How to track ChatGPT and AI traffic in Google Analytics 4, step by step

AI engines are sending real visitors to real websites, and GA4 can show you yours: which engines, which landing pages, and whether those visitors convert. The setup takes about ten minutes. This guide walks through it step by step, gives you the exact regex to use, and is honest about the blind spots, because GA4 systematically undercounts AI traffic and you should know why before you read the numbers.

How AI traffic shows up in GA4

When someone clicks a link in an AI answer, the visit arrives in GA4 as a referral, with the engine’s domain as the session source: chatgpt.com (or the older chat.openai.com), perplexity.ai, copilot.microsoft.com, gemini.google.com, claude.ai. ChatGPT also appends utm_source=chatgpt.com to many outbound links, which makes those sessions easy to isolate.

GA4 does not group these into an “AI” channel by default; they are scattered through your Referral channel alongside every other website that links to you. The job, therefore, is to (1) find them, and (2) give them a channel of their own so they show up in your standard reports from now on.

Step 1: check what AI traffic you already have

  1. Open Reports, then Acquisition, then Traffic acquisition.

  2. Change the primary dimension to “Session source / medium” (click the dimension dropdown above the table).

  3. Search for chatgpt in the table’s search box. Repeat for perplexity, copilot, claude and gemini.google.

  4. Widen the date range to the last 90 days. AI referral volumes are small for most sites; a week of data often rounds to nothing.

If rows appear, you have AI traffic. Note which engines and roughly how much: this is your baseline.

Step 2: build a comparison with one regex

For an ad-hoc view across all AI engines at once, use a filter with a regular expression:

  1. In Traffic acquisition, click Add filter.
  2. Dimension: Session source. Match type: matches regex.
  3. Paste this expression:
chatgpt\.com|chat\.openai\.com|perplexity\.ai|copilot\.microsoft\.com|claude\.ai|gemini\.google\.com

Apply it, and the table shows only AI-referred sessions: users, engaged sessions, conversions, revenue if you track it. Add “Landing page” as a secondary dimension to see which pages AI engines actually send people to; they are frequently not your homepage, and often not the pages you would have guessed.

Step 3: create a permanent AI channel

To stop repeating step 2 forever, give AI referrals their own channel:

  1. Open Admin, then Data display, then Channel groups.

  2. Create a new channel group (GA4 will not let you break the default one, which is fine; a copy is safer anyway).

  3. Add a new channel, name it something like “AI referrals”.

  4. Set the condition: Source matches regex, and paste the same expression from step 2.

  5. Drag your new channel above “Referral” in the ordering. Channel rules apply top-down, so AI sources must be claimed before the generic Referral rule catches them.

  6. Save. From now on, every standard report that offers channel grouping can show “AI referrals” as its own line.

One quirk to expect: custom channel groups apply to data going forward and to historical data in reports, but attribution nuances mean small differences against the ad-hoc filter. Directionally they will agree.

The honest caveats: why GA4 undercounts AI traffic

Read your AI numbers as a floor, not a ceiling, for three reasons:

So a GA4 report saying “AI sends us 40 sessions a month” really means “at least 40 sessions arrived with an AI referrer”. The true number of people influenced by AI answers about you is larger, and GA4 cannot see it.

Clicks are the lagging indicator; citations are the leading one

This is the structural limit of analytics for AI search: GA4 shows clicks after the fact. By the time an AI-referred visit lands, the engine already decided, days or weeks earlier, to cite you rather than a competitor. And when it decides the other way, GA4 shows you nothing at all: there is no report for the click that went to someone else.

That earlier decision is measurable, just not in GA4. The SearchScore Tracker asks six engines (ChatGPT, Claude, Gemini, DeepSeek, Grok and Perplexity) the questions your customers actually ask, every week, and records which engines cited you, at what position, with what sentiment, and who got named instead when you were left out. GA4 tells you what the citations you already earn are worth; the Tracker tells you whether you are earning them. The two together give you the full funnel: cited, clicked, converted.

Frequently asked questions

How do I see ChatGPT traffic in Google Analytics?

Open Reports, then Acquisition, then Traffic acquisition, set the dimension to Session source / medium and search for “chatgpt”. Clicks from ChatGPT arrive as referrals from chatgpt.com (or the older chat.openai.com), and many carry utm_source=chatgpt.com. For all AI engines at once, filter Session source with a regex covering the engines’ domains.

Why does my AI traffic look so low in GA4?

Partly because it genuinely is small for most sites, and partly because GA4 undercounts it: clicks from AI apps and many free-tier sessions pass no referrer and land in Direct, and most AI answers never produce a click at all, even when they influence the buyer. Treat GA4’s AI referral number as a minimum.

What regex matches AI referral sources in GA4?

A practical expression covering the major engines is: chatgpt.com|chat.openai.com|perplexity.ai|copilot.microsoft.com|claude.ai|gemini.google.com, applied to the Session source dimension with the “matches regex” condition. Review it every few months, because engines change domains and link behaviour.

Can GA4 tell me how often ChatGPT mentions my brand?

No. GA4 only sees visits to your site, so it records a click after a citation has already happened, and records nothing when an engine cites a competitor instead. Measuring the citations themselves requires querying the engines directly, which is what a citation tracker such as the SearchScore Tracker does on a weekly cycle across six engines.

Related reading

Part of Pillar Article — see all guides in this series →