If you have started to measure ChatGPT Ads conversions, you have probably noticed there is little established practice yet. The channel is new, the reporting is still developing, and most tracking advice was written for Meta and Google. This post covers the practical question: how do you get store orders to OpenAI Ads, and how far can you trust the numbers that come back?
It sits next to our earlier piece on why ChatGPT is becoming an ad channel for ecommerce, which covers the why. This one covers the how, with the limits left in.
Implemented is not approved
This distinction matters, so here it is plainly.
| What we can say | What we do not say |
|---|---|
| SyncBeacon has an OpenAI Ads destination, shipped in the product | That SyncBeacon is OpenAI-approved or an official OpenAI partner |
| Conversion events are sent to OpenAI Ads from store orders | That the integration is listed in any OpenAI marketplace |
| It is covered by our own automated tests | That OpenAI has validated the integration on its side |
Treat the data as a working implementation against OpenAI's documented capabilities. As the channel and its APIs change, details here may change too.
The measurement path
OpenAI Ads measurement follows the same pattern as other platforms, and the server side does the reliable part.
- A shopper clicks a ChatGPT ad and lands on your store. The URL carries a click reference (the
opprefparameter). - If you use the SyncBeacon browser script, it captures that reference and remembers it, so it is still available when the order arrives later.
- The order arrives at SyncBeacon from your store (a Shopify or WooCommerce webhook, or the Shopware order sync).
- SyncBeacon sends the conversion to OpenAI Ads, with the click reference when it has one.
The reason to send from the server is the one that applies on every platform: the browser alone misses events. The server-side tracking page explains the mechanism. For this channel, click-reference capture is the piece to get right, because a conversion with no click reference has less to match on.
What gets sent to OpenAI Ads
| Data | Sent? |
|---|---|
| Conversion events: order created, checkout started, items added, and similar | Yes |
| Order amounts and line items | Yes |
| Click reference (oppref) | Yes, when captured |
| IP address and user agent | Yes, when consent allows |
| Hashed email | Only for consented visitors |
| Phone number | No |
We deliberately do not claim that city, ZIP or country from store orders reach OpenAI. Assume matching rests on the click reference and the hashed email, nothing more.
Consent applies as it does elsewhere. If a visitor explicitly denies marketing consent through a detected banner, the event is stored for your records and not forwarded. When consent is missing, marketing identifiers are stripped from what is sent. Our GDPR page explains how consent gating works.
Setup: ChatGPT Ads conversion tracking in SyncBeacon
Setting up ChatGPT Ads conversion tracking takes about the same steps as any destination, with one difference: it is not a browser sign-in.
- Connect your store (Shopify, WooCommerce or Shopware).
- In your OpenAI Ads account, create an Ads API key.
- In SyncBeacon, open Integrations, choose OpenAI Ads and paste the key.
- SyncBeacon can create the pixel and conversions key for you when your ad account allows it. If it does not, attach the Pixel ID and conversions key manually.
- Install the browser script if you want click references captured.
- Place a test order and check its status on the Conversions page.
The step-by-step page is OpenAI Ads conversion tracking. If you are on the free Pilot plan, remember it includes one ad platform, so OpenAI Ads would be your only destination. Compare plans on the pricing page.
Checking that the data is real
For a new channel, verify before you trust.
1. Reconcile with your store
Count orders in your store that came through ChatGPT ad links over a set period, then compare with what OpenAI Ads reports for the same period. Small differences are normal, since attribution windows and rules are OpenAI's. A large gap is a signal to investigate the click reference and consent.
2. Look at delivery status in SyncBeacon
The Conversions page lists each event and its status per destination. Failed deliveries retry up to 24 times, but permanent errors such as bad credentials stop immediately and are shown as errors. If you see errors, fix the connection before judging campaign performance.
3. Check what consent removed
If a share of orders arrives without a click reference or hashed email, check how many of those shoppers declined marketing consent. That is expected behaviour, but it affects how complete the numbers look.
A sensible rollout before you scale spend
Because the channel is new and the tooling around it is still settling, a staged approach costs little and protects your budget.
- Week one: prove delivery. Run a small budget, place test orders, and confirm each one reaches OpenAI Ads with the right amount and line items. Do not judge campaign results yet.
- Week two: check capture. Look at how many real orders carry a click reference. If most do not, check that the browser script is installed on the landing pages ChatGPT ads point to, and whether your consent banner is blocking the script for most visitors.
- Then: compare, carefully. Set the channel beside your others using the same definitions, such as orders in your store credited to each channel under one attribution model, rather than each platform's self-reported totals.
- Only then: move budget. Raise spend in steps and watch your store revenue, not just the ad report.
None of this is specific to OpenAI. It is the same discipline you would apply to any channel where the reporting is young and the match signals are thin, and skipping it is how teams end up scaling budget on numbers they have never checked.
Limits you should plan around
- Thin matching. No phone number is sent, and the email hash is consent-gated. Do not expect Meta-level match quality.
- Refunds are not synced. Cancellations are recorded in SyncBeacon but not passed to ad platforms as revenue adjustments.
- Moving platform. OpenAI Ads is young. Reporting definitions, attribution windows and API capabilities can change, so re-check your setup periodically.
- Apples and oranges. OpenAI's attribution rules are its own. Do not compare its conversion counts directly to Meta's or Google's without adjusting for differences.
- Not a promise of results. Tracking shows what happened. It does not make campaigns perform.
Beyond conversion counts
Once events flow, two things help make the channel comparable with the rest. First, the Beacon Attribution Engine (Growth and Enterprise plans) can show ChatGPT Ads next to your other channels, with spend pulled in from the connected OpenAI account. Read the figures as attribution-adjusted, not exact. The attribution guide explains the models.
Second, if you run product campaigns on OpenAI Ads, SyncBeacon can pause a linked OpenAI campaign when a product goes out of stock. It is implemented against the Ads API capabilities available to us, and works through the same product-to-campaign mapping described in the out-of-stock ad pause guide.
Related reading
- OpenAI Ads conversion tracking (setup page)
- OpenAI Ads for ecommerce: why ChatGPT is a real ad channel
- Meta Conversions API without a developer
- Ecommerce conversion tracking
Get started
Get ChatGPT Ads orders measured
Connect your store and OpenAI Ads account, then check delivery per event in the dashboard.
This article does not claim OpenAI approval, partner certification or a marketplace listing. OpenAI Ads capabilities can change. Not legal advice; consent obligations depend on your jurisdiction.