More people now start their research in ChatGPT, Perplexity, Gemini, Copilot and Claude, and some of them click through to the sources those assistants cite. If you are working on answer engine optimization, you need to know whether that is happening for your site, which pages get the visits, and whether those visitors do anything useful. Google Analytics 4 can show you a lot of this, but not by default and not perfectly. This guide shows you how to track AI traffic in GA4 step by step, what the numbers can and cannot tell you, and how to turn the data into content decisions.
What AI Traffic Looks Like in GA4
When someone clicks a link inside an AI assistant, one of three things usually happens:
- The browser passes a referrer. GA4 records the assistant’s domain as the session source, for example
perplexity.ai / referralorgemini.google.com / referral. This is the easy case. - The assistant adds its own tag. ChatGPT adds
utm_source=chatgpt.comto many of the links it shows, so those visits arrive withchatgpt.comas the source. Depending on what else is passed, the medium may show asreferralor(not set). - Nothing is passed. Desktop and mobile apps, privacy settings and copy-pasted links often strip the referrer. These visits land in Direct, and there is no reliable way to separate them out.
Google’s own AI features are a separate case. Clicks from AI Overviews and AI Mode in Google Search arrive as ordinary google / organic traffic. At the time of writing, Search Console includes them in the normal Performance totals and does not offer a filter to separate them. Keep that in mind when you report on “AI traffic”: what GA4 can isolate is traffic from standalone assistants, not all AI-influenced visits.
Step 1: Check Which AI Sources Already Send Visits
Start with the standard report before building anything.
- In GA4, open Reports > Acquisition > Traffic acquisition.
- Change the primary dimension to Session source / medium.
- Set the date range to the last 90 days, or longer if your site is small.
- Click Add filter, choose Session source, set the match type to matches regex, and paste the expression below.
.*(chatgpt|openai|perplexity|gemini\.google|bard\.google|copilot\.microsoft|claude\.ai|meta\.ai|you\.com|poe\.com|deepseek|grok\.com).*
The leading and trailing .* mean the pattern also catches subdomains and variations such as chat.openai.com. New assistants appear regularly, so look through your full referral list every few months and add any AI domains you find. If a domain you do not recognise shows up with a handful of sessions, check it before adding it; some referral sources are spam.
Step 2: Create an “AI Assistants” Custom Channel Group
Filtering a report each time is slow. A custom channel group gives AI traffic its own row in every acquisition report, next to Organic Search, Paid Search and Referral. Custom channel groups also apply to your historical data, so you see past AI visits as soon as you save it.
- Go to Admin > Data display > Channel groups.
- Click Create new channel group. GA4 copies the default channels in for you. Name it something like “Channels with AI”.
- Click Add new channel, name it AI Assistants, and add a condition: Source > matches regex > the expression from Step 1.
- Save the channel, then use Reorder to move AI Assistants above Referral. GA4 assigns each session to the first channel whose rules it matches, so if Referral comes first, AI visits will stay there.
- Save the group. In the Traffic acquisition report, open the dimension menu and choose your new channel group instead of the default one.
Your account can have a limited number of custom channel groups, so use one group for all your custom changes rather than creating a new group for each idea.
Step 3: Find the Pages AI Assistants Send People To
Knowing that ChatGPT sent 40 sessions is interesting. Knowing it sent them to three specific guides is useful, because it tells you what kind of content gets cited.
- Open Explore and start a Free form exploration.
- Import the dimensions Session source, your custom channel group, and Landing page + query string.
- Import the metrics Sessions, Engaged sessions, Engagement rate and Key events.
- Put the channel group and landing page in Rows and the metrics in Values. Add a filter for the AI Assistants channel.
Then look for patterns. Are the cited pages long guides, comparison pages, tools or pages with clear definitions? Do they answer a specific question in the first paragraph? Do they include original data or step-by-step instructions? These patterns are your best evidence of what answer engines value on your site. Our guide to AEO and GEO explains how AI search engines choose sources, and it is a good checklist to compare against.
Step 4: Measure Whether AI Visitors Are Worth It
Traffic only matters if it leads somewhere. Mark the actions that matter to your business as key events in GA4 (form submissions, calls, sign-ups, purchases) if you have not already. Then compare the AI Assistants channel with Organic Search on engagement rate, average engagement time and key event rate.
Be careful with small numbers. A channel with 25 sessions and 2 conversions can look spectacular or terrible depending on the month. Look at trends over a quarter, and treat the comparison as a direction, not a verdict. Many sites find AI referral volume is still small compared with Google organic search, so set expectations accordingly when you report it.
Step 5: Look Beyond GA4 for What It Misses
Because much AI influence never shows up as a tracked click, add a few other signals:
- Ask new leads how they found you. Add an optional “How did you hear about us?” field to your contact form, with ChatGPT or another AI assistant as an option. Self-reported answers catch visits that arrived as Direct.
- Watch branded search. People who learn about you in an AI answer often search your brand name afterwards. A rise in branded impressions in Search Console can be a sign of AI exposure.
- Check AI crawler access in your server logs. Crawlers such as GPTBot, OAI-SearchBot and ChatGPT-User (OpenAI), PerplexityBot (Perplexity) and ClaudeBot (Anthropic) identify themselves in their user agent. If your logs or hosting dashboard show them fetching your pages, those systems can at least read your content. Use the AI Crawler robots.txt Tester to confirm your robots.txt is not blocking the ones you want.
- Test prompts by hand. Ask the assistants the questions your customers ask and note whether your site is cited. Record the date and prompt, because answers change.
Common Tracking Mistakes
- Leaving AI traffic inside Referral. If the channel order is wrong, the AI Assistants row stays empty. Always move it above Referral.
- Adding UTM tags to your own internal links to “track” AