ChatGPT Ads

ChatGPT Ads is the advertising system that OpenAI began testing in ChatGPT in February 2026. It places paid, labeled advertisements below some responses for eligible users and provides campaign-buying and measurement services to advertisers. The system remains a phased product: whether a person can see ads and whether an advertiser can buy them directly both depend on plan, age, region, and rollout status.^{[2][3][7][10]}

OpenAI says advertising does not affect the text of ChatGPT's answers. Ad selection is handled on systems separate from the chat model and uses relevance signals and advertiser bids. Advertisers can submit creatives, context hints, targeting settings, and bids, but they cannot shape, rank, or alter the answer itself.^[7][9][14]

History and rollout

OpenAI first announced its advertising plan on January 16, 2026. The announcement described a future US test for logged-in adults on the Free and Go plans and said the first placements would appear beneath relevant answers. It also set out company principles covering answer independence, conversation privacy, user controls, and a continuing ad-free option.^[2]

The US test began on February 9. At launch, it applied to logged-in adult users on Free and Go, while Plus, Pro, Business, Enterprise, and Education accounts were excluded. OpenAI initially sold campaigns directly to a small group of advertisers. It later expanded buying through agency and technology partners.^[3][4]

Scroll sideways for more →

The table records announcements, not a guarantee that every eligible user received ads on those dates. User-facing rollout and advertiser access are also different. For example, the August European announcement addressed ads reaching users, while a later update addressed when advertisers could use Ads Manager without going through OpenAI or a partner.^[3][5][6]

Who may see ads

OpenAI's current help documentation says ads may appear on Free and Go. Plus, Pro, Business, Enterprise, and Edu accounts do not have ads. This makes the common phrase "paid plans are ad-free" imprecise because Go is a paid, lower-cost plan that may include advertising. ChatGPT Plus, Pro, and the listed organizational plans are the specific ad-free tiers documented as of September 2026.^[7]

OpenAI says it does not show ads to accounts it identifies as belonging to people under 18, using account information and age prediction where available. The original test was limited to logged-in adults. The current help page also describes ads in logged-out experiences, where placements are limited to material appropriate for all audiences; it limits what it calls mature-content ads to eligible logged-in users.^[3][7]

Availability is phased by market and account. OpenAI's country list concerns self-service access for advertisers and changes as the beta expands. Signing up for an Ads Manager account does not grant buying access before a country is supported.^[8][10]

Free users in eligible regions can select an Ads-Free mode without paying, but it lowers message limits and removes access to some features. OpenAI identifies image generation and deep research as examples of features that may be unavailable in that mode. Plus and Pro remove ads without that particular Free-tier reduction.^[7]

Placement and answer separation

Ads can appear after the end of a ChatGPT response. Each placement is labeled "Sponsored" and is visually separated from the answer. A displayed ad is a paid placement and does not mean OpenAI endorses the advertiser, product, or service.^[7]

An ad unit can contain an advertiser name, logo, headline, descriptive copy, landing page, image, or multiple items. During the test, a response can be followed by one or more ad units, and a unit can include more than one advertiser when the system finds eligible matches.^[7][9]

OpenAI says the advertising system runs separately from the chat model. Advertisers do not control answer generation or the order of content within an answer. OpenAI's Advertising Terms separately reserve to OpenAI control over advertiser eligibility, campaign acceptance, inventory allocation, ad delivery and presentation, policy enforcement, measurement, and reporting.^[7][14]

Those boundaries do not mean advertiser inputs are irrelevant to the ad that appears. Advertisers provide the ad, destination, budget, bid, context hints, and permitted targeting choices. The ads system uses those materials in its separate selection process.^[9][14]

Selection and personalization

The selection system considers the context and intent of the current conversation, the ad's landing page, title and copy, advertiser-provided context hints and targeting settings, and expected relevance and outcomes. When more than one ad is eligible, OpenAI says ranking also considers advertiser bids.^[7][9]

Context hints describe conversations, topics, or keywords for which an advertiser believes its offer may be relevant. OpenAI's documentation says they are not exact-match search keywords and do not guarantee that an ad will appear beside a particular conversation.^[9]

