ChatGPT Ads conversion tracking for B2B SaaS: clicks to revenue

OpenAI added cost-per-click buying to ChatGPT Ads on May 5, 2026. Its words: "Advertisers are only charged based on a click outcome" (OpenAI).

A click is not a customer, though. For B2B SaaS, a trial, a call and a first invoice sit in between. This guide covers ChatGPT Ads conversion tracking from click to money, as of November 2026.

How do ChatGPT Ads work?#

OpenAI's own pages say this, as of November 2026.

Three details matter for B2B SaaS.

• Context hints are not keywords. OpenAI says "Hints are not exact-match keywords."
• Ads don't buy answers. OpenAI says "Ads do not influence the answers ChatGPT gives you." Earning a mention is a separate job. See how to rank in ChatGPT.
• You never see the chat. Advertisers get aggregate data only, not "your chats, chat history, memories, or personal details" (OpenAI). Your landing page and billing data have to tell the story.

What is ChatGPT Ads conversion tracking?#

ChatGPT Ads conversion tracking is the process of linking each ad click to the sign-ups and payments it led to.

OpenAI gives you two senders (OpenAI Ads docs):

1. The Measurement Pixel. A script in your page <head> sends browser events with oaiq("measure", ...).
2. The Conversions API. Your server sends events to OpenAI with an API key.

For SaaS, the server matters most. Renewals and invoices never touch a browser.

OpenAI's standard events include trial_started, registration_completed, lead_created, subscription_created and order_created (OpenAI Ads docs). Here is how I'd map them:

Why order_created? OpenAI says its ROAS "depends on the values sent with order_created events" (OpenAI Ads docs).

Capture oppref and UTMs on the landing page#

The oppref value#

oppref is OpenAI's click value. The docs call it "An opaque, OpenAI-provided attribution identifier. Pass the original string without modification" (OpenAI Ads docs).

The Pixel "captures oppref from the landing page URL." It stores it "in a first-party __oppref cookie" for later page views (OpenAI Ads docs).

The server side is your job. The Conversions API doesn't capture oppref for you. So:

1. Read oppref on arrival. Take it from the URL or the __oppref cookie.
2. Save it at sign-up. Store it on the account record.
3. Copy it into Stripe. Put it in the customer's metadata with your UTMs.
4. Read it in the webhook. Send it back with the paid event, unchanged.

Watch for redirects that drop query strings.

UTMs, added by you#

OpenAI's ad reference names the landing field creative.target_url. It says nothing about tracking or UTM parameters (OpenAI Ads docs). So add your own tags to the URL, for example:

?utm_source=chatgpt&utm_medium=paid&utm_campaign=crm-switchers

Use utm_medium=paid to separate ad clicks from organic ChatGPT referrals. Our Stripe UTM tracking guide shows how to carry UTMs into Stripe.

How do you send conversions with the Conversions API?#

Everything below is from OpenAI's Conversions API and conversion tracking pages, as of November 2026.

Endpoint. Events go to https://bzr.openai.com/v1/events?pid=${PIXEL_ID} with a Bearer key. You create the key through api.ads.openai.com.

Required fields. Each event needs id, type, timestamp_ms and data. Web events also need source_url. action_source is "conditionally required," with values such as web, offline and phone_call.

Amounts. Send money "as integers in the standard ISO 4217 minor unit." OpenAI's example is 12999 for $129.99 (OpenAI Ads docs). Stripe also stores amounts in minor units (Stripe Docs). So Stripe's amount_paid can go across as is. Don't divide by 100 here.

Timing. The timestamp must be "within the last 7 days." Send the paid event when the webhook fires, not in a weekly batch.

Hashing. Identifiers that point to a person must be SHA-256 hashed and sent as "a lowercase, 64-character hexadecimal string."

Fields like countries, ip_address and user_agent go unhashed.

Deduplication. It keys on Pixel ID, event name and id. OpenAI keeps the first event and "ignores later duplicates." Use one value for the Pixel's event_id and the server's id, and reuse it on retries.

Batches. Up to 1,000 events per call, but "If one event in the batch fails, the full batch fails."

Testing. Add "validate_only": true to check a request without saving it. SaaS marketing attribution explains why paid events beat sign-ups.

Judge ChatGPT Ads ROAS on Stripe revenue#

OpenAI's ROAS is useful for bidding, not for your profit and loss. The click window is 7, 14 or 30 days, with 30 as the default. The view-through window is 0 or 1 day, defaulting to 1 (OpenAI Ads docs). Conversions update in daily processing, so allow "at least one day."

Three reasons the platform number will differ from Stripe:

• View-through credit. The 1-day view window counts people who never clicked. Set it to 0 for click-only numbers.
• The 30-day window. A deal that closes on day 45 shows in Stripe, not in the platform.
• Renewals. Send only the first payment, and platform ROAS ignores month two onward.

So build a second number from Stripe:

A worked example (made-up numbers). Assumption: $3,000 spend, five paying customers on a $200 monthly plan. First-payment ROAS is $1,000 ÷ $3,000, or 0.33. If all five stay three months, 90-day ROAS is 1.0. Compare both with your margin. What is a good ROAS? shows how to set break-even.

What's still unknown#

As of November 2026, OpenAI's pages leave some gaps. Plan around them.

• Minimum budgets. Daily minimums "depend on the account currency." No number is published (OpenAI Ads docs).
• Self-serve access. Ads Manager is in beta and opening "gradually." Your account may not have it yet.
• How long oppref lives. The docs don't state a cookie lifetime.

A small-budget test plan#

I'd treat the first month as a measurement test, not a growth bet.

1. Week 0: wire it up. Install the Pixel and server sender. Test with validate_only. Buy your plan in Stripe test mode and check that oppref and UTMs reach the webhook.
2. Pick one offer. One chat_card campaign, one landing page, context hints that describe the problem you solve.
3. Set a cap you can lose. A daily budget near your account's minimum, and a total fixed in advance.
4. Optimize for the deepest event with volume. If payments are rare, start with trial_started. Keep sending order_created.
5. Note your baseline. Record organic ChatGPT sign-ups for the four weeks before. A marketing incrementality test helps.
6. Decide on Stripe. After the click window closes, compare first-payment ROAS with break-even. Scale, change the offer, or stop.

Where Kratly fits#

Kratly has a ChatGPT Ads connection built and in testing: spend and results by day next to Stripe revenue, and verified first payments sent back as conversions, with hashed email sent only if you allow it. Until it's verified on a live account, tag ChatGPT clicks with utm_source=chatgpt and Kratly shows them next to their payments, which gives you the Stripe-based ROAS view above.