ChatGPT Ads Attribution: Click Windows, View-Through, and Reporting Delays
How ChatGPT Ads Attributes a Conversion
OpenAI's measurement documentation lists the conditions explicitly. The event must come from a data source connected to your ad account - the conversion tracking setup covers those sources - and it must match a conversion event configured for the campaign. The event has to occur within the applicable attribution window, and OpenAI has to be able to connect it to an eligible ad click using available measurement signals. Those signals include the OpenAI click reference called oppref, eligible advanced matching information, and modeled measurement where available. These four govern the click-attributed Conversions metric; view-through attribution swaps the click condition for a qualifying impression. If any of the four fails, the click-attributed conversion never appears in Ads Manager, no matter how real it was.
The connection layer deserves a closer look. The oppref parameter is appended to the landing page URL when an ad is clicked, and the OpenAI Pixel captures it and stores it in a first-party cookie so later conversion events can be tied back to the click. When direct click information is incomplete, automatic advanced matching fills the gap - the pixel detects supported customer information on your site, normalizes and hashes it with SHA-256 in the browser, and attaches it to conversion events without sending raw data. Where modeled measurement is available, OpenAI may also estimate attribution for otherwise unattributed events using aggregated patterns.
The Click Window and the One-Day View-Through Window
Click-through attribution measures conversions that follow a click inside the applicable configured click window. View-through is a different instrument: when it is available for your account, Ads Manager shows one-day view-through conversions - labeled VTA (1d) - counting conversions that happen within one day after an eligible ad impression, with no changes required to your pixel integration. The precedence rule matters when both could apply: if an impression and a click both qualify for the same conversion, the click gets the credit and the view-through does not.
View-through numbers live separately from the main metric at the campaign, ad group, and ad levels, with the breakdown shown when you hover over the Conversions column and exportable to CSV. OpenAI frames them as supplemental reporting in the strictest sense - VTA (1d) does not affect CPA, post-click conversion rate, bidding, billing, or conversion optimization. Unlike platforms where post-view credit feeds bidding directly, ChatGPT Ads keeps VTA (1d) outside optimization - and treating it as more than context, the way the attribution window habits from those platforms encourage, will inflate what you believe your campaigns are optimizing toward.
What Counts in the Conversions Metric - and What Does Not
Ads Manager displays a single Conversions total for the selected scope and date range, and that total includes click-through conversions only. View-through conversions never fold into it - they report separately - which makes the Conversions column the number that bidding and optimization actually respond to. Two more subtleties shape what the column means. Order Created Sales and Order Created ROAS are calculated only from attributed order_created events, so other events never contribute to those columns. And reported conversion totals may include modeled conversions where available - estimates OpenAI adds for otherwise unattributed events - which is measurement completeness by design, not a bug, but it changes what your total represents.
The 24-48 Hour Reporting Delay
OpenAI's reporting documentation carries an explicit note: allow 24-48 hours for attributed conversions to be included in reporting. The delay has operational consequences beyond patience. Pausing or judging a campaign on same-day conversion counts reads noise as signal, and campaign updates themselves are not instant - after a pause, ads may keep serving for up to 24 hours, and both reporting and charges can be adjusted afterward to reflect that activity. The disciplined reading order is delivery first, conversions a day or two later, and decisions only after the window closes.
Why Ads Manager and Your Tracker Disagree
Every tracker-vs-Ads-Manager comparison in ChatGPT Ads breaks down into the same components. Windows differ: your tracker counts conversions on its own lookback rules, while Ads Manager applies its configured click window and, separately, one-day view-through. Identifiers differ: the tracker attributes what its own click parameter catches, while OpenAI attributes through oppref, advanced matching, and - where available - modeled estimates your tracker will never see. Timing differs: the 24-48 hour reporting inclusion means same-day comparisons are structurally unfair. The same structural gap exists in Meta's tracker discrepancies and in TikTok's, and the reading discipline is identical too: compare trends and per-campaign ratios over stable periods, not single-day totals, and treat the platform column as the number that bidding actually uses.
Improving the Signals That Attribution Depends On
Attribution quality follows signal quality, and OpenAI's improvement checklist is short. Install the pixel across relevant pages and fire events when the intended action actually happens. Preserve the oppref click reference through redirects and landing-page navigation, and include it with server-side events when it is available - it is the strongest connective tissue between a click and a conversion. Run the pixel and the conversions API together with shared event IDs so the copies deduplicate instead of doubling. MOST delivers tracked conversions server-side to ChatGPT Ads alongside Meta, TikTok, Reddit, Snapchat, and Pinterest, and the free Pixel Activator validates one event end-to-end at pixel.way2.us before recurring delivery starts.