If ad personalization is enabled, the system may also use signals from past chats, memory, and prior ad interactions. Turning personalization off prevents other chat threads, ad history, and saved topics from informing selection, but it does not stop the current thread from being used for contextual matching. OpenAI also lists broad context such as general location and language. Personalized ads were not initially available in the European Economic Area or Switzerland.^[7][12]

Temporary Chats do not contain ads. Actions such as hiding, reporting, viewing, or clicking an ad are not themselves added to ChatGPT memory. If a user explicitly shares an ad into a conversation with the "Ask ChatGPT" control, that new conversation follows the user's ordinary chat-history settings.^[7]

Privacy and user controls

OpenAI says advertisers do not receive ChatGPT conversations, chat history, memories, names, email addresses, precise locations, IP addresses, or sensitive information. For ordinary campaign reporting, advertisers receive aggregate, non-identifying measures such as total views and clicks.^[3][7]

This restriction on advertiser access does not mean ChatGPT processes no information for advertising. Depending on settings and region, OpenAI may process the current chat, past chats, memory, ad interactions, account or device information, and advertiser-provided information to select or measure ads. Its privacy policy also says it may receive information from advertisers or other data partners, such as information about purchases, for measurement and relevance.^[7][12]

Eligible users can turn personalization off, hide or report an ad, inspect why an ad was shown, and clear data used for ads. If ads data is cleared, OpenAI says it stops using that data for selection and removes it from its servers within 30 days. Users who leave personalization off can still receive contextual ads based on the current conversation.^[7]

Advertiser platform

Ads Manager Beta

Ads Manager Beta is OpenAI's self-service interface for creating, launching, changing, and measuring campaigns. Advertisers can use a guided workflow or bulk upload. They can set a goal, budget, bid, and pacing; upload creative; and inspect impressions, clicks, and spend at the campaign, ad-group, and ad levels. Reporting is available in tables, charts, and CSV exports. Account functions include members, permissions, API keys, billing information, and change logs.^[1][4][8]

OpenAI also offers an Advertiser API for programmatic campaign management and performance queries. Its developer documentation includes product feeds, campaign targeting, conversion-optimized campaigns, and custom audiences. Ads Manager remains labeled beta, and specific features or access can vary during rollout.^[8][10][11]

Buying and auction

ChatGPT Ads supports cost per thousand impressions, or CPM, and cost per click, or CPC. A Reach campaign uses a maximum CPM bid and charges per 1,000 impressions. A Clicks campaign uses a maximum CPC bid and charges for a valid click. OpenAI first used CPM during the pilot and added CPC in May 2026.^[4][9]

OpenAI describes its selection mechanism as a "relevance-weighted, second-price auction" among eligible ads. The public documentation does not disclose the relevance formula or the clearing-price calculation. It therefore does not support assuming that the winner simply pays the second-highest unweighted bid.^[9]

Measurement

Ads Manager reports impressions, clicks, spend, click-through rate, average CPC, average CPM, and conversions. Advertisers can append static UTM parameters to landing-page URLs. For actions after a click, OpenAI supports a browser-based Measurement Pixel and a server-side Conversions API. Documented conversion examples include purchases, leads, and signups.^[4][9][11]

Conversion tools allow an advertiser to send website events for attribution and optimization. They are separate from access to ChatGPT conversations. OpenAI says the routine campaign insights it returns to advertisers are aggregated, while its privacy policy discloses that OpenAI may receive advertiser-side information for measurement and ad relevance.^[4][7][12]

Advertising policy

OpenAI's placement policy excludes ads from conversations it classifies as sensitive, emotionally vulnerable, or unsafe for brands. Its examples include personal or mental-health discussions, child-safety content, political content, fraud, self-harm, weapons, and cyber abuse.^[7][13]

The content policy is a separate question from placement context. OpenAI initially concentrated on consumer categories such as household goods, travel, local services, digital products, and education. By August 2026, it allowed some approved advertisers in financial services, healthcare and medicine, and legal services under manual, market-specific restrictions. Political advertising remained prohibited. Rules apply to the ad creative and the destination page, not only to the headline shown in ChatGPT.^[13]

