ChatGPT Ads Glossary: The Terms, Defined From the Documentation
By The InPromptAds Team 8 min read
Last verified 15 August 2026. Version 1.0. Refreshed monthly, whether or not anything changed.
Most wrong answers about ChatGPT ads come from three pairs of terms being treated as synonyms. Each pair is defined below and each distinction costs someone real money.
Advertiser eligibility is not ad serving. The entry commitment is not the campaign minimum. Context hints are not keywords.
Terms are grouped by what they govern. Where OpenAI does not document something, the entry says so rather than filling the gap.
Access and eligibility
Advertiser eligibility. Whether a business can register for an advertiser account, determined by the country the business is registered in rather than where its customers are. Nine countries are listed as Available as of 13 August 2026. See also serving market.
Serving market. A country where ChatGPT users are shown ads. Distinct from advertiser eligibility and currently a smaller set. Brazil appears on the advertiser list while Brazilian users are not served, which is the clearest demonstration that the two are separate permissions.
Geotargets catalogue. OpenAI's published CSV of locations selectable in the campaign targeting picker. Contains 42,110 rows, of which 41,488 are United States postal codes, and lists 249 countries including many that serve no ads. A targeting picker, not an availability list.
Ads Admin role. The permission required to create an advertiser account, granted at admin.openai.com under Users, then Direct roles. A missing role, rather than a platform restriction, is the usual reason someone cannot create an account.
Ad-eligible pool. The set of users who can be shown ads: total users minus Plus, Pro and Business plan users, minus accounts identified as under 18, minus Free users who take the Ads-Free option, minus sensitive contexts. OpenAI has published the size of none of these exclusions.
Ads-Free option. A setting allowing Free users to stop seeing ads in exchange for lower message limits. Take-up rate unpublished.
Buying and auction
Reach objective. A campaign type optimising for impressions, paid per 1,000 impressions on a CPM basis.
Clicks objective. A campaign type optimising for clicks, paid per valid click on a CPC basis. What counts as a valid click is not documented.
Maximum bid. The ceiling an advertiser sets at ad group level, as a maximum CPM bid for Reach campaigns or a maximum CPC bid for Clicks campaigns. Because the auction is second-price, the maximum bid is not the expected price. OpenAI recommends starting CPC campaigns at $3 to $5 USD.
Relevance-weighted second-price auction. The mechanism OpenAI uses to select among eligible ads. Relevance is weighted into selection, so a more relevant ad can beat a higher bid, and the winner pays a price derived from the next competing bid rather than their own maximum.
Bid-strength guidance. An in-product signal in Ads Manager indicating whether a bid is likely to be competitive or may limit delivery. Calculation method undocumented.
Entry commitment. The 200,000 USD commitment required to enter the ad programme, removed on 5 May 2026. Frequently and wrongly reported as the removal of all minimum spend.
Campaign minimum. The daily floor per campaign, still in force: 25 USD in the United States, 15 GBP, 2,500 JPY, 25,000 KRW, 25 AUD, 25 CAD, 25 NZD, 40 BRL, 150 MXN.
Targeting and matching
Context hints. Descriptions of conversations, topics or keywords where an advertiser's product may be relevant, set at ad group level. OpenAI states they guide ad matching but are not exact-match keywords and do not guarantee delivery in specific conversations.
Selection signals. The inputs OpenAI documents as feeding ad selection: the context and intent of the current conversation, the ad's landing page, its title and copy, advertiser-provided context hints, and, where ads personalisation is enabled, select signals from the user's broader ChatGPT experience.
Ads personalisation. A user setting that, when enabled, allows signals from the user's broader ChatGPT experience to inform ad selection beyond the current conversation.
Sensitive user journeys. A category in OpenAI's ad policies where no ads run. Named alongside conversations about personal health, mental health and politics. The boundary is not defined in the policy, which makes it the least predictable constraint on available inventory.
Creative and delivery
Ad unit. The delivered advertisement, containing advertiser name, favicon, title, copy, landing page and image asset. Appears below ChatGPT responses.
Image asset. The creative. PNG or JPG, square, no larger than 1200 by 1200 pixels, publicly accessible via a URL that opens directly to the image, hosted on public Google Drive, AWS S3, CloudFront, or the advertiser's own domain or CDN. OpenAI states no minimum size, despite a circulating 640 pixel claim.
Character limits. Not documented by OpenAI. Third-party claims range from 16 to 50 characters for the title and 32 to 100 for the description, disagreeing by more than threefold.
