ChatGPT Ads reporting and privacy: what you can see, what stays private

What reporting shows you

OpenAI's reporting is aggregated. You see the numbers that describe performance, not the people behind them:

• Impressions
• Clicks
• Spend
• Conversions

That is enough to judge whether a campaign works and to compare creatives, but it is a summary. There is no conversation-level detail, and there never will be one, because the chat is not shared.

What advertisers never receive

This is the part that surprises advertisers used to search and social. OpenAI is explicit about it.

Advertisers never receive chats, chat history, memories, names, emails, precise location or IP addresses. The ad is matched inside ChatGPT using the context of the conversation, and when personalization is on, select signals from the user's broader ChatGPT use. Those signals stay inside ChatGPT and are never passed to the advertiser.

In other words, the targeting can be rich while the reporting stays anonymous. You benefit from the match without seeing the data that produced it.

What the ad can and cannot do to the answer

Ads never influence answers. Advertisers cannot shape, rank or alter what ChatGPT says, and the ad is visually separated from the response and labeled as sponsored.

This matters for planning as much as for trust: you are buying placement next to a relevant moment, not a mention inside the answer. The copy and landing page carry the message, because nothing else will.

The controls users hold

Users have direct control over ads, and each control affects what you can reach.

• They can hide an individual ad or report it.
• They can turn off personalization, after which matching relies on the current conversation only.
• They can clear their ads data.
• They can switch to an ads-free Free plan with lower usage limits.

Personalized ads are also not initially available in the EEA or Switzerland, where matching is based on the current thread only. Treat non-personalized matching as the baseline your copy must work in.

Where your measurement has to live

Because reporting is aggregated and no personal data crosses over, attribution is your job, on your site.

OpenAI's reporting shows the click. Whether that click became a lead depends on your pixel, the Conversions API and preserving the click reference when the visitor lands. If that chain is missing, you will see clicks and spend with no way to tell which turned into pipeline.

Set it up before launch:

• Fire a pixel event on the actions that matter, not just page views.
• Send server-side events through the Conversions API for resilience.
• Keep the click reference intact from landing to conversion so the two can be joined.

Plan around the limits, not against them

The privacy model is not a gap to work around; it is the design. You get relevant placement and honest, aggregated numbers, and the user keeps their conversation.

The advertisers who do well here accept that split early. They pour effort into the two things they own — the creative and the landing experience — and they build measurement on their own side so aggregated reporting is enough to steer by.

None of this is unique to ChatGPT in spirit; it is only stricter here. Search and social have been tightening what advertisers can see for years, and privacy-safe, aggregated reporting is the direction the whole industry is moving. Building your own measurement now is an investment that keeps paying off well beyond this one channel.