails, market making or infrastructure, your buyer is likely to be on a paid or workplace plan, and on OpenAI’s description that buyer never sees the ad. They still see the answer. A retail exchange or card chasing first-time buyers is in a different position, because that audience is far more likely to be on the Free plan.

Paying for ChatGPT ads vs earning the answer: how do they compare?

The ad buys reach to Free and Go users next to an answer; earning the answer buys the recommendation itself, on every plan and every engine, and it keeps working after the spend stops. Most crypto brands need the earned side first, because without it the ad sits under an answer that names a competitor.

Should a crypto brand buy ChatGPT ads?

Only after checking two things: whether your category is eligible at all, and whether the answer above the slot already names you. If the answer names your competitor, an ad underneath it is a poor use of the money.

How do you measure whether the earned answer is working?

Run a fixed set of buyer prompts across engines on a schedule and record, per answer, whether you were named, cited or absent. Ads report their own numbers. The earned side needs a measurement you control.

citeOS runs 20 buyer-style prompts across 5 engines for 100 observation points per audit and publishes every weight on the methodology page. The do-it-yourself version is in AI visibility tracking. Whatever you use, keep the prompts fixed and expect noise: the same question can return a different list an hour later, so judge the trend across weeks, not a single run.

For the difference between the two disciplines more broadly, AEO vs SEO and LLM SEO for crypto brands cover what changes when the reader is a model.

How was this checked?

What else do people ask about ChatGPT ads?

Disclosure. Published by Emergence Media, which operates citeOS. Contact: Sagar Saxena on X. AI answers are non-deterministic, and OpenAI’s ad programme is still described as a test, so details may have changed since the date above. Nothing here is investment advice.