To track ChatGPT Ads conversions, create a conversion data source in Ads Manager, send the chosen business event through the OpenAI Pixel, the Conversions API, or both, and select the matching event in your campaign. Then verify that OpenAI receives the event and can attribute it within the applicable window. Installing a Pixel alone does not prove a purchase or lead will appear in a campaign report. The current OpenAI conversion-measurement guide explains the conditions.
This article is a practical QA path for paid ads, not an organic AI visibility guide or a claim that Screpy has run these campaigns. If you need the broader channel picture, start with how ChatGPT Ads work. Use the cost guide later when your measured outcome is reliable enough to evaluate.
Which conversion event and data source should you choose?
Choose one action with business meaning. For an ecommerce store, Purchase is more informative than a product-page view if the goal is sales. For a service company, a completed qualified-lead form may be more useful than a generic page visit. The event received from your data source must match a standard or custom event configured for the campaign and fall within the applicable attribution rules. OpenAI's measurement guide describes click matching; its newer results guide also documents eligible one-day view-through reporting when that window is selected. A visible event in a debugging tool is therefore only the first checkpoint, not proof of campaign attribution.
The browser Pixel can send events from pages where the user acts. The Conversions API can send eligible events server-side. You may use either or both, depending on your implementation. For Shopify merchants using the official integration, the app can create or connect an OpenAI Pixel and send eligible commerce events; verify the event and account in the Shopify setup flow rather than assuming every store action is mapped automatically. OpenAI's Shopify guide describes that connection.
Prepare a small event inventory before configuration: event name, where it fires, whether a test action can be performed, whether value/currency are needed, and which campaign will optimize toward it. A hypothetical store might test one $40 order, but the dollar figure is merely a QA example—not a campaign result. Follow OpenAI's current developer event schema for actual field formats.
Only send data where permitted. OpenAI's guidance calls for clear user information and necessary consent under applicable law. Consent controls and privacy policy review belong in the implementation checklist, not as afterthoughts once reporting is live.
How do you preserve attribution and avoid double-counting?
OpenAI can append an oppref click reference to the landing-page URL. Its Pixel can capture that reference in a first-party cookie, and eligible server-side events can include it when available. Check that redirects preserve the query parameter all the way to the usable landing page. A redirect that silently drops it may weaken the direct click signal, even if a general analytics tool still shows a visit. That is a reason to test the complete URL path, not merely the final page's status code. OpenAI explains oppref and matching.
If the Pixel and Conversions API send the same purchase, give those two copies the same event ID so OpenAI can deduplicate them. Do not invent a new ID independently in the browser and backend for a single order. The simplified event journey is:
ad click with oppref → landing page → purchase ID 123 → Pixel event ID 123 + API event ID 123 → one intended conversion
This is an illustrative flow, not a claim that every conversion can be attributed or that an order number should always be exposed as an event ID. Use the platform's supported identifier format and privacy rules. The QA questions are: Was the same action sent twice? Did both paths use the same event ID? Did the campaign select the corresponding event? Did consent conditions allow either path? A received event and an attributed campaign conversion remain different states.
Screpy's website audit can help locate redirect or destination issues on the site side. It does not test the OpenAI Pixel, CAPI or campaign attribution; use Ads Manager and your event implementation for those checks.
Why does Ads Manager show zero conversions?
Work from event transport to attribution. A zero in the campaign report does not identify which stage failed:
| Observation | What to inspect next |
|---|---|
| No event reaches the data source | Pixel installation, API response, trigger, browser consent and the correct Ads Manager account |
| Event received but campaign total is zero | Event name/type selected in the campaign, date range, attribution window and eligible ad interaction |
| Two events for one purchase | Pixel/API event IDs and duplicate browser triggers |
| Clicks appear but same-day conversions do not | Event timing, reporting delay, test path and whether the test purchase followed an eligible ad click |
OpenAI says a reportable conversion depends on an eligible ad interaction and the selected reporting window, not only on receiving a matching event. Click-through and, when enabled, eligible one-day view-through conversions may be included; the current results guide explains how to change the window and view the breakdown. OpenAI may also use other eligible matching or modeled measurement signals where available, so a missing oppref does not prove that every attribution will fail; preserving it is still a sensible QA step.
Keep the time window long enough for reporting. OpenAI's results guidance says conversion data can take 24–48 hours to settle. Check the account time zone and campaign dates before comparing a storefront order at 23:55 with the Ads Manager day. If the event is received but no attribution appears after the documented window, gather event IDs, timestamps, campaign event setting, destination URL and any available click reference for support. Do not “fix” the situation by firing a conversion on every page load.
Why do Ads Manager and GA4 report different numbers?
They answer different questions. UTM-tagged URLs can help GA4 identify sessions from paid ChatGPT clicks, while Ads Manager reports platform events and conversions using its own eligibility, attribution windows, deduplication and, where available, modeled measurement. Different date boundaries, consent conditions and browser storage can change either total. OpenAI says discrepancies do not necessarily mean that one system is broken. See its measurement caveats and reporting guide.
For a meaningful comparison, write down the precise metric: paid ad clicks, GA4 sessions, store orders or Ads Manager attributed purchases. Give each a date range, time zone and attribution window. A single ad click may not create a GA4 session under every consent or browser condition; a store order can exist without being attributable to that ad. Resist converting this into a one-to-one reconciliation exercise if the underlying definitions differ.
Paid ads also should not be folded into organic AI visibility reporting. Screpy's ChatGPT traffic guide addresses unpaid referral/visibility questions. OpenAI says advertisers can add static UTM parameters to ad URLs; use a naming convention that explicitly distinguishes paid campaigns from organic ChatGPT referrals.
What should you validate before scaling spend?
Run one documented test path with a permitted test event, and save the expected and observed result at each stage:
- A real ad destination opens with the intended UTM parameters and preserves
opprefif it is present. Redirects do not lose the user or the relevant query values. - The intended business action fires exactly the configured event, with values in the platform's supported format. An unrelated page view does not fire Purchase.
- The event reaches the correct data source and Ads Manager account. If Pixel and API both send it, the same action shares an event ID.
- The campaign is configured for the corresponding event and the action is within its attribution window. Wait for normal reporting before declaring a failure.
- The team records consent behavior, time zone and test transaction so a later discrepancy can be explained without sending personal data into an ad-hoc spreadsheet.
Once the path is sound, judge the channel against your own economics, not a generic “good conversion rate.” The ChatGPT Ads cost guide distinguishes your bid, media spend and cost per actual business result. If your store is new to the official app, complete its setup preflight before making a campaign optimization decision. Data quality is part of the test, not a task postponed until after budget increases.