Insights are the performance reports inside ChatGPT Ads Manager that show how your campaigns, ad groups, and ads are doing, using metrics like impressions, clicks, spend, click-through rate, and cost per click. They cover delivery and cost at an aggregate level, so they are separate from conversion tracking, which OpenAI reports through its pixel and Conversions API.
If you buy ChatGPT Ads, insights are where you decide what to pause, scale, or rewrite each week. Skip them and you keep paying for ad groups that burn budget without proving they move clicks or cost in the right direction.
In ChatGPT Ads, insights are the reporting layer that turns raw ad delivery into numbers you can act on, available in the Ads Manager interface and through a reporting API. That API returns six metrics: impressions, clicks, spend, click-through rate, average cost per click, and average cost per mille. You can view them at four levels: ad account, campaign, ad group, and ad.
Each report can use a time bucket of hourly, daily, or monthly, plus one optional breakdown at a time, by product, country, or device. Data refreshes hourly, so same-day activity shows without an overnight wait. Conversions, revenue, and return on ad spend are not columns in this feed; they depend on the OpenAI pixel and Conversions API as a separate measurement path.
Insights sit above the auction and below your analytics stack. They report what OpenAI charged and how often your ad was shown and clicked, with no view into the prompts, conversations, or audience details behind those numbers. AirOps connects that delivery data to the questions people actually ask AI assistants, so reporting can feed content decisions.
Resources: see how to structure and measure ChatGPT Ads campaigns from first launch.
Insights work as a pipeline that starts the moment your ad serves and ends with a table or chart you read. Here is the path your data takes.
Log events: As your ads serve inside ChatGPT, OpenAI records each impression, click, and the spend attached to it.
Aggregate: Those events roll up across four levels, from a single ad up to the whole ad account, so you can zoom in or out.
Request a report: In Ads Manager or through the reporting API, you choose a level, a date range, and a time bucket of hourly, daily, or monthly.
Apply a breakdown: You can split one report by product, country, or device, one dimension at a time, to see where delivery concentrates.
Read or export: Results appear as tables and charts, and you can export to CSV or pull them into a reporting tool.
The output tells you how much you spent and how your ads were delivered and clicked. It does not tell you which prompts triggered an ad or who saw it, so pair it with your own conversion and site data to judge business impact.
ChatGPT Ads is a young channel with no published benchmarks, so insights are the only first-hand evidence you have for whether to keep spending. Every decision to raise a budget, kill an ad group, or walk away rests on what these reports actually show.
They guide budget: Spend, CPC, and CPM in insights tell you what each click and thousand impressions actually cost, so you can move money toward the campaigns earning their keep.
They expose waste early: Without checking insights, an ad group can quietly spend for days at a rising CPC and a falling click-through rate before anyone notices the budget drained with little to show.
They set expectations: Because ChatGPT shows one ad per answer, click-through rates read lower than search, and insights give you the baseline to judge your own trend instead of chasing a number from another platform.
SEO managers use insights to compare ChatGPT Ads click costs against the organic AI citations they already track, deciding where paid support is worth it.
Growth marketers use insights to spot which campaigns hold a stable cost per click as they scale spend week over week.
Demand gen leads use insights to export spend and delivery data into a wider report that shows finance how the channel performs against others.
The same six metrics roll up across four levels, from the whole ad account down to a single ad, and the level you read changes the story completely, because an account that looks healthy overall can still hide one ad group quietly draining budget at a bad cost per click.
A breakdown splits a single report by product, country, or device, but only one dimension at a time, which is how you locate where spend and clicks concentrate when ChatGPT Ads gives you no query-level or demographic detail to work from.
Because clicks live in insights while conversions live in the pixel and Conversions API, the attribution window you set decides how many days after a click a conversion still counts, which quietly changes whether a campaign looks efficient or wasteful.
Track impressions, clicks, spend, CTR, CPC, and CPM for every ChatGPT Ads campaign in one place.
See same-day performance, since insights refresh hourly instead of waiting for an overnight batch.
Pull reports at four levels, from ad account down to a single ad.
Export to CSV or feed the reporting API into your own dashboards.
Compare delivery by product, country, or device to find where budget works hardest.
Check insights on a fixed weekly cadence, so you compare like periods instead of reacting to hourly noise.
Read at the ad group level first, because that is where cost and click-through problems usually show before they reach the account total.
Set one consistent attribution window across campaigns, so conversion comparisons stay honest.
Use a single breakdown per question, since stacking dimensions is not supported and muddies what you learn.
Export to CSV or the reporting API for anything finance-facing, so your numbers match the source instead of a screenshot.
Pair insights with your own site and conversion data, because the reports stop at delivery and cost.
Avoid judging ChatGPT Ads by Google or Meta benchmarks. The channel shows one ad per answer and has no published industry baselines, so a click-through rate that looks low against search can still be healthy here; build your read from your own trend line.
