OpenAI's ad rollout puts sponsored links next to ChatGPT's answers. The answer itself - the brand the engine actually names when someone asks a real question - still isn't for sale.
Start with the dated facts, because this space moves fast and old numbers age a page badly.
The format itself: ads are clearly labeled as sponsored and visually separated from the answer, matched to the conversation's topic and the user's interaction history. The rollout has included carousel units showing multiple products from one advertiser, conversion-tracking campaign objectives, and third-party measurement integrations - the standard toolkit of a performance ad product, not an experiment anymore. Users on Free can opt out in exchange for fewer daily messages, or upgrade to a paid tier to remove ads entirely.
OpenAI has been explicit about the boundary. Its own language: "Ads do not influence the answers ChatGPT gives you, and we keep your conversations with ChatGPT private from advertisers." The company says ads run on a separate system from the model, and that advertisers can't shape, rank, or alter what ChatGPT says in response to a question. Advertisers see aggregate performance data - views, clicks - not the conversation itself.
Take that as OpenAI's claim, not an independently audited fact - no outside party has verified the wall between ad serving and answer generation. But it's the claim the product is built and marketed on, and it matches how the format actually looks: the sponsored link sits below the answer, labeled, separate. It isn't woven into the sentence that names a brand. That naming - who ChatGPT actually recommends when someone asks "what's the best X for Y" - is still a retrieval and ranking outcome, the same mechanism it was before any ad shipped.
Two different things now sit on the same screen. Confusing them is the mistake to avoid.
Paid search never killed organic search economics. It clarified them. Once a business could buy the top of the page, being the answer someone finds without paying became a distinct, measurable asset - and businesses that only ever paid for placement never built one. The same split is opening up in AI answers. A sponsored link next to a ChatGPT response is rented shelf space. Being the brand ChatGPT names, unprompted, when a buyer asks a real question is inventory nobody sold you and nobody can repossess when your budget runs out.
The parallel has a limit: AI answers aren't a ranked list of ten blue links, they're usually one or two brands named directly, and there's no query string to reverse-engineer. We've written about that difference in how to navigate AEO - the retrieval mechanics are not SEO mechanics wearing a new label.
If you're weighing a ChatGPT ad budget, the ad economics are a media-buying decision - run the numbers the way you'd run any paid channel test. But there's a question underneath it that a campaign dashboard won't answer: are you in ChatGPT's organic answers today, before you rent the space next to them?
If the answer is yes, ads add reach next to a recommendation you already have. If the answer is no, or you don't know, an ad budget is buying attention around an absence - a user sees your sponsored link, then reads an answer that names someone else. That's a worse outcome than either paying or not paying alone. Knowing which case you're in requires reading actual answers to real buyer questions, not a citation count or a click report. We cover the difference between being cited and being recommended in what Bing's AI citation data can and can't tell you - the same "quoted vs. picked" gap applies here.
Some things aren't settled yet, and a page like this owes you the unknowns as much as the facts.
We measure the organic answer layer specifically: a frozen set of real buyer questions, asked across ChatGPT, Claude, Gemini, and Perplexity every wave, tracking whether your brand is named, where, and against which competitors - independent of anything either you or a competitor spent on ads. Ad spend is a separate line item worth its own analysis; it isn't a substitute for knowing whether the unpaid answer already has you in it. Every count we publish re-derives to a specific question and a specific answer, the same way we'd want claims about ad neutrality to be checked.