TL;DR
- The ad renders below the ChatGPT answer and never changes it.
- Allowed categories are narrow, and self-serve is blocked in most countries.
- CPC campaigns open at $3-5 per click.
- OpenAI can now refuse ads that conflict with its competitive position.
- Being cited inside the answer stays organic and has no price.
ChatGPT Ads reached a $1 billion annualized revenue run rate inside 200 days and now sells self-serve in 52 countries. OpenAI started testing sponsored placements below ChatGPT answers in February 2026, opened the Ads Manager in May, added 31 European markets in August, and extended self-serve buying to India, Europe, the Middle East, and North Africa on August 31, 2026. Ads reach logged-in adults on the Free and Go plans only. Plus, Pro, Business, Enterprise, and Education stay ad-free.
Definition: ChatGPT ads are sponsored placements that OpenAI shows below the answer in ChatGPT conversations. They appear only for logged-in adults on the Free and Go plans, carry a clear sponsored label, and stay visually separate from the organic answer. OpenAI matches ads to the topic of the conversation, and advertisers never see the chats themselves.
Where the rollout stands in September 2026
OpenAI moved fast on this. The company published its ads principles in mid January 2026 and had live tests running in the US within a month. By the end of August it was describing ChatGPT Ads as a global platform carrying tens of thousands of advertisers. Here is the timeline, drawn from OpenAI’s own announcements.
- January 16, 2026: OpenAI published its ads principles and confirmed US testing would start within weeks. ChatGPT Go launched in the US the same day at $8 per month, after reaching 171 countries since August 2025.
- February 9, 2026: testing began in the US for logged-in adults on the Free and Go tiers.
- March 26, 2026: OpenAI reported no measured impact on consumer trust and low ad dismissal rates, then expanded the pilot to Canada, Australia, and New Zealand.
- May 5, 2026: a beta self-serve Ads Manager opened, along with CPC bidding and conversion measurement.
- May 7, 2026: OpenAI said pilots in the UK, Mexico, Brazil, Japan, and South Korea would begin in the coming weeks.
- August 11, 2026: OpenAI said ChatGPT Ads had launched in the UK, Mexico, Brazil, Japan, and South Korea, and that it would keep expanding to more markets this year.
- August 18, 2026: OpenAI announced that ChatGPT Ads would reach 31 European markets, its largest expansion to that point, and confirmed it had added conversion optimization, geo-targeting, and custom audiences.
- August 31, 2026: OpenAI said the business had reached $1 billion in annualized revenue run rate in under 200 days, and opened self-serve buying across India, Europe, the Middle East, and North Africa.
- September 10, 2026: version 1.6 of the ad policies added OpenAI’s right to decline ads that conflict with its business interests or competitive position.
The business logic is simple. Sam Altman put it plainly when the plan was announced: “a lot of people want to use a lot of AI and don’t want to pay.” CNBC reported that OpenAI was on track for a $20 billion annualized revenue run rate in 2025 while carrying enormous infrastructure commitments. Ads fund the free tier, which OpenAI says keeps ChatGPT available to more than 1 billion weekly active users. Seven months in, that ad business was running at a $1 billion annualized rate.
What the ads look like
OpenAI’s help center documentation describes one format. The ad renders below a relevant conversation and contains six parts: advertiser name, favicon logo, headline, description copy, an image, and a landing page link. It is labeled as sponsored and kept visually separate from the answer.
Two exclusions matter. Accounts that are, or appear to be, under 18 never see ads. Ads are also kept out of sensitive conversations, which OpenAI’s placement policy defines as mental and personal health chats, political content, and a list of brand-unsafe contexts running from graphic violence to misinformation. One nuance catches people out. Since April 2026 medical, legal, and financial advice conversations are no longer categorically blocked. The block applies to the sensitive context, not to the whole subject.
OpenAI has previewed where this goes next. Its January announcement showed mockups of conversational units where a user can ask the ad follow-up questions before deciding. Those are concepts for now. Nothing inside the answer itself carries payment, and OpenAI has said it plans more formats, objectives, and buying models over time.
Brand safety and advertiser vetting
OpenAI’s stated policy is to place ads only near chats that are safe, appropriate, and aligned with user trust. Review runs at three levels: the advertiser, the creative and its landing page, then the placement. For brands, that vetting cuts both ways: cleaner inventory, slower access.
