How to Track Leads From ChatGPT, Perplexity, Gemini, and Google AI Overviews in 2026

You track leads from ChatGPT, Perplexity, Gemini, and Google AI Overviews with three layers working together: GA4's AI Assistant channel and a custom channel group for referral visits, a required "How did you hear about us?" question on every form and call, and CRM fields that store both answers. No single layer is complete, because much AI-influenced traffic arrives with no referrer.

This is the question I get right after "how do I show up in ChatGPT." Owners start doing the work, and then they open Google Analytics and see a few dozen sessions from chatgpt.com and wonder if any of it matters. Meanwhile their front desk keeps hearing "I asked ChatGPT and it said you were good." Both things are true. The analytics is undercounting, and the fix is to stop relying on analytics alone.

Why is AI traffic so hard to see?

• Zero-click answers: the assistant answers the question and names you, and the person calls you directly or searches your name later. No session from the assistant ever happens.
• Missing referrers: mobile apps, privacy settings, and copy-pasted links often strip the referrer, so the visit lands in Direct.
• Inconsistent tagging: ChatGPT appends utm_source=chatgpt.com to many links it cites, but analyses this year found plenty of links in its newer answer formats without it.
• Google's AI features count as Google: clicks from AI Overviews and AI Mode come from google.com and are classified as Organic Search, mixed in with regular blue-link clicks.

So the goal is not perfect attribution. It is a defensible estimate that tells you whether your AI visibility work is producing customers. If you have not started that work yet, read how local businesses get visible in AI search first.

What changed in GA4 in 2026?

Google added an "AI Assistant" channel to GA4's default channel grouping. Industry coverage dates the announcement to May 2026, with a gradual rollout to properties over the following weeks. Google's traffic-source documentation describes it as visits arriving from sources like ChatGPT, Gemini, DeepSeek, Copilot, or Grok, and states that it excludes Google's AI Overviews and AI Mode. When GA4 recognizes one of these referrers, it sets the medium to ai-assistant and groups the session into that channel automatically.

Two practical notes. It is not retroactive, so anything before your property got the channel stays in Referral. And Google has not published the full list of recognized referrers. Some reports found Perplexity and Claude visits still landing in plain Referral. That is why I still set up a custom channel group.

How do I set up GA4 to catch AI referrals?

1. Check the built-in channel

In GA4, open the Traffic acquisition report and look for AI Assistant in the Session default channel group. If it is there, you are getting the built-in classification. Then switch the dimension to Session source to see which assistants are actually sending traffic.

2. Add a custom channel group as a safety net

Under Admin, Data display, Channel groups, copy the default group and add a channel called "AI Search" above Referral. Set the condition to Session source matching a regex like chatgpt|openai|perplexity|gemini|copilot|claude|deepseek|grok|meta.ai and the medium not equal to cpc. Order matters: it must sit above Referral or those sessions get claimed first. This catches assistants the built-in list misses, and you can apply it to historical data for comparison.

3. Mark the right key events

Sessions are vanity. Make sure form submissions, tap-to-call clicks, booking confirmations, and chat starts are marked as key events, so you can see AI Assistant conversions, not just visits.

4. Extend data retention

GA4 defaults event-level data retention to 2 months, which limits how far back explorations go. Change it to 14 months under Admin, Data collection and modification, Data retention. Otherwise you will not be able to compare this quarter to last year.

What can I do about Google AI Overviews and AI Mode?

GA4 cannot separate these clicks from regular organic search. Search Console now helps a little: Google began rolling out generative AI performance reporting in mid-2026 that shows impressions from AI Overviews and AI Mode by page, though industry coverage reports it does not include clicks or queries. Use it as a visibility signal. If a service page gains AI Overview impressions and organic leads to that page rise in the same period, that is reasonable evidence. Pair it with self-reported data, below, because people who found you in an AI Overview will often say "Google" when asked.

Do UTM parameters work for AI traffic?

Partially, and only for links you control. You cannot tag the links an assistant decides to cite. ChatGPT's own utm_source=chatgpt.com tag helps when it is present, but it is not consistent. Where UTMs do help: links you place on your Google Business Profile website button, in directory listings that AI tools read heavily, and in profiles you control. Tagging your Google Business Profile link, for example with utm_source=google and utm_medium=organic and utm_campaign=gbp, separates Maps and profile visits from regular organic search, which matters because a lot of AI answers send people to your profile first.

One warning: never put UTMs on internal links within your own site. It overwrites the original source and wrecks the very attribution you are trying to build.

Why "How did you hear about us?" is the most important field

Self-reported attribution is the only layer that catches the zero-click recommendation, which is the most common way AI search sends local customers. Someone asks Gemini for a pediatric dentist, sees your name, and calls. Analytics sees nothing. The caller will tell you if you ask.

• Make it required on every web form, as a dropdown, with "AI assistant (ChatGPT, Gemini, Perplexity, etc.)" as its own option alongside Google search, Google Maps, referral, social, TV or radio, and other.
• Add an optional free-text "anything else?" box. The specific answers ("ChatGPT said you were the only one open Saturday") are gold for content planning.
• Ask on every phone call and record it in the CRM before the call ends. If an AI receptionist answers your phone, have it ask the question and write the answer to the contact record.
• Ask again at booking or onboarding for anyone who skipped it, since the person who chose you is not always the person who filled out the form.

Expect messy answers. Plenty of people will say "Google" when they mean an AI Overview, or "online." That is fine. You are looking for trends over months, not courtroom evidence.

How should I set up lead source fields in my CRM?

Store the two kinds of attribution separately and never let one overwrite the other. Analytics data tells you what the browser saw. Self-reported data tells you what the person remembers. The gap between them is the insight.

The first three are captured with a few lines of code that read the referrer and UTMs on the first visit, store them, and pass them into hidden form fields. Any decent custom website build should include this. The CRM side, with source fields tied to pipeline stages and revenue, is standard in the CRM systems we set up.

What should the monthly AI attribution report look like?

• AI Assistant sessions and key events from GA4, by assistant, compared to last month.
• Leads, booked jobs, and revenue where first touch source was an AI assistant.
• Leads, booked jobs, and revenue where the self-reported source was an AI assistant, including the ones analytics marked as Direct or Organic.
• AI Overview and AI Mode impressions from Search Console for your top service pages.
• Your citation check: whether the main assistants name you for your top 10 buyer questions, a habit covered in my GEO checklist.

In practice, the self-reported number is usually several times larger than the analytics number for local service businesses, because so much of the influence is zero-click. If you only looked at GA4, you might conclude AI search does not matter for you. Your phone log may tell a different story.

We build this reporting into client dashboards through our analytics and reporting services, so AI-sourced leads, booked jobs, and revenue sit next to every other channel. If you want help wiring up GA4, your forms, and your CRM so the numbers line up, book a call with me.