60% of ChatGPT Ad Conversions Happen After the Click Window Closes
Only about 40% of conversions attributable to ChatGPT Ads happen in the immediate post-click session. The remaining 60% occur hours, days, or weeks later through channels with no traceable browser link to the original ChatGPT interaction. The oppref click identifier lives in a first-party cookie with a 30-day lifespan, but Safari ITP caps client-set cookies at 7 days — silently expiring the attribution link for the majority of mobile shoppers. Server-side capture of the oppref at landing and delayed conversion upload via the Conversions API are the only ways to close this gap.
The Conversation Gap: ChatGPT Ads Don’t Convert Like Search
ChatGPT users are researching, not buying — and the conversion happens days or weeks later through a channel with no link to the original ad.
Early advertiser data puts the scale of this gap at roughly 60%. Only about 40% of conversions attributable to ChatGPT Ads happen in the immediate post-click session. The remaining 60% occur hours, days, or weeks later, through channels that have no traceable link back to the original ChatGPT interaction.
This isn’t a measurement failure. It’s the structural reality of conversational AI advertising. A Google search ad catches someone who typed “buy running shoes” — the intent is immediate, the conversion path is short, and the click-to-purchase window is tight. A ChatGPT ad catches someone mid-conversation — researching, comparing, discovering. The purchase decision is still forming.
Only about 40% of conversions attributable to ChatGPT Ads happen in the immediate post-click session, with the remaining 60% occurring hours, days, or weeks later through channels that have no traceable browser link to the original interaction.
The ecommerce conversion timeline confirms this pattern across all channels. Industry data shows that while 75% of ecommerce conversions happen within 24 hours of initial engagement, 90% occur by day 12 after the first click. The final 5–10% take more than four weeks. For ChatGPT — where the initial interaction is research, not purchase — the curve shifts even further toward delayed conversion.
If you’re measuring ChatGPT Ads performance using only same-session conversions, you’re seeing less than half the picture. And if your conversion optimisation campaign is feeding OpenAI incomplete data, the algorithm is learning from a biased sample.
You may be interested in: OpenAI Launched a Conversions API for ChatGPT Ads — WooCommerce Has No Native Integration Yet
The Oppref Cookie Works for 30 Days — Except on Safari
OpenAI sets a 30-day attribution cookie, but Safari ITP expires it in 7 days — silently breaking attribution for the majority of mobile shoppers.
When a user clicks a ChatGPT ad, OpenAI appends the oppref identifier to the landing page URL. The OAIQ pixel SDK captures this parameter and stores it in a first-party cookie named __oppref on the merchant’s domain with a 720-hour time-to-live — 30 days.
In theory, this 30-day window should be long enough to capture most delayed conversions. In practice, Safari’s Intelligent Tracking Prevention caps client-set JavaScript cookies at 7 days. After day 7, the __oppref cookie silently expires. Any purchase after that point cannot be attributed to the ChatGPT ad click via the browser pixel.
The oppref click identifier is stored in a first-party __oppref cookie with a 720-hour (30-day) lifespan, but Safari ITP caps client-set JavaScript cookies at 7 days — silently expiring ChatGPT ad attribution for Safari and iOS users.
This is the same ITP mechanism that degrades Meta’s fbclid and Google’s gclid on Safari. The pattern is identical: the platform sets a long-lived cookie, Safari overrides the expiry, and the advertiser loses attribution on delayed conversions without realising it.
| Attribution Method | Intended Window | Safari ITP Reality | Delayed Conversions Captured |
|---|---|---|---|
| Browser pixel (__oppref cookie) | 30 days (720 hours) | 7 days (ITP cap) | Same-session + 7 days only |
| Server-side oppref capture | Unlimited (stored in database) | Not affected by ITP | Full conversion window |
| Conversions API (delayed upload) | 7-day upload deadline | Not affected by ITP | Any conversion within 7 days of event |
For WooCommerce stores where a significant share of traffic comes from iOS devices, the Safari ITP cap is not an edge case — it’s the majority of mobile attribution. The 30-day oppref window that OpenAI designed is functionally a 7-day window for every Safari user.
Your GA4 Can’t See the Gap
ChatGPT traffic shows as Direct in GA4 because the referrer gets stripped — making the attribution gap completely invisible in your analytics.
The attribution gap is made worse by a second problem: GA4 can’t reliably identify ChatGPT traffic at all. When a user clicks a ChatGPT ad, the click opens in ChatGPT’s in-app WebView. Mobile browsers and in-app environments frequently strip the referrer header before the page loads. The result is that ChatGPT-driven traffic appears as Direct in GA4.
ChatGPT ad clicks open in an in-app WebView and mobile browsers frequently strip the referrer header, causing ChatGPT-driven traffic to appear as Direct in GA4 — making the attribution gap invisible without server-side capture.
You can’t measure a gap you can’t see. If ChatGPT traffic is being bucketed as Direct in GA4, your attribution reports show no ChatGPT contribution at all. The 40% of same-session conversions that should be credited to ChatGPT are being attributed to Direct. The 60% of delayed conversions are distributed across whatever channel the user happened to return through — organic search, email, a bookmark.
The fix requires two layers. First, UTM parameters on the ChatGPT ad’s destination URL so GA4 can identify the traffic source regardless of referrer stripping. Second, server-side capture of the oppref identifier so the delayed conversions can be linked back to the original click even when the cookie has expired.
Without both layers, ChatGPT Ads look like they don’t work — not because they don’t convert, but because you can’t see the conversions they drive.
Server-Side Capture Closes the Window
Capturing the oppref at landing and storing it server-side — in your database, not a browser cookie — keeps the attribution chain intact for the full conversion window.