The category list is the part most brands miss, and it is the first thing to check. OpenAI’s ad content policy still limits advertising to a narrow set of consumer verticals: household and consumer goods, local services, travel and entertainment, and digital products and education. Financial services, healthcare and medicine, and legal services are reviewed case by case, and only in the US. Everything else is disallowed for now. If you sell gambling, alcohol, tobacco, recreational drugs, or adult products, or you advertise individual housing or job listings, you cannot buy these ads at any price today.
Targeting and what happens with user data
Ad selection runs on conversational context. OpenAI matches ads using the topic and intent of the current chat, the ad’s landing page and copy, and, when personalization is on, signals from the user’s broader ChatGPT history. That last part means ChatGPT’s memory shapes what each user sees, in ads as well as answers.
Advertisers steer this with context hints at the ad group level. Hints describe the conversations, topics, or keywords where a product fits. They guide matching, they do not behave like exact-match keywords, and they guarantee nothing. Writing good hints takes knowing how buyers actually phrase things, which is why we treat prompt language as its own research job.
On data, OpenAI’s commitments are specific. Advertisers never see chats, chat history, memories, or personal details. They receive aggregate performance data only, such as views and clicks. OpenAI says it never sells conversation data to advertisers. Users can turn personalization off, clear their ad data, dismiss any ad, and see why a given ad appeared. Free users can even opt out of ads entirely, in exchange for lower usage limits and reduced feature access: fewer messages, and no access to some tools such as image generation and deep research.
Pricing, buying, and measurement
Three buying models exist today. Reach campaigns pay per thousand impressions (CPM). Clicks campaigns pay per valid click (CPC), added in May 2026. Conversion optimization arrived by August, and OpenAI says CPC and outcome-optimized bidding now account for most campaigns. Delivery runs through a relevance-weighted second-price auction. For CPC campaigns, OpenAI’s help center recommends a starting maximum bid of $3-5 per click.
Access comes through two doors. The Ads Manager at ads.openai.com handles registration, budgets, bids, pacing, creative upload, and reporting, and OpenAI still labels it a beta product. Self-serve access is gated by country, and it is the billing entity’s country that counts, not the audience you want to reach. Fifty-two countries are on the list. Israel, China, and every country in Southeast Asia are not. The second door is partners: OpenAI named Dentsu, Omnicom, Publicis, and WPP on the agency side, plus Adobe, Criteo, Kargo, Pacvue, and StackAdapt on the technology side, and says the ecosystem has since passed 50 technology and measurement partners.
Measurement covers impressions, clicks, spend, CTR, average CPC, average CPM, and conversions. The OpenAI Pixel and Conversions API handle post-click actions, and the company now calls both foundations of the platform rather than recent additions. UTM parameters persist on ad clicks, so tagged traffic lands in the same reports as your organic ChatGPT referrals in GA4 and sits beside the AI search metrics worth tracking. OpenAI does not publish its attribution window.
What OpenAI has not published: rate cards, CPM floors, category benchmarks, or auction win rates. Every specific CPM figure circulating right now comes from third parties. We would not build a budget model on numbers OpenAI has not confirmed.
The ad sits below the answer, and the answer still decides
OpenAI’s first stated principle is answer independence. Ads never change what ChatGPT says. In OpenAI’s words, ads “run on separate systems from our chat model, and advertisers have no ability to shape, rank, or alter ChatGPT’s responses.” The recommendation inside the answer comes from the model and the sources it cites. Money buys the slot underneath, and only that.
There is a second question OpenAI has not answered. It says advertising does not influence the answer. It has not said whether advertising influences which sources the answer cites. Those are different claims, and OpenAI has published no statement either way, not on ads.openai.com, not in the help center, and not in any of its ads announcements. Until the company addresses it directly, treat citation independence as undocumented rather than guaranteed.
Most coverage misses the strategic point underneath all of this. When a user asks “what’s the best CRM for a small agency,” ChatGPT names brands before any ad renders. A paid slot below the answer does not replace a citation inside it. The work of getting your brand picked as a source in ChatGPT answers stays organic, and that position has no listed price.
The pattern matches what we see across AI surfaces: most of the influence happens before anyone clicks. That argues for a strategy built for answers without clicks, and for measuring your share of AI answers the way you once measured rankings.
There is a second-order effect to watch. On buying-intent prompts, the ad below the answer now competes with the cited sources for whatever clicks remain. Referral patterns on commercial topics will shift as the platform scales. Watch your own data before assuming anything about how much.
