What Are ChatGPT Ads, and What Do They Mean for Marketers?

"ChatGPT ads" is now two questions wearing one phrase. Some people searching it want to know why an advertisement has started appearing under their answers and how to stop it. Others want to know how to buy one.

Both have real answers as of late 2026, and they are worth taking in order, because the first explains the second. OpenAI began testing advertising inside ChatGPT earlier this year, and it has since published both a user-facing explanation of how ads work and an advertiser-facing description of how to buy them. Between those two documents you can get a clear picture without relying on the speculation that filled the gap beforehand.

Does ChatGPT have ads?

Yes — on some plans. OpenAI's help centre is direct about which: "Ads may appear for users on the Free and Go plans. Plus, Pro, Business, Enterprise, and Edu accounts will not have ads."

On where they appear: "Ads can appear below the end of a response. Ads are clearly labeled as sponsored and visually separated from ChatGPT's response."

The rollout is recent and still partial. OpenAI states that "Ad testing started in the United States on February 9, 2026," and that "We're gradually rolling out ads to eligible users on the Free and Go plans in select regions." So if you do not see ads, that may be your plan, your country, or simply that the rollout has not reached your account.

Two exclusions are worth knowing. OpenAI has said it will not show ads in conversations about health, mental health or politics, and will not serve ads to accounts identified as belonging to under-18s.

Why OpenAI started selling ads at all

The stated reason is demand for a free tier that does not cost OpenAI nothing to run. Sam Altman framed it this way when the plan was announced: "It is clear to us that a lot of people want to use a lot of AI and don't want to pay, so we are hopeful a business model like this can work."

That is worth taking at face value, because it explains the shape of the product. Inference is expensive, the free tier is enormous, and subscription conversion has a ceiling. Advertising is the obvious third option, and the design choices — ads only on the non-paying tiers, placed below the answer rather than inside it — follow directly from wanting the revenue without damaging the paid product.

How to tell a sponsored card from an answer

Two signals, both deliberate. The card is labelled as sponsored, and it is visually separated from the response rather than appended to its text. The response itself ends first; the ad sits underneath it as a distinct element.

This is a meaningfully clearer separation than most ad formats on the open web offer, and it is the main reason the "ads are ruining ChatGPT" reaction has been softer than many expected. It is still advertising, and it still occupies attention you did not ask to give — but it is not disguised as part of the answer.

Visibility in AI answers is now a channel of its own. Get a free SEO audit and see where your brand currently stands.

Do ads change the answers ChatGPT gives?

This is the question that matters most, and OpenAI answers it flatly: "No. Ads do not influence ChatGPT's answers. Ads run on separate systems from our chat model."

The advertiser-facing material makes the same commitment from the other direction, listing among its principles that "ChatGPT's answers stay independent."

How much weight to put on that is a judgement call, and it is reasonable to be sceptical of any ad-funded platform's account of its own neutrality. But the architectural claim is specific and falsifiable rather than vague: the ad selection system is described as separate from the model that writes the response, and the ad is placed below a completed answer rather than woven into it. That is a meaningfully different design from sponsored results blended into a ranked list, which is what search advertising normalised.

What data selects the ad

OpenAI lists the signals used. Without personalisation: "What you're discussing in your current chat thread" and "Basic context such as general location or language." With personalisation enabled, it adds "Your current chat thread," "How you interact with ads," and "Past chats and memory." It also notes that "If memory is on, ChatGPT may save and use memories and reference recent chats when selecting an ad."

What advertisers do not get

The limit is stated plainly: "Advertisers never receive your chats, chat history, memories, name, email, precise location, IP address, or sensitive information." On the buying side, the same point appears as "Advertisers receive aggregated performance insights" with no "access to individual conversations."

So the targeting happens on OpenAI's side of the wall. An advertiser picks who they want to reach and sees aggregate results; they do not see the conversation that triggered the placement.

How do you turn ChatGPT ads off?