Account name and logo. Set in Settings, Account info. Ads will not serve until this is complete, and the failure produces no error message.
OAI-AdsBot. The OpenAI user agent relevant to ad landing pages, distinct from GPTBot. Landing pages blocking OpenAI user agents or IP addresses are a listed cause of ad rejection.
Feeds
Product feed. A catalogue of products submitted for Product Ads campaigns, via CSV or TXT upload, hosted HTTPS URL, or SFTP connection. Reported catalogue range is roughly 1,000 to 2 million products.
Feed expiry. Items expire after two weeks, which is why OpenAI recommends hosted URL or automated SFTP over manual upload.
Ads-eligible products. Products must be marked ads-eligible to appear in campaigns. Uploading a catalogue does not make it biddable.
Products tab. A reporting view, not an inventory view. Products appear only after an active feed campaign delivers, then after reporting catches up.
Measurement
Ads Manager Beta reporting. Includes impressions, clicks, spend, click-through rate, average CPC, average CPM and conversions.
Conversion measurement. Configurable in Ads Manager Beta. Set it up before launch; it cannot be backfilled.
Static tracking parameters. UTM and similar parameters added to landing page URLs, which persist on ad clicks and can be read in existing analytics tools.
Reporting latency. Up to seven hours between delivery and reporting. OpenAI advises allowing 24 hours before flagging a delivery issue.
Benchmarks. None exist. OpenAI states there are no performance benchmarks across advertisers, industries or campaign types. Any published ChatGPT ads benchmark is a single account, a relabelled figure from another platform, or a model.
Answer Independence. OpenAI's stated principle that advertising does not influence the assistant's answers. Governs answer generation, and does not address whether user behaviour changes on a response carrying both an answer and an ad.
Review and approval
Business verification. Identity and business verification, run through Persona against the entity's registered details.
Account review. A separate step from verification. Passing verification does not mean the account is approved to run campaigns.
Ad review. Per-ad review, ongoing after the account is live. Rejected ads show in the Ads tab; hovering a Not serving ad gives the reason. Re-review can be requested.
Immutable account fields. Country, currency and time zone, set at signup and not changeable afterwards. Correcting them requires a new advertiser account created with a different email address, since an address already used cannot be reused.
Case-by-case categories. Financial and health advertising, approved individually with proof of licensure. Outside the United States, ads for financial services and health services are generally prohibited, so this route is a US carve-out.
How this page is maintained
Refreshed monthly against OpenAI's advertiser documentation, ad policies, availability page and feed documentation, with the version line updated on every check including checks that find nothing. Definitional changes are recorded in the changelog.
Terms marked as undocumented are the ones worth watching. Several of them, including character limits and the sensitive-journey boundary, are questions any advertiser with an account could resolve publicly.
Quick answers
What is the difference between advertiser eligibility and serving markets on ChatGPT ads? Advertiser eligibility is whether a business can register an account, determined by the country the business is registered in. A serving market is a country where users are shown ads. They are separate permissions, and Brazil is on the advertiser list while Brazilian users are not served.
Are context hints the same as keywords in ChatGPT ads? No. OpenAI states context hints guide ad matching but are not exact-match keywords and do not guarantee delivery in specific conversations.
Did OpenAI remove the minimum spend for ChatGPT ads? OpenAI removed the 200,000 USD entry commitment on 5 May 2026. The campaign minimum remains: 25 USD per day per campaign in the United States, with local equivalents in other markets. The two are frequently conflated.
What are ChatGPT ad character limits? OpenAI does not publish them. Third-party claims range from 16 to 50 characters for the title and 32 to 100 for the description, disagreeing by more than threefold.
What is a relevance-weighted second-price auction? The mechanism OpenAI uses to select among eligible ads. Relevance is weighted into selection, so a more relevant ad can beat a higher bid, and the winner pays a price derived from the next competing bid rather than their own maximum.
What is the ad-eligible pool on ChatGPT? Total users minus Plus, Pro and Business plan users, minus accounts identified as under 18, minus Free users who take the Ads-Free option, minus sensitive contexts where no ads run. OpenAI has published the size of none of these exclusions.
Related reading
- ChatGPT ads eligibility by country — the two lists and the catalogue that misleads
- How the ChatGPT ads auction works
- ChatGPT ads changelog — every documented change, dated
Changelog
15 August 2026. v1.0 published.
Field kit
Tools
Site
OpenAI (primary)
The InPromptAds Team
Writes about AI-native advertising and the shift from search to conversation.
