The data you can and cannot get from ChatGPT Ads

With ChatGPT Ads, standard metrics such as impressions, clicks, CTR, CPM, and CPC come in from the moment you turn the ad on, and conversions are captured once you install the pixel or the Conversions API. What advertisers receive are these aggregated performance figures, not the conversations users had.

You cannot get the raw conversation text, a 1:1 link between a single click and a single conversation, or how many times a question actually came up within a category (absolute frequency). OpenAI's help article on ads in ChatGPT states that advertisers cannot access chats, chat history, memories, or personal details, and receive only aggregated performance information such as views and clicks.

Standard metrics you get as soon as the ad is on

Impressions, clicks, CTR (click-through rate), spend, average CPC, and average CPM come in from the moment you turn the ad on, in the same form as on any other ad channel. If you have ever run ads, there are almost no new metrics to learn.

These metrics count exactly how many times users saw an ad below an answer and how many times they actually clicked it. According to OpenAI's basics guide to ads, the maximum bid is set at the ad group level, with reach campaigns bidding on CPM and click campaigns bidding on CPC. For CPC, OpenAI recommends starting at $3–5 per click.

For the first two days of a campaign, track only CTR and average CPC. If CTR stays at the bottom of the category, you will want to fix the ad copy first, but the targeting signal is what needs fixing first. ChatGPT chooses ads based on the context and intent of the conversation rather than keywords, and for each ad group, advertisers enter context hints (context_hints in the documentation) that describe which conversations and topics the ad suits. Context hints are not exact-match keywords and do not guarantee delivery, but because they are the signal advertisers put into matching directly, they are the first thing to adjust.

If you want your brand's ad to appear when a user asks about "a serum for flaky winter skin," put words that express that situation and concern into the context hints. If the hints are off, the ad runs in the wrong conversations no matter how good the copy is.

When do conversion metrics start being captured?

Conversions, CVR, CPA, and conversion value are captured only after advertisers install the pixel or the Conversions API on their own sites. The list of supported events in OpenAI's developer documentation includes 12 standard events such as purchases, leads, and registrations, and you can define your own (custom) events on top of them. The pixel function takes the form oaiq().

According to OpenAI's reporting documentation, a conversion is credited to an ad when it happens within a set period after the ad is clicked (7, 14, or 30 days; 30 by default), and the dedicated conversion report also counts conversions within 1 day of seeing the ad (view-through conversions) separately. Within that period, a user who came from a ChatGPT conversation and then paid via search or Instagram may still be counted as a conversion in the ChatGPT Ads dashboard. Treating the dashboard numbers as the channel's full performance skews ad spend decisions.

So also send server events from GA4 or your own CRM through the Conversions API, and separately verify the revenue the ChatGPT channel actually added within the marketing mix (the incremental effect). If you set up this dual measurement from the start, there will be no later dispute over "did this channel really deliver results?"

How to use ad group performance for content work

Read CTR and CVR by ad group together, and they become diagnostic material, not just an ad report card. Because ad groups divide topics through their context hints, differences in performance between groups show which topics draw a response. If CTR is high but CVR is low within the same ad group, that signals a gap between the promise the ad made and the trust the landing page earns.

PION uses ChatGPT Ads performance as material for prioritizing content work. Here is how we carry ad group performance over into it.

• Topics from ad groups with high CVR go first into organic search content work.
• For groups with high CTR but low CVR, strengthen the landing page's trust materials (brand materials, case studies) before the ad copy.

Which data can't advertisers get?

Four of these, including raw conversation text, are data OpenAI says it does not give advertisers, and the absolute frequency of questions within a category cannot be extracted in the first place because of how the auction works.

Absolute frequency is the most misunderstood item. OpenAI selects ads through a relevance-weighted second-price auction, so the impression count advertisers see is the number of auctions won. Even an ad group with 100 impressions cannot be read as that topic having been asked about 100 times.

Even for the same question, if relevance or bid is low, the ad loses the auction and impressions are recorded as 0. So ChatGPT Ads data only shows what goes on inside the conversations where your ads are currently running; it does not show questions your ads have not yet reached. How often a question actually comes up in the market can only be captured by cross-measuring with external tools such as Otterly, Semrush, and Adthena.

Ad placements and answer bodies are run separately, so you also choose who handles each one separately. How to split the two and which items to compare in a proposal are covered in What a ChatGPT marketing agency actually does.