fm.

How to track ChatGPT and AI traffic in GA4: one saved report, and the form question that found $960K

GA4 can show you the visits AI assistants send. It can't show you the buyers who asked an assistant and then searched your name. The report I ran at Pipe17, its numbers and the form question that covered the gap.

Key takeaways

  • To track ChatGPT and AI traffic in GA4, save one report filtered on session source for the assistants you care about, and read it by source and by landing page.
  • Take a baseline first. My first 90 days at Pipe17 showed 343 sessions from AI assistants, 203 of them from ChatGPT, with 8 conversions.
  • GA4 can't show the prompt, and it only sees buyers who click through from the assistant. Many don't.
  • One question on the form covered the gap. Read against CRM source, it showed AI-referred buyers were $960K in closed-won revenue, 32% of new-business wins, all of it credited to paid search.

01How to track ChatGPT and AI traffic in GA4

To track ChatGPT and AI traffic in GA4, build one saved report that filters session source for the AI assistants, then break it down by source and by landing page. Rerun it on a schedule, and pair it with a "how did you hear about us" question on your forms, because GA4 only counts the buyers who clicked a link inside the assistant.

This note is the report and the form question. The wider measurement, including prompt panels, citation share and a custom channel group for AI assistants, is in how to measure AI visibility, and I won't repeat it here.

02The saved report

  1. Filter hostname to the marketing site.If an app, a help center or another site shares the property, their sessions will blur the picture. Keep only the site buyers land on.
  2. Filter session source.Mine matched session source containing chatgpt, openai, perplexity, claude, copilot, gemini and a few others. Match on a fragment of the name, because the same assistant can arrive under more than one domain.
  3. Report by source.Sessions, engaged sessions and conversions for each assistant. This is the table that tells you which assistant matters for your category.
  4. Report landing pages by source.The same filter, broken down by landing page. This is the view you act on.
  5. Schedule it.A scheduled job reran my report on the first of each month and emailed it to marketing leadership with the change since the last run. A report nobody has to remember to open is the one that gets read.

03The baseline from my work at Pipe17

Pipe17 sells an OrderOps platform to commerce brands and 3PLs. The first 90 days of the report showed 343 sessions from AI assistants and 8 conversions.

SourceSessions in the first 90 days
ChatGPT203
Gemini48
Perplexity44
Claude35
Copilot11
All AI assistants343, with 8 conversions

That was about 2.7% of Google organic sessions on the same site. On its face it's a small channel. By my last monthly report, integration pages were the strongest group of AI-referral landing pages after the homepage, which told me what the assistants were citing the site for: specific questions about connecting one system to another.

04What GA4 can't tell you

  • The prompt. GA4 sees the visit and never the question. The landing page is the best proxy you have for what the assistant cited you for.
  • The medium can move. GA4 later started filing this traffic under its own medium. A report keyed on medium equals referral will undercount, which is one more reason to filter on source.
  • Buyers who don't click. Someone who reads an answer, opens a new tab and types your name arrives as brand search or direct. GA4 has no way to connect that visit to the assistant.

The third limit is the big one, and it's where the money was.

05The form question that found $960K

Paid search looked like the best revenue channel in the company on paper. When I read the click paths, 14 of 15 paid-search sales-qualified opportunities had first arrived by searching for the company's own name. So where had they heard the name?

The forms asked buyers how they heard about us. I read those answers against the source recorded in the CRM.

  • The share of respondents naming an AI assistant had gone from 6.6% to 16.5% in a year.
  • Four opportunities carried a written answer naming an assistant. All four were logged as paid search.
  • The ones that closed came to $960K in closed-won revenue. That was 32% of new-business wins, and 80% of the closed-won revenue paid search had been given credit for.
  • HubSpot's own AI referral source had captured 7 contacts and no deals.

The path is easy to follow once you see it. An assistant recommends the product. The buyer searches the brand name, clicks the brand ad and fills in the demo form. Paid search gets the credit. The answer was filled in on only about 21% of sales-qualified opportunity deals, so the real figure is higher than what I could prove.

To run the same check, add the question to every form that creates a sales conversation, leave it as free text so people use their own words, and read the answers next to the CRM source every month. It's one field, and no tracking pixel would have found what it found.

06What I'd do with the numbers

Use the landing page view to decide which pages to improve, and the form answers to decide how much the channel is worth. I also designed a 20-page freshness test, ten refreshed pages against ten matched controls over six weeks, built on real edits instead of changed dates. I don't have a result from it to report.

How this view sits next to first-touch and influence reporting is in B2B marketing attribution for long sales cycles. What makes an assistant cite a page in the first place is in how to rank in ChatGPT. If you want this set up and tied to pipeline, that's part of my work as an AEO consultant, and the free AEO checker tests whether the assistants can find, read and quote your site.

FAQQuestions people ask

How do I track ChatGPT traffic in GA4?
Build one saved report filtered on session source. At Pipe17, Farhan Munshi filtered hostname to the marketing site, then session source containing chatgpt, openai, perplexity, claude, copilot and gemini, and reported sessions, engaged sessions and conversions by source, then landing pages by source. A scheduled job reran it monthly and emailed the change since the last run.
How much traffic do ChatGPT and other AI assistants send?
It depends on the site, so take a baseline before you judge it. Farhan Munshi's first 90 days of tracking at Pipe17 showed 343 sessions from AI assistants, 203 of them from ChatGPT, with 8 conversions. That was about 2.7% of the site's Google organic sessions. The session count understated what the assistants were worth.
Can GA4 show which prompt someone typed into ChatGPT?
No. GA4 sees the visit and its source. It can't see the prompt. Farhan Munshi uses the landing page as the best proxy for what an assistant cited a site for. At Pipe17, integration pages were the strongest group of AI-referral landing pages after the homepage by his last monthly report.
Why does AI-referred revenue show up as paid search or direct in the CRM?
Because many buyers don't click the assistant's link. At Pipe17, Farhan Munshi found the usual path was an assistant recommending the product, the buyer searching the brand name, clicking the brand ad and filling in the demo form. Paid search got the credit. HubSpot's own AI referral source had captured 7 contacts and no deals.
How do you measure revenue from AI assistants?
Ask buyers how they heard about you and read the answers against the recorded source. At Pipe17, Farhan Munshi found four opportunities whose written answer named an AI assistant, all logged as paid search. The ones that closed came to $960K in closed-won revenue, 32% of new-business wins. The question was filled in on only about 21% of opportunities.