How to find ChatGPT traffic in GA4 (and other AI assistants)

AI assistants send traffic that’s easy to miss. It’s spread across several sources, labelled as ordinary referrals, and sometimes missing a referrer altogether. But it’s often high-intent: someone who clicks a source in a ChatGPT answer about tools in your category is already comparing options.

This guide shows the quick check, then the permanent setup, then how to read the numbers.

The two-minute check

1. In GA4, open Reports → Acquisition → Traffic acquisition.
2. Click the dimension dropdown above the table (it usually says “Session primary channel group”) and choose Session source / medium.
3. Type chatgpt in the search box above the table.
4. Set the date range to the last 90 days and compare with the previous period.

You’ll usually see chatgpt.com / referral. ChatGPT adds utm_source=chatgpt.com to many of the links it cites, which is why the source is labelled so clearly. Repeat the search for perplexity, copilot, gemini and claude.

AI assistant sources to look for

See all AI traffic at once with a regex

Instead of searching one by one, use an exploration with a regex filter:

1. Go to Explore → Free form.
2. Add the dimensions Session source and Landing page + query string, and the metrics Sessions, Engaged sessions and Key events.
3. Drag Session source into Rows.
4. Add a filter: Session source → matches regex → paste the expression below.

.*(chatgpt\.com|chat\.openai\.com|perplexity\.ai|copilot\.microsoft\.com|gemini\.google\.com|claude\.ai).*

Add new assistants to the list as they appear in your referral report. Keep the regex in a shared document so everyone uses the same one.

Make it permanent: an AI channel group

Explorations are good for digging, but a custom channel makes AI traffic show up in every standard report next to Organic Search, Paid and Email. GA4 lets you do this with a custom channel group:

1. Go to Admin → Data display → Channel groups. You need Editor access or above.
2. Click Create new channel group. GA4 starts you with a copy of the default group.
3. Name it, for example “Channels with AI”.
4. Click Add new channel, name it “AI assistants”, and set the condition: Source → matches regex → the expression above.
5. Click Reorder and drag “AI assistants” above “Referral”. Channels are matched top to bottom, so it must come first or AI visits will be counted as Referral.
6. Click Save group.

Google says custom channel groups can be applied to your reports retroactively, so you’ll see historical AI traffic too. Standard properties can have two custom channel groups in addition to the default one.

To use it, open Traffic acquisition, click the dimension dropdown and pick your new channel group instead of the default.

Find which pages AI sends people to

Knowing the total is useful. Knowing the landing pages is what tells you what to do next. In your exploration, put Landing page in Rows and keep the AI regex filter. You’ll typically see three kinds of page:

Compare AI visitors with everyone else

A comparison makes the numbers easier to read. In any standard report, click Add comparison at the top, choose Session source, set it to match your AI regex, and save. Add a second comparison for “Session default channel group exactly matches Organic Search”. Every report you open, from landing pages to conversions, now shows AI visitors and organic visitors side by side.

If you use Looker Studio, connect your GA4 property and create a chart using your custom channel group as the dimension. Filter it to the “AI assistants” channel and add sessions, engaged sessions and key events. A single page with that chart, a landing-page table and a month-over-month comparison is enough for most teams.

AI visits showing up as Direct or Unassigned

If your AI channel looks smaller than expected, some visits may be landing elsewhere:

• Direct: clicks from desktop and mobile apps sometimes arrive without a referrer. Watch for spikes in Direct traffic to pages that AI cites, such as pricing or a specific guide.
• Unassigned: visits with a UTM source but no medium can land here. Check Session source / medium for chatgpt.com entries without “referral”.
• Referral: if the AI channel isn’t above Referral in your channel group, AI visits still show here.

You can’t recover every lost referrer, but you can spot patterns. A page that suddenly gets more Direct traffic at the same time as it starts appearing in ChatGPT answers is a strong hint.

Measure quality, not just volume

AI traffic is usually small compared with Google, so judge it by what it does. Compare the AI assistants channel with Organic Search on:

• Engagement rate: are AI visitors reading?
• Key events per session: signups, demo requests, purchases.
• Landing pages: are they arriving on buying pages or general guides?

Add a “How did you hear about us?” field with an “AI assistant (ChatGPT, Perplexity, etc.)” option to signup and demo forms. It catches people who read about you in an AI answer and came back later through Google or by typing your URL, which GA4 can’t attribute.

