OpenAI ChatGPT Ads Target Users via Uploaded Email and Phone Lists

OpenAI allows advertisers to target specific ChatGPT users by uploading lists containing personal identifiers, according to an investigation by 404 Media published in Gizmodo. The platform’s advertising system matches these uploaded identifiers against its user base to serve personalized visual and text ads within conversations.

Advertisers can use email addresses, phone numbers, or unique Android advertising IDs to build these audiences. OpenAI’s system checks if the uploaded accounts link to active ChatGPT users. If a match is found, that user becomes eligible to receive the specific advertisement, effectively allowing marketers to reach individuals they have already identified through other means.

Data Handling and Privacy Concerns

The investigation highlights potential risks in how OpenAI handles the uploaded data. It remains unclear whether the platform hashes raw user data before processing it, meaning sensitive information could theoretically sit exposed on the servers. OpenAI states that it deletes all uploaded files within 24 hours of processing the data, but the interim storage method has not been fully detailed.

Additionally, the platform does not reveal to advertisers when a specific account is matched. This prevents users from using the ad tool as a reconnaissance technique to discover if someone else has a ChatGPT profile. However, the initial upload of raw identifiers still requires trust in OpenAI’s internal data protection mechanisms during that 24-hour window.

Targeting Limits and Opt-Out Gaps

OpenAI imposes a minimum threshold for these custom audience lists. Advertisers must provide a list of at least 25,000 entries to create an audience based on uploaded identifiers. This restriction prevents marketers from targeting single individuals repeatedly with ads, but it still allows for broad, precise targeting of large groups of known contacts.

Users can opt out of personalized advertising in their settings, but this control may not be comprehensive. OpenAI notes that the opt-out setting excludes ads based on past chats, activity, and preferences. However, ads may still be served based on the user’s current chat context, meaning the opt-out does not fully block all personalized ad delivery methods.

What should you do?

Security professionals and users should review their ChatGPT privacy settings to understand the extent of data collection. While no patch fixes this business model, users can limit exposure by adjusting ad preferences. Organizations should monitor how their employee data might be used in external advertising campaigns if those identifiers are leaked or sold.

What to do and how to stay safe: OpenAI

• Review your ChatGPT account settings to opt out of personalized advertising where possible.
• Avoid using the same email addresses or phone numbers for high-risk services and your ChatGPT account.
• Monitor for unexpected ads in your conversations that reference recent or sensitive topics.
• Assume that your ChatGPT usage data may be linked to external identifiers if you have reused contact info.

Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation

General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

Can advertisers see if they successfully matched a specific user?

No, OpenAI does not show the advertiser when it hits on a matched account, preventing profile discovery via ads.

How many entries are needed to target a custom audience?

Advertisers must upload a list of at least 25,000 entries to create an audience based on identifiers.

Does opting out of personalized ads block all targeted advertising?

No, opting out excludes ads based on past activity, but ads may still be based on your current chat.