There are four controls plus two plan-level options. OpenAI lists the controls as: "Turn off ad personalization at any time; Dismiss ads and share feedback; See why you're being shown a particular ad; Clear data used for ads at any time."

Note what the first one does and does not do. Turning off ad *personalisation* reduces the signals used to choose an ad; it does not remove advertising. For that you need one of the plan-level routes, which OpenAI gives as "Upgrade to an ad-free subscription plan (Plus or Pro)" or "Choose a free, no-ads option with additional limits" — that second option trading advertising for a lower daily message allowance.

If you are weighing the paid route purely to escape advertising, the ad question is one line in a larger decision about models and features; our breakdown of ChatGPT Go versus Plus covers where that line falls.

Before you budget for paid placement, find out what you already earn for free. Get a free SEO audit and see which queries you show up for.

How do you advertise on ChatGPT?

There are two routes, and which one applies to you depends mostly on your budget.

The self-serve route. OpenAI has launched what it describes as "a beta self-serve Ads Manager that allows advertisers to sign up and purchase ads directly to appear in ChatGPT." Beta is the operative word: expect a narrower feature set than a mature ad platform and expect it to change.

The partner route. Buying is also available "through partners," and OpenAI names them: Dentsu, Omnicom, Publicis, WPP, Adobe, Criteo, Kargo, Pacvue and StackAdapt. If your media buying already runs through a holding company or one of those platforms, that is likely to be the practical path.

What you can buy

The format is the sponsored card that appears beneath a response. OpenAI says it will "continue to evolve our ad platform with new formats, objectives, and capabilities," which is a clear signal that the single-format picture today is not the end state.

Pricing works on the two models every performance marketer already knows. OpenAI confirms "We'll continue to support CPM and CPC buying" — impressions or clicks.

What the creative needs to do differently

One practical difference from search advertising is worth planning for. A search ad interrupts someone who typed a few words. A ChatGPT sponsored card follows a complete, often lengthy answer that has already addressed the question in detail.

So the reader arriving at your card is better informed than a searcher at the equivalent moment, and generic category messaging is likely to land worse here than it does on search. The copy that works is the copy that adds something the answer did not: a specific offer, a trial, a comparison the model stayed neutral about, a concrete next step.

The corollary is that measurement windows may behave differently too. Someone deep in a research conversation is further along the decision than their first search suggested, which can show up as better conversion rates on lower click volume.

How you measure it

Measurement is more developed than the format list would suggest. OpenAI describes "Conversions API and pixel-based measurement," used to attribute outcomes such as "a purchase, lead, sign-up, or other meaningful action." That is a familiar stack: a pixel on your site, a server-side API for events the pixel cannot see, and conversion reporting against them.

What you do not get is conversation-level data. Reporting is aggregate by design.

When is this worth your budget, and when is it not?

Here is the verdict, use case by use case, rather than a table that refuses to choose.

Worth testing now: considered purchases with a research phase. Software, financial products, B2B services, travel, anything where a person asks an assistant several questions before deciding. The reason is intent quality. A sponsored card under a detailed answer about choosing between two categories of product reaches someone mid-decision, and that is historically the most valuable moment in any channel.

Worth testing now: brands already running large paid search programmes. You have the pixel, the conversion API, the creative process and the measurement discipline. The marginal cost of adding a beta channel is small, and the information you get about an emerging surface is worth something on its own.

Not yet: brands whose audience is on Plus, Pro, Business or Enterprise. This is the structural catch and it is widely underweighted. Ads only serve to Free and Go users. If you sell to engineers, developers, consultants or enterprise teams — the people most likely to be on a paid tier — a large share of your target audience is, by design, unreachable here. Check which tier your buyers are plausibly on before committing budget.

Not yet: anything in the excluded categories. Health, mental health and politics are off the table. If your product sits adjacent to those, expect inconsistent delivery.

Not yet: brands with no measurement in place. A beta ad platform in a new medium is the worst possible place to learn attribution. Get the pixel and conversion events working on an established channel first.

What are the limitations you should plan around?