Track conversions from AI visitors

Traffic only matters if it turns into customers. Make sure your important actions, such as signups, demo requests, trials and purchases, are marked as key events in Admin → Data display → Events. Then, in your exploration, add Key events and Session key event rate as metrics alongside sessions.

Look at the rate, not the count. A channel with 300 sessions and a 4% key event rate is worth more attention than one with 3,000 sessions and 0.5%. If AI visitors convert well, that’s the strongest argument for investing in AI visibility, and the number leadership will ask for first.

For B2B companies with longer sales cycles, also pass the first-touch source into your CRM. A deal that started with a ChatGPT visit three months ago won’t show up in GA4’s conversion reports, but it will in the CRM if the source was captured at signup.

What GA4 can’t show you

• AI Overviews and AI Mode clicks. Google counts these inside overall search traffic, so in GA4 they look like google / organic.
• Visits with no referrer. Some clicks from AI apps arrive without a referrer and land in Direct, especially from desktop and mobile apps.
• People who read but don’t click. Most people don’t click sources at all: in Pew Research’s study, people clicked a link inside an AI summary on just 1% of visits that showed one.
• Mentions without links. If ChatGPT names you without citing your site, there’s nothing to click.

So treat GA4’s AI channel as a floor, not the full picture. Pair it with Bing’s AI Performance report for Copilot citations, and a monthly prompt check in ChatGPT and Perplexity for mentions. Our guide to citations vs brand mentions shows how to report them together.

Check your server logs too

GA4 only sees visitors who load your page in a browser and run its tracking script. Your server logs see something GA4 can’t: AI assistants fetching your pages for users. When someone asks ChatGPT about a page, a request from ChatGPT-User often appears in your logs. Perplexity-User works the same way.

grep -E "ChatGPT-User|Perplexity-User" access.log \
  | awk '{print $7}' | sort | uniq -c | sort -rn | head -20

That list shows which pages AI assistants read on behalf of people, even when nobody clicks through. It’s a useful complement to GA4’s landing-page report. If you’re setting up crawler access at the same time, our guide to OpenAI’s crawlers explains which bot is which.

Reading the numbers: an example

Say your AI channel shows 400 sessions last month, up from 250. ChatGPT sent 280, Perplexity 90, Copilot 30. Engagement rate is 68%, against 52% for Organic Search. The top landing pages are your comparison page, your pricing page and a guide on choosing tools in your category. Twelve signups came from the AI channel, and another eight people chose “AI assistant” in the signup form.

That tells you AI visitors are fewer but more engaged than search visitors, they arrive on buying pages, and the real influence is larger than GA4 alone shows. The action: keep the comparison and pricing pages current, and create similar pages for the next two competitors buyers ask about. Our ChatGPT citation playbook covers how.

Set up an alert for changes

AI traffic can change quickly when a page starts or stops being cited. GA4’s Insights can notify you. Go to Reports → Reports snapshot, open the Insights card, choose Create, and set a custom insight: a weekly check where sessions for your AI regex source change by more than, say, 30%. Have it emailed to the people who own your key pages, so changes are noticed quickly.

A sudden drop is usually worth investigating the same week. Check that robots.txt and your CDN still allow AI crawlers, that the landing page still loads its content in HTML, and whether a competitor’s page has replaced yours in the answers for that topic.

A monthly AI traffic report

AI assistant traffic, [month]

Sessions:        [n] (last month [n], [+/-]%)
  ChatGPT [n] | Perplexity [n] | Copilot [n] | Gemini [n] | Claude [n]
Engagement rate: [x]% (Organic Search: [x]%)
Key events:      [n] signups / demos from AI channel
Top landing pages: [page] [n], [page] [n], [page] [n]
"Heard about us via AI" (form): [n]
Notes: [new pages cited, drops, actions next month]

Common mistakes

• Searching only “openai”. Most ChatGPT traffic now shows as chatgpt.com.
• Forgetting other assistants. Perplexity, Copilot, Gemini and Claude all send visits.
• Changing the regex without telling anyone. Reports stop matching month to month.
• Expecting AI Overviews in GA4. They’re in Google organic, not a separate source.
• Skipping the “How did you hear about us?” field. It catches the visits analytics can’t attribute.