Growth marketer operator · San Jose, CA | ZuAI: 10K → 2M users at $0.02 CAC | $300k/mo ad spend managed
Ar.Bhavesh Pro
ChatGPT Ads

ChatGPT Ads attribution: measurement checklist

the checks to run before and after you spend on ChatGPT Ads: events, oppref, deduplication, MMP setup, and why Ads Manager and analytics disagree.

last updated 5 November 2026

ChatGPT Ads measurement holds up when four things are true before launch: the OpenAI pixel is in the head of every page, oppref survives every redirect between the ad and the conversion, a real conversion event fires and appears in Ads Manager, and the value you send matches what you actually earn. after launch the job changes. decide which source you budget from, expect Ads Manager to disagree with your analytics, and judge trends rather than reconciling every row.

Before you spend

work down this list in order. each item depends on the one above it.

  1. the pixel is in the <head> of every page, initialised with your Pixel ID. setup is in the pixel chapter.
  2. oppref survives every redirect. click your own ad path and check the parameter is still on the URL when the pixel loads. shorteners, locale redirects and http to https hops are the usual culprits.
  3. a real conversion fires and appears in Ads Manager. not page_viewed, which fires on its own. an actual measure event from a test purchase or sign-up.
  4. the value matches what you earn. if you send a value, it should be the real one, in the right currency.
  5. the event you will optimise toward is a standard event you can select. custom events cannot drive the Conversions objective, and the event is fixed once the campaign exists.
  6. if you run the pixel and the Conversions API together, both send the same event ID for the same conversion, built from your order ID. otherwise every conversion is counted twice. details in the Conversions API chapter.
  7. your Conversions API calls carry oppref whenever you have it.
  8. for an app on the store route: your measurement partner has the Pixel ID and Conversions API key, the install and in-app events are mapped, and the partner’s attribution link is the ad’s destination.
  9. you have decided which source you budget from: Ads Manager, your analytics, or for an app usually your measurement partner. decide now, before the numbers disagree.
  10. you know what you can afford to pay. the break-even CAC calculator gives the ceiling, and the LTV to CAC and payback calculator shows how long your cash is tied up before a user pays it back.
  11. your UTMs are clean, if you tag landing URLs with them. run them through the UTM checker so a capital letter does not split one campaign into two rows.

Why Ads Manager and your analytics will disagree

they will, and OpenAI says plainly that a difference does not by itself mean an error. the gap comes from different attribution windows, time zones, consent and cookie conditions, deduplication rules, and modelled conversions, where OpenAI estimates attribution for events it could not tie to a click directly.

for apps there are two more reasons. your measurement partner uses its own attribution windows and settings, and its attribution is click based, so there are no view-through installs in it. and attributed events reach Ads Manager after 24 to 48 hours, so the last two days in Ads Manager are not the final count.

the fix is not reconciliation. pick the source you budget from, keep it, and read direction over weeks rather than chasing every row.

Five breaks and what causes them

  • a green pixel and zero conversions: the measure calls were never added on the pages where the action completes.
  • every conversion counted twice: the pixel and the Conversions API send different event IDs, usually two random numbers generated separately.
  • conversions that do not tie back to clicks: a redirect dropped oppref before the pixel saw it, or the server event went out without it.
  • an app campaign showing nothing in Ads Manager: attributed installs arrive after 24 to 48 hours, and only for events your partner has mapped.
  • the campaign optimising toward the wrong thing: the event is fixed at creation, so the fix is a new campaign, not an edit.

After launch, every week

  • read the trend against your chosen source, not the daily row.
  • leave a Conversions campaign alone until its review date. edits restart learning to some degree.
  • judge cost per conversion against what a customer is worth and how fast the money comes back, not on its own.
  • compare the channel with your others in the channel CAC comparison, which shows where moving budget actually buys more users.

the wider habit of reading paid through numbers the platforms cannot blur is in reading paid signal after ATT. the ChatGPT Ads specifics behind this checklist are in the conversion tracking guide and the app installs guide.

this is the last chapter. a printable version of every check above is offered at the foot of the playbook hub.

measurement is one leg of the ChatGPT Ads playbook. if you want a readout you can trust set up on your own account, the breakdown call is free. Let’s talk Growth.

this page is part of the ChatGPT Ads playbook.

next

More in this playbook

Want this run on your app?

the playbook is free and yours. if you would rather it was executed than read, tell me what you are working on.

answers go to a private google sheet and to my inbox, used only to reply and follow up. the newsletter is opt in and separate.

Want this run on your app?

the breakdown call is free. bring one real growth problem, leave with the fix.

Let's talk Growth