The distinction matters: an approved health-related advertiser may be eligible to run a compliant campaign in a permitted market, but OpenAI says the campaign still cannot appear beside a user's personal health conversation.^[7][13]

Reported scale and revenue

On August 31, 2026, OpenAI reported that ChatGPT Ads had reached a $1 billion annualized revenue run rate less than 200 days after launch. It also reported tens of thousands of advertisers, more than 50 technology and measurement partners, and availability in more than 40 countries through its sales team and partners. In the same announcement, OpenAI said ChatGPT as a whole had more than one billion weekly active users.^[6]

These are dated, company-reported figures. The announcement did not publish an audit or a calculation methodology. An annualized run rate extrapolates a current revenue pace over a year; it is not the amount of ad revenue accumulated during the first 200 days. The weekly-active-user figure covers ChatGPT overall, not the subset shown ads.^[6]

Independent study

An August 2026 study by Emma Lurie, Ro Encarnacion, Sorelle Friedler, and Danae Metaxa audited the initial US rollout with 91 sock-puppet accounts. The researchers used a 3 by 3 design intended to signal three racial or ethnic categories and three income levels. Across 335 prompts, they collected more than 3,000 ads from 186 advertisers.^[15]

The authors reported that their accounts began receiving ads 14 days after creation and that lower-income personas were more likely to receive ads regardless of the signaled racial category. They also found that the early inventory leaned toward consumer goods and that ads were clearly separated from the language model's response.^[15]

The study provides independent observations, but it does not describe the entire advertising system. It used simulated accounts, covered an early phase of the US test, and cannot by itself establish the current global system's causal targeting logic or its effects on real users. The version available in September 2026 was an arXiv preprint marked for publication at AIES 2026.^[15]

References

1. ^OpenAI. "Advertise in ChatGPT." Accessed September 3, 2026. ads.openai.com
2. ^Fidji Simo. "Our approach to advertising and expanding access to ChatGPT." OpenAI, January 16, 2026. openai.com/...-to-advertising-and-expanding-access
3. ^OpenAI. "Testing ads in ChatGPT." Originally published February 9, 2026; updated August 11, 2026. openai.com/...testing-ads-in-chatgpt
4. ^OpenAI. "New ways to buy ChatGPT ads." May 5, 2026. openai.com/...new-ways-to-buy-chatgpt-ads
5. ^OpenAI. "ChatGPT Ads expands across Europe." Originally published August 18, 2026; updated August 31, 2026. openai.com/...chatgpt-ads-expands-across-europe
6. ^OpenAI. "A milestone in expanding access to AI." August 31, 2026. openai.com/...anding-access-to-ai-with-chatgpt-ads
7. ^OpenAI Help Center. "Ads in ChatGPT." Accessed September 3, 2026. help.openai.com/...20001047-ads-in-chatgpt
8. ^OpenAI Help Center. "Ads Manager Beta Overview." Accessed September 3, 2026. help.openai.com/...20001206
9. ^OpenAI Help Center. "Ads in ChatGPT: The Basics." Accessed September 3, 2026. help.openai.com/...01207-ads-in-chatgpt-the-basics
10. ^OpenAI Help Center. "Ads Manager Availability." Accessed September 3, 2026. help.openai.com/...20001245-ads-manager-availability
11. ^OpenAI Developers. "Ads." Accessed September 3, 2026. developers.openai.com/ads
12. ^OpenAI. "Privacy Policy." Accessed September 3, 2026. openai.com/...privacy-policy
13. ^OpenAI. "Ad policies." Updated August 31, 2026. openai.com/...ad-policies
14. ^OpenAI. "Advertising Terms." Updated August 25, 2026. openai.com/...advertising-terms
15. ^Emma Lurie, Ro Encarnacion, Sorelle A. Friedler, and Danae Metaxa. "The Beginning of ChatGPT Ads." arXiv:2608.05008, submitted August 5, 2026. arxiv.org/...2608.05008