AirOps: connects your ChatGPT Ads spend and delivery data to the AI prompts and citations your brand is winning or missing, so insights inform what content to build next.
OpenAI Ads Manager: the native interface where insights live, with table and chart views, four report levels, and CSV export.
Supermetrics: pulls ChatGPT Ads metrics on an hourly refresh into spreadsheets and BI tools so you can blend them with other channels.
Open Ads Manager: Log into ads.openai.com and find the insights view for a campaign you already run. This takes nothing but access and shows you the metrics available today.
Pick your core metrics: Decide which of the six reported metrics matter for this campaign, usually spend, CPC, and click-through rate, so you are not distracted by every column.
Set a reporting cadence: Choose a weekly day to review, settle on one date range and time bucket you will use every time, and note where your baseline starts.
Add one breakdown: Split a report by product, country, or device to see where delivery concentrates, changing the dimension only when you have a new question.
Connect conversions: Install the OpenAI pixel or Conversions API and set a consistent attribution window, so insights on spend finally sit next to the outcomes they are meant to drive.
Insights are the performance reports in ChatGPT Ads that show how campaigns deliver and what they cost.
You read them at four levels and in hourly, daily, or monthly buckets, with one optional breakdown by product, country, or device.
The main constraint is scope: reporting stops at aggregate delivery and cost, with no query-level, demographic, or conversion data built in.
The main risk is misreading a low click-through rate against search benchmarks that do not apply to a one-ad-per-answer channel.
The leverage is cadence: a steady weekly read paired with your own conversion data turns raw numbers into budget decisions.
Insights and conversion tracking answer two different questions and live on two separate data paths. Insights are the delivery and cost reports built into Ads Manager, covering impressions, clicks, spend, click-through rate, average cost per click, and average cost per mille across your account, campaigns, ad groups, and ads. Conversion tracking sits outside that feed, running through the OpenAI pixel and the Conversions API to record actions like purchases, leads, and sign-ups on your own site. The reporting feed does not carry conversions, revenue, or return on ad spend as columns, so you cannot judge profitability from insights alone. In practice you use insights to see how efficiently your budget bought attention, then lean on conversion tracking to see whether that attention turned into business. Reading only one of the two gives you half the picture, which is why experienced buyers wire up both before they scale a campaign.
A weekly review is the right default for most ChatGPT Ads accounts, with a quick mid-week glance if a campaign is new or spending fast. The data refreshes hourly, which tempts advertisers to watch it all day, but hourly swings in a young channel are mostly noise and lead to overreacting. A steady weekly cadence lets you compare the same length of time week over week, so a change in cost per click or click-through rate means something instead of reflecting a slow afternoon. Check more often only in two cases: the first days of a new campaign, when you are confirming delivery actually started, and any week you make a budget or bid change, where you want to see the effect before it compounds. Set a fixed day and a fixed date range, and resist pulling the report twice a day out of anxiety; the trend line is what guides decisions.
Day-to-day swings in insights are normal in ChatGPT Ads. They come from how the channel delivers, and they rarely signal a reporting glitch. ChatGPT shows one ad per answer, so your impressions depend on how many relevant conversations happen that day, which is far less predictable than a search results page that loads on demand. Auction prices move as other advertisers enter and leave, so your average cost per click can drift without any change on your side. Small daily volumes also exaggerate percentages: a handful of extra clicks can swing click-through rate sharply when the denominator is low. The hourly refresh adds another layer, since a report pulled at noon captures a partial day. The fix is to stop reading single days and compare full weeks against each other, where these sources of variance average out and a real trend becomes visible.
You influence insights indirectly, because the reports only mirror decisions you make elsewhere in the account. You cannot edit a metric or change how OpenAI calculates it, but you shape every number by choosing budgets, bids, targeting, and creative. Raise a bid and your average cost per click usually moves; tighten targeting and your impressions and click-through rate shift; pause a weak ad group and the account averages change the next day. What you control on the reporting side is the lens: the report level, the date range, the time bucket, and the single breakdown you apply. Those choices decide what story the same underlying data tells, which is why two people can look at one campaign and reach different conclusions. Treat insights as a mirror of your account settings, adjust the settings deliberately, and give each change enough time to show up before you judge it.
Good performance is defined against your own goals and history, because OpenAI has not published cross-advertiser benchmarks for ChatGPT Ads. There is no official figure for a healthy click-through rate, cost per click, or cost per mille, and no auction or impression-share view to tell you how you stack up against competitors. That absence means a number is good when it moves your cost in the right direction week over week and stays within the budget you set to hit a target cost per outcome. Because ChatGPT serves a single ad per answer, click-through rates tend to read lower than on search, so judging them against Google figures will mislead you. Set a baseline in your first few weeks, decide the cost per conversion the channel needs to justify its spend, and measure every later week against that line. Good means improving on yourself, until shared benchmarks eventually exist.