ChatGPT ads vs Google AI Mode ads vs Perplexity
The three big conversational surfaces have taken three different paths. Google is testing sponsored formats in AI Mode, starting with retail and expanding into travel. Perplexity went the other way: it shut down its ad tests in February 2026, saying labeled ads can erode trust, despite reporting 780 million monthly queries.
| ChatGPT ads | Google AI Mode ads | Perplexity | |
|---|---|---|---|
| Status (September 2026) | Self-serve buying in 52 countries, more through the OpenAI ads team and partners; $1 billion annualized run rate reported on August 31, 2026 | Testing sponsored formats; retail first, travel and other categories next; Direct Offers for ready-to-buy shoppers | Ads discontinued February 2026; no stated plans to return |
| Formats | One labeled unit below the answer: name, headline, copy, image, landing link | Sponsored units below organic results; shopping formats with product listings | None |
| Targeting | Conversation topic and intent, advertiser context hints, optional personalization from past chats and memory | Query and conversational intent through the existing Google Ads stack | Not applicable |
| Organic citations alongside | Answer and its cited sources render first; ads never appear inside the answer | AI Mode cites sources inline; sponsored units sit below with labels | Citations are the only way brands appear |
Organic citations exist on every surface, paid slots on two, and on both of those the citations sit closer to the answer than the ad does. The sponsored units inside Google’s AI Mode sit below the cited response the same way ChatGPT’s do, and Perplexity, the smallest of the three by query volume, has no paid layer at all.
Perplexity’s exit is worth a second look. The company told the Financial Times that users need to believe they are getting the best possible answer, and that visible ads make people second-guess it. Perplexity now leans on subscriptions, with paid plans running $20 to $200 per month. OpenAI went the other way and added ChatGPT Go at $8 a month with ads, a step Perplexity does not match below $20, so the two have split on who pays and how. One well-funded AI engine looked at ad revenue and decided trust was worth more. That tells you how carefully OpenAI will guard the line between ads and answers.
What this means if you already do GEO or AEO
Ads have not changed the inputs the model uses to pick sources. It still selects on relevance, clarity, and authority, so existing GEO work keeps its full value. If anything, paid slots arriving below answers raises the value of the free slots inside them. Our read on where AI search adoption actually stands explains why this surface deserves the attention, and we tracked what GPT-5 meant for content teams when the answer engine itself changed.
Commerce is the direction of travel. OpenAI keeps building shopping surfaces, and we profiled OpenAI’s shopping agent when it appeared. Google is further along: its February 2026 letter confirmed agent-driven checkout rolling out inside AI Mode with Etsy and Wayfair. Ads, answers, and checkout are converging into one surface, and brands need to be present in all three layers.
Budget-wise, treat ChatGPT ads as experimental spend for now. The thinking in our guide to splitting budget between SEO and GEO still applies: fund the organic layer first, because it compounds and the paid layer does not.
What advertisers should do now, and what to skip
Worth doing now:
- Check two lists before anything else: whether your billing country has self-serve access, and whether your category is one OpenAI currently allows. Both rule people out.
- Create an Ads Manager account at ads.openai.com, or route through your agency. All four major holding companies have access.
- Run a small CPC test on decision-stage topics where a click leads to a measurable action. Start near OpenAI’s suggested $3-5 bid range.
- Tag every landing URL with UTM parameters before launch. The parameters persist, so attribution works from day one.
- Write context hints in the language buyers actually type into ChatGPT, pulled from real prompt research.
- Match landing pages to the conversation. Someone arriving from a comparison chat needs specifics, fast.
Worth skipping:
- Long spending commitments. OpenAI says more formats, objectives, and buying options are coming, and the ad policy has been revised six times since March 2026.
- Anyone selling placement inside ChatGPT answers. Those slots are never for sale, on any surface.
- Budget plans built on CPM figures OpenAI has not published.
- Assuming approval is permanent. Since September 10, 2026 OpenAI reserves the right to refuse ads that conflict with its business interests or competitive position, which matters if you sell anything OpenAI also sells.
Where to start this week
Check whether your billing country has self-serve access and whether your category is allowed. Brief whoever owns paid media on the format, the exclusions, and the $3-5 CPC starting point. Then baseline your organic presence before spending anything: our free AI visibility tools score a page for answer readiness and track how often the engines name your brand.
If the ads make sense for you, run them on top of a cited position rather than as a substitute for one. The brands that win this surface will hold both the citation and the slot below it.