Four, and the first is the one people miss.

Reach is capped by the business model. Every user who finds the ads irritating enough to act has two documented escape routes — upgrade, or accept fewer daily messages. Both remove them from your addressable audience. The ad tier is, structurally, the segment least willing to pay for AI, which has obvious implications for which products it suits.

It is a beta. Self-serve features, formats, objectives and reporting are all explicitly described as still evolving. Anything you optimise for this quarter may work differently next quarter.

Aggregate-only reporting limits optimisation. No conversation-level data means you cannot see the query context that produced a conversion. You are optimising against outcomes, not against the prompts behind them.

Brand safety controls are largely undocumented. The category exclusions OpenAI has described are protections for users, not adjacency controls for advertisers. There is no published equivalent of the placement or topic exclusion lists that mature ad platforms offer, so you have limited say over the kind of conversation your card appears beneath.

It is not the same thing as being recommended. This is the distinction that matters most for long-term strategy. Buying a sponsored card places you *beside* the answer. Being named *inside* the answer — because the model's sources treat you as a credible option — is a different mechanism entirely, with different economics, and it does not stop when you stop paying.

Paid placement rents attention. Being cited in the answer earns it. Run a free audit and see how your brand appears in AI search today.

How do ChatGPT ads compare with Google Ads?

Three honest differences, and one similarity worth naming.

Audience composition is inverted. Google's paid results reach essentially everyone who searches. ChatGPT ads reach only the subset of users on Free and Go, in regions where the rollout has landed. Scale is not comparable today, and the composition is skewed toward non-payers in a way search traffic is not.

Placement is less entangled with the answer. A Google ad sits in the ranked results, competing for the same attention as the organic listings. A ChatGPT sponsored card sits below a finished answer, labelled and visually separated. Lower intrusiveness, and probably lower click-through as a result.

Targeting is more opaque to you and less exposed to the advertiser. You get aggregate performance, not query-level detail. For advertisers used to search-term reports, that is a real loss of control — and it is deliberate, because the alternative would mean exposing conversations.

One more asymmetry: competitive visibility. In search you can see, more or less, who else is bidding on your terms and what they are saying. In an assistant there is no public results page to inspect, no equivalent of running a query to audit competitors' ad copy. Competitive intelligence in this channel is currently much thinner, and that cuts both ways — your rivals cannot watch you either.

The similarity: the buying mechanics are familiar. CPC and CPM, a pixel, a conversions API, a self-serve manager, agency partners. OpenAI has not invented a new way to transact; it has built a recognisable ad platform on an unfamiliar surface. That is good news for anyone who already runs paid media, because the learning curve is about the medium, not the machinery.

A reasonable way to size the test: treat it as a research budget rather than a performance line. Spend enough to get statistically meaningful conversion data on one product and one audience, hold it for a full buying cycle, and judge it against your existing paid search cost per acquisition rather than against a hoped-for number.

Final Thoughts

The short version: ChatGPT shows sponsored cards below answers to Free and Go users in a widening set of regions, Plus and Pro remain ad-free, personalisation can be switched off but advertising itself cannot be on those plans, and brands can buy placements through a beta self-serve manager or through named agency and platform partners on a CPC or CPM basis with pixel and conversions-API measurement.

The longer version is a strategic point about where the ceiling is. Advertising inside an assistant is a rented position next to the answer, in a channel whose audience is capped by the business model to people who have declined to pay. It is worth testing, particularly if you already run performance media and sell something people research before buying. It is not a substitute for the thing that actually compounds, which is being the source an assistant draws on when it answers without being paid to. If you only have the resources to pursue one of the two this year, our step-by-step checklist for ranking in ChatGPT, Claude and Perplexity is where the durable return is.

One practical closing note: verify the plan and country details before you brief a campaign. This surface has changed repeatedly through 2026 and OpenAI's own help centre is the only source worth trusting on what is live today.

Find out whether AI assistants already mention your brand — and what to fix if they do not. Claim your free SEO audit.