The server-side model captures the oppref parameter the moment the user lands on your WooCommerce store. Instead of relying on a browser cookie that Safari will expire in 7 days, the server reads the URL parameter and stores it in a database record tied to the user session or customer account.
When that user returns days or weeks later and completes a purchase, the server retrieves the stored oppref and includes it in the Conversions API event sent to OpenAI. The attribution chain is intact regardless of cookie expiry, ad blockers, device switches, or browser changes.
This is the same architecture that fixes Meta’s fbclid and Google’s gclid degradation on Safari. The click identifier is captured once, at the source, and stored where the browser can’t touch it. The Conversions API then transmits the conversion with the original click ID attached — server-to-server, with no browser dependency.
Server-side capture of the oppref identifier at landing — stored in a first-party database, not a browser cookie — survives Safari ITP, ad blockers, and session breaks, keeping the attribution chain intact for the full 30-day window.
For WooCommerce stores already running server-side pipelines for Meta CAPI and Google Enhanced Conversions, the oppref capture is one more URL parameter stored at landing. The infrastructure that closes the attribution gap for ChatGPT is the same infrastructure that closes it for every other platform.
You may be interested in: Your Consent Mode v2 Looks Green and Transmits Nothing: The Silent WooCommerce Conversion Loss You Can’t Get Back
The 7-Day Upload Deadline You Can’t Miss
The Conversions API accepts event timestamps within a 7-day window — delayed conversions uploaded after that deadline are permanently lost.
Even with server-side oppref capture, there’s a constraint: the OpenAI Conversions API requires event timestamps to fall within approximately 7 days of the event occurrence. Events older than 7 days cannot be uploaded or attributed.
The OpenAI Conversions API accepts events with timestamps up to 7 days old, giving WooCommerce stores a narrow window to upload delayed conversions before they become permanently unattributable.
This creates a tension. The conversion might happen 14 or 21 days after the click — well within the 30-day oppref window — but the conversion event itself must be uploaded within 7 days of when it occurs. A server-side pipeline that sends events in near real-time eliminates this deadline entirely. The purchase fires the WooCommerce hook, the server captures it, and the Conversions API event is sent within seconds — not batched at the end of the week.
Stores that batch-upload conversions weekly or rely on manual exports risk hitting the 7-day wall. The conversion happened, the oppref was captured, the attribution was possible — but the upload was too late.
Conversion-optimised ChatGPT Ads campaigns launched on June 5, 2026, for advertisers who had their pixel or CAPI configured by June 1. The algorithm’s ability to optimise delivery toward conversions depends entirely on the completeness and timeliness of the conversion data it receives. Feed it 40% of your conversions and it optimises toward a biased sample. Feed it close to 100% and it learns your actual customer profile.
Key Takeaways
- 60% of ChatGPT ad conversions happen after the click session: Conversational AI advertising drives research-stage discovery, not immediate purchase. Same-session measurement captures less than half the value.
- The oppref cookie expires in 7 days on Safari, not 30: Safari ITP overrides OpenAI’s 720-hour cookie with a 7-day cap, silently breaking attribution for iOS and Safari users on delayed conversions.
- GA4 can’t see the gap: ChatGPT clicks open in an in-app WebView that strips the referrer, so ChatGPT-driven traffic appears as Direct — making the attribution gap invisible in standard analytics.
- Server-side oppref capture is the fix: Storing the click identifier in your database at landing — not in a browser cookie — keeps the attribution chain intact past Safari ITP, ad blockers, and device switches.
- The Conversions API has a 7-day upload deadline: Delayed conversions must be uploaded within 7 days of the event. A real-time server-side pipeline eliminates this deadline entirely.
ChatGPT Ads appear inside conversational AI sessions where users are researching, not buying. The user may be comparing options, building a shortlist, or discovering a product for the first time. The purchase decision happens hours, days, or weeks later through a different channel — a direct visit, an email click, or a Google search. This delayed conversion pattern is structural to conversational AI advertising, not a measurement error.
OpenAI sets the oppref cookie with a 720-hour (30-day) time-to-live as a first-party cookie on the merchant domain. However, Safari ITP caps client-set JavaScript cookies at 7 days. After day 7, the oppref cookie silently expires and any subsequent purchase by that user cannot be attributed back to the ChatGPT ad click via the browser pixel alone.
When a user lands on your site from a ChatGPT ad, the oppref parameter is appended to the URL. Your server reads this parameter at landing and stores it in your database tied to the user session or customer record. When the user later converts — even days or weeks later — your server retrieves the stored oppref and includes it in the Conversions API event. This survives cookie expiry, ad blockers, and device switches.
The OpenAI Conversions API accepts event timestamps within approximately 7 days of the event occurrence. Events older than 7 days cannot be uploaded or attributed. This means WooCommerce stores need to upload delayed conversions promptly — a server-side pipeline that sends events in near real-time avoids this deadline entirely.
References
- Lapis — ChatGPT Ads Conversion Tracking: Pixel Setup, Conversions API, and Attribution Guide (May 2026)
- OpenAI Developers — JavaScript Pixel Documentation
- OpenAI Developers — Conversions API Documentation
- Opensend — Time to First Purchase Statistics for eCommerce Stores (2026)
- PPC Land — ChatGPT Ads Conversion Optimization Timeline (May 2026)
- SureThing — ChatGPT Ads in 2026: Costs, CTR Benchmarks and How to Advertise
- Digital Applied — ChatGPT Ads Attribution: A Measurement Playbook (July 2026)
- Measure Marketing Pro — How to Track OpenAI ChatGPT Ads (May 2026)
If your WooCommerce store is running ChatGPT Ads, the conversion data you can see in same-session reporting is less than half the story. Talk to Seresa about building server-side oppref capture and real-time Conversions API delivery that closes the 60% attribution gap.