How ChatGPT ad targeting works—and what users can change

A report by 404 Media, as recounted by Gizmodo, says two ChatGPT ad campaigns showed advertisers could upload identifiers—including email addresses, phone numbers and Android advertising IDs—to build audiences matched against ChatGPT users. OpenAI’s Ads Manager documentation lists the identifiers it accepts and sets a key limit: advertisers need 25,000 matched users for audience inclusion or bid adjustments, not merely 25,000 uploaded rows. (Gizmodo’s account attributes the finding to 404 Media reporter Joseph Cox.)

OpenAI says it began testing ChatGPT ads in the United States on February 9, 2026; eligible Free and Go users in selected regions may see them. The rules for advertiser-created audiences and the controls available to users answer different questions: matching affects who may be included in an audience, while personalization settings affect which signals can shape an ad.

Which identifiers can advertisers use?

OpenAI lists email addresses, phone numbers, SHA-256-hashed email addresses and phone numbers, and Google Advertising ID (GAID) as accepted identifiers for custom audiences. The same list can contain more rows than the number of people who actually match: invalid, duplicate and unmatched entries do not count toward the audience threshold. OpenAI’s custom-audience rules also say uploaded audience files are deleted after processing, usually no later than 24 hours after upload.

The 25,000-user threshold applies to audience inclusion and bid adjustments. OpenAI says audiences with fewer matched users can still be used to exclude people, and recommends at least 100,000 users for inclusion or bid adjustments. Ads Manager does not reveal which individuals matched or let advertisers select specific people from an audience. So a one-person list cannot satisfy the documented inclusion threshold, and advertisers do not get a roster of matched ChatGPT users. A match also does not guarantee that a person will see an ad.

What does ad personalization change?

OpenAI says ad selection can consider the current conversation, advertiser-provided context hints and targeting choices, the ad and its landing page, and—when personalization is enabled—selected signals from a user’s broader ChatGPT experience. Turning personalization off does not remove ads: OpenAI says the current chat may still inform them, while other chats, ad history and topics are not used for ad selection. (OpenAI’s explanation of ads in ChatGPT describes these signals and controls.)

OpenAI says advertisers do not receive chats, chat histories, memories or personal details. It says advertisers receive aggregated, non-identifying performance information, such as total views or clicks.

NeoTeo has also covered OpenAI’s Sponsored Agents test inside ChatGPT, a separate advertiser-facing format.

Choosing between personalization and no ads

Turning personalization off changes which signals may inform an ad; it does not switch ads off. OpenAI describes separate choices for users who want an ad-free experience. The practical differences are:

For a Free user who wants to remove ads without choosing a paid subscription, Ads-Free is the documented option, with lower usage limits and potentially reduced feature access.