Cost Per Booked Call: How to Measure It for ChatGPT Ads

Clicks are cheap to count and easy to misread. For any business that sells through a conversation (plumbers, law firms, dentists, agencies, B2B software), the number that actually tells you whether ChatGPT Ads is working is cost per booked call: how much you spend to get one real appointment on the calendar.

This guide shows how to calculate it, how to set a target you can afford, and the mistakes that quietly make the number wrong.

Cost per booked call = ad spend ÷ booked calls from those ads

If a campaign spent $1,200 last month and produced 16 booked calls, your cost per booked call is $75.

Simple. The hard part is making sure both numbers are honest.

A campaign with cheap clicks and lots of form fills can still lose money if few of those leads ever book. Cost per booked call cuts through that.

Write it down before you measure anything. A booked call should mean an appointment that exists on a calendar with a real person: a scheduled consultation, a service visit, a sales call or a demo.

Don't count calendar holds you created yourself, duplicate bookings from the same person, or calls that were cancelled before they happened. If your CRM has stages, map "booked" to one stage and stick to it.

This is where most measurement breaks. A booked call usually happens hours or days after the click, often in your CRM rather than on your website.

1. Track the first action on your site (a form fill or booking) with OpenAI's Measurement Pixel.
2. Keep OpenAI's click reference (oppref) with the lead in your CRM.
3. Send the booking back to OpenAI through the Conversions API as appointment_scheduled, within 7 days.

We walk through the setup step by step in ChatGPT Ads conversion tracking.

Match the spend to the same date range and the same campaigns, ad groups or context hints you're measuring. If you're comparing ad groups, use each ad group's own spend, not the campaign total split evenly.

Your target cost per booked call comes from your economics, not from a benchmark:

Maximum cost per booked call = average job value × close rate × the share of revenue you're willing to spend on ads

Example for a plumbing company:

At $60 per booked call, this campaign is profitable. At $120, it isn't, even if the clicks look cheap.

Once you trust the number, compare it across your structure:

• Campaign level: is ChatGPT Ads worth it at all?
• Ad group level: which customer needs book calls cheaply?
• Context hint level: if each key hint has its own ad group (the Big Hint Campaign method), you can compare hints directly.

Move budget toward what books calls below your target, and rewrite or pause what doesn't.

1. Counting leads as bookings. A form fill is not a booked call.
2. Ignoring the delay. Bookings arrive days after clicks, and OpenAI says attributed conversions can take 24 to 48 hours to show. Judge last week, not yesterday.
3. Tiny samples. Two booked calls at $40 each isn't a trend. Wait for at least 5 to 10 before making big moves.
4. Double counting. The same booking sent from both the Pixel and the Conversions API without a shared event ID gets counted twice.
5. Stopping at the booking. The next step is cost per paid job. A cheap booked call that never closes isn't cheap.

Big Hint connects ChatGPT Ads to the leads and booked calls in your CRM and shows cost per booked call by campaign, ad group and context hint, with what to change next.

• OpenAI Developers: Conversions API
• OpenAI Help Center: Troubleshooting conversion tracking for ChatGPT Ads

The plumbing example is illustrative. Last reviewed October 27, 2026.