ChatGPT Ads
ChatGPT Ads is OpenAI's paid advertising format: a sponsored card with a headline, description, image and link, shown below a ChatGPT answer when the topic of the conversation matches the advertiser's targeting. The card is always labeled Sponsored and never changes the answer itself. Users across 31 European countries have been seeing the ads since August 24, 2026. For campaign setup and management see our ChatGPT Ads agency page.
In short
| What it is | A sponsored card below the ChatGPT answer; for e-commerce, also a recommended product from a feed |
| Who sees the ads | Only users on the free tiers, Free and Go. Paid tiers and under-18 accounts see no ads |
| How targeting works | Contextual signals, i.e. a description of the situation in the conversation, not keywords |
| What it costs | An auction priced per click, with CPM buying and conversion optimization also available. No minimum commitments |
| What it is not | A bought place inside the answer itself. Organic recommendations are the discipline of GEO |
How ChatGPT Ads work
Format and availability
The ad is a small card below the answer: a headline, a short description, a square image and a link to the advertiser's site. For e-commerce, the system can show a recommended product from the ads feed, so the card behaves like a product ad inside a conversation. The format started in the US in February 2026, ads have been shown across 31 European countries since August 24, 2026, and since August 31, 2026 the self-serve OpenAI Ads Manager is open to advertisers across Europe. Sensitive topics such as health or politics stay ad-free.
Targeting through contextual signals
The advertiser does not enter keywords but describes the situation where the offer should appear: who is solving what problem and why this product is the answer. The system then matches the ad semantically against the meaning of the whole conversation. At launch in the EU there is no targeting on chat history or personal profiles, only on the conversation topic, approximate location, device, language and time of day; personalization requires explicit user consent. A signal is a hint, not a rule: it does not guarantee delivery, and there is no report of the conversations that triggered an ad.
Account, measurement and feed
Every advertiser needs their own OpenAI Ads Manager account with business verification. Country, currency and timezone lock after account creation, and the payment card must be issued to the same company the account is verified for. Conversions are measured with a browser pixel and the server-side Conversions API, ideally both at once, deduplicated through a shared event ID. The product feed is built to be compatible with the Google Shopping feed; the key field is is_ads_eligible, plus conversational attributes that get a product recommended for descriptive questions.
From our own practice
We treat ChatGPT Ads as a controlled test: a separate budget, criteria agreed upfront and evaluation over months, not weeks. The auction is new and public benchmarks are unofficial and contradictory, so the value of the first campaigns lies mainly in the data and the cheaper auction, not in guaranteed returns.
We write signals as customer situations and keep one ad group per intent type; otherwise results cannot be read. Paid ChatGPT traffic is separated from organic referrals with UTM parameters from day one, so the two never merge into a single source in analytics. For e-commerce, most of the preparation lives in the feed: it is the same work as conversational attributes in Google Merchant Center, so data prepared once gets used twice.
Common mistakes
- A signal written as a keyword list. The system matches the meaning of a situation, not words. A weak signal means irrelevant delivery.
- Mixed intents in one ad group. Purchase, comparison and research conversations behave differently and cannot be evaluated together.
- An account created with the wrong country or currency. These settings lock, and a mistake means starting over.
- A payment card issued to a different company. A mismatch with the verified entity is the most common reason verification fails. The model where an agency pays and re-invoices media does not work here.
- Judging the channel after two weeks. Or funding it by moving budget out of working performance campaigns instead of a separate test.
Related terms
See also ChatGPT Shopping, agentic commerce, AI Mode, GEO and AI visibility.
Frequently asked questions
When can European companies advertise?
Since August 31, 2026, when OpenAI opened the self-serve Ads Manager across all 31 European countries. Ads have been shown to users since August 24, 2026. You create the account yourself and go through business verification; measurement and feed preparation should run in parallel.
How much do ChatGPT Ads cost?
You pay per click in an auction, typically low single-digit dollars per click; the minimum daily budget depends on the account currency and there are no minimum commitments. Beyond CPC, the platform also offers CPM buying and conversion optimization. OpenAI publishes no official benchmarks; your real cost per conversion will only come from your own test.
Do I need a product feed?
Only e-commerce stores do. Lead-driven businesses use the text card with a link and measure key site actions: forms, inquiries, registrations.
Do ads influence the ChatGPT answer?
No. The card renders separately below the answer and always carries the Sponsored label. You cannot buy your way into the answer itself; organic recommendations are the domain of GEO.
How we can help
We set up the account and verification, deploy measurement, prepare the feed and launch the first campaigns as a controlled test with an honest verdict. Details are on the ChatGPT Ads agency page.