Measurement
How to track leads from ChatGPT, Perplexity and Google’s AI Overviews
Visits from AI assistants often don’t look like what they are. Here’s how to see as much of that traffic as you can, and how to fill the gaps analytics leaves.
insightfig4 min read
Key takeaways
- Some AI visits arrive with a source you can segment in GA4. Many don’t, and are counted as Direct.
- Google counts AI Overviews and AI Mode in ordinary Search Console data. Separate AI reporting is still rolling out and shows impressions only.
- A “How did you hear about us?” field with AI options is the simplest way to catch what analytics misses.
- Judge AI search on qualified inquiries, using several imperfect signals together.
When a prospect finds your business through an AI assistant, the visit often doesn’t look like it. Some arrive labelled with the assistant’s name. Others arrive with no referrer at all and are counted as Direct, as if the person had typed your address. Clicks from Google’s AI Overviews look like ordinary Google organic traffic. If you only look at default reports, AI search is mostly invisible.
Here’s how to see as much of it as you can, and how to fill the gaps. The details reflect how these tools behave as of October 2026. They change often, so check them against your own data.
1. Group the AI referrals GA4 can identify
When an assistant passes a referrer or tags its links, the visit carries a recognizable source, such as chatgpt.com, perplexity.ai, copilot.microsoft.com, gemini.google.com or claude.ai. Don’t rely on GA4’s default channels to catch all of them. Define the group yourself:
- 01In GA4, go to Admin, then Data display, then Channel groups, and create a new channel group by copying the default one.
- 02Add a channel called “AI assistants” with the condition Session source matches regex, using a pattern such as
chatgpt\.com|openai\.com|perplexity\.ai|copilot\.microsoft\.com|gemini\.google\.com|claude\.ai. - 03Move the new channel above Referral, so these sessions aren’t absorbed into it. GA4 applies channel rules in order.
- 04Use the channel group in your acquisition reports and explorations, and make sure your form submissions, calls and bookings are marked as key events, so you can see which AI visits became inquiries.
Review the pattern every few months. New assistants appear, and existing ones change domains and link behaviour.
2. Know what Search Console can and can’t tell you
Google counts impressions and clicks from AI Overviews and AI Mode within your normal Search Console performance totals, and the clicks reach your site as Google organic traffic. In 2026 Google began rolling out separate generative AI reporting in Search Console, but it shows impressions only, and not every property has it yet.
Treat it as a visibility signal: which pages appear in AI features, and whether that’s growing. Don’t expect it to tell you which questions led to an inquiry. Bing Webmaster Tools is worth checking too, since Microsoft Copilot draws on Bing’s index.
3. Ask people how they found you
The most reliable way to catch AI-driven inquiries is also the simplest: ask. Add a “How did you hear about us?” field to your contact and booking forms, with options such as “ChatGPT or another AI assistant” and “Google search”, plus a free-text option.
Self-reported answers aren’t perfect. People forget, and they tend to name the last thing they did. But for considered purchases they regularly surface sources analytics can’t see: an assistant’s recommendation, a podcast, a colleague. Store the answer with the lead in your CRM, so you can compare it with what analytics recorded.
4. Check what the assistants actually say
Analytics shows the visits you received, not the answers where you were missing. Keep a fixed list of 20 to 50 questions that matter to your business, written the way prospects ask them. Each month, put them to the assistants your prospects use and record whether your business is mentioned, cited or recommended, who appears instead, and which sources the answer relies on.
Answers vary from one run to the next, so look at the trend across the whole list rather than any single answer. Specialist tools can automate this at scale. For most businesses, a consistent manual check is enough to start.
| Source | How it usually appears | What to use |
|---|---|---|
| AI assistants that pass a referrer or tag links | A named source, such as chatgpt.com | A custom “AI assistants” channel in GA4 |
| AI assistants that pass nothing | Direct traffic | A “How did you hear about us?” field |
| Google AI Overviews and AI Mode | Google organic traffic | Search Console, including AI reporting where available |
| Answers where you don’t appear | Nowhere in analytics | A monthly check of your priority questions |
5. Connect it to inquiries and clients
Traffic from AI assistants only matters if it produces business. Connect form submissions, calls and bookings to your CRM, record what happened to each inquiry (qualified, booked, won or lost), and report AI search the same way as every other channel: qualified inquiries, consultations and clients won.
None of these signals is complete. Read together, they’re enough to decide whether AI search deserves more investment. If you’d like help setting this up, see Measurement & Analytics, or start with the basics in AEO and GEO explained.