For most of its life, ChatGPT was the one major consumer surface with no advertising on it. That changed in 2026. OpenAI started a US pilot in February, opened a self-serve Ads Manager in May, and by October 2026 had opened self-serve buying in 89 countries, the UK and the EU included. This guide is about the mechanics: how an ad gets matched to a conversation, and why that is different from search. If you want current prices, markets and policy, those change monthly, so we keep them on our ChatGPT Ads page. If you want the setup steps and ad specs, see how to advertise on ChatGPT.
Where do the ads actually appear?
Ads show up as clearly labelled "Sponsored" cards below an AI response, not woven into the answer itself. OpenAI has been firm on that separation: the ad sits beneath the reply and does not influence what the model says. Ads are only shown to logged-in adults on the Free and Go tiers. Plus, Pro and Enterprise users don't see them. That matters for planning: you are reaching the broad consumer base, not the paying power users.
What are context hints?
ChatGPT Ads do not run on keywords. Instead, each ad group carries context hints: plain-language descriptions of the conversations, needs and use cases where your offer is relevant. OpenAI describes them as broad thematic signals, not exact keyword matches. Rather than bidding on the keyword "project management software", you describe the moment: "a small team lead comparing tools to manage client projects and deadlines." You are targeting a situation, not a string of text.
The signal behind this is rich. People tell ChatGPT what they are actually trying to do (compare options, solve a problem, plan a purchase) in full sentences, across several turns. That is closer to genuine intent than a two-word search query.
How does an ad get matched to a conversation?
When a conversation is eligible for an ad, OpenAI's systems compare its topic and intent with advertisers' context hints, then run an auction among the relevant ads. Several things feed that decision:
- The current conversation: the strongest signal, and in Europe the main one.
- Location, device, language and time: country targeting plus iOS app, Android app or web.
- Wider ChatGPT history: used for personalisation where it is enabled. It is switched off in the EEA and Switzerland at launch.
- Your bid and budget: CPM (pay for views) or CPC (pay for clicks).
Some conversations never show ads at all. OpenAI keeps ads away from mental-health and personal-health conversations, and from conversations about prohibited categories like gambling.
How is it different from Google Ads keyword targeting?
Google Ads still starts from the query, even with broad match and AI Max stretching what a keyword can match. ChatGPT Ads start from the conversation. In practice that changes three things:
- No match types or negative keyword lists. You control relevance by how precisely you describe the situation, and by splitting distinct use cases into separate ad groups.
- No search-term report in the usual sense. You optimise on outcomes (CTR, conversions, CPA by ad group) rather than mining queries.
- Intent shows up earlier. Many ChatGPT conversations are research and comparison, so expect a longer path to conversion than a bottom-funnel brand search.
That's why the two platforms complement each other. Our breakdown of how Google Ads works in 2026 covers the search side of the same shift towards intent and AI.
Can you retarget or use your own customer data?
Partly. Since mid-May 2026, advertisers can upload email or phone lists as custom audiences, then include them, exclude them, or adjust bids for them (from 0.1× to 10×). Lists need at least 25,000 matched users, and OpenAI recommends 100,000 or more. Excluding existing customers is often the most useful first step for acquisition budgets. Classic pixel-based website retargeting isn't documented yet, so for re-engaging site visitors the established channels still do the heavy lifting. Our retargeting service covers those.
How do you measure it?
OpenAI ships the tools you would expect from a serious ad platform: a pixel for browser-side conversions and a Conversions API for server-side events, deduplicated with a shared event ID, plus click-through and 1-day view-through attribution. That is the same measurement discipline we apply everywhere: instrument the conversion path properly before you scale, and lean on server-side signals as browser tracking keeps degrading. Without that, a new channel just becomes a new place to lose attribution.
Who is the best early fit?
The strongest early fit is any eligible category where the decision is considered and the conversation is rich, because that is where contextual matching has the most to work with:
- B2B software: buyers compare tools and features in detail.
- Ecommerce and premium consumer goods: people weigh options before buying.
- Travel: multi-step planning conversations.
- Education: long, question-led research journeys.
Some sectors can't take part. Gambling is prohibited in every market, although in our own test we found it already listed in Ads Manager's onboarding dropdown (we shared it on LinkedIn). Financial services and health are generally prohibited outside the US. For those brands the opportunity is organic: being the brand ChatGPT cites in its answers. Our guide to GEO and AEO explains how.
Where does it fit in a media plan?
Treat ChatGPT Ads as a new high-intent surface, not a replacement for anything. ChatGPT still refers a small fraction of the traffic Google does, so it is a test budget today. It lives or dies on the same fundamentals as every other channel: clean measurement, a sharp offer, and a landing experience built to convert.
This is the early, land-grab phase: cheaper attention, less competition, and a format still being defined. If you want to test ChatGPT Ads properly, with tracking and a plan behind it, see our ChatGPT Ads service or run it as part of a coordinated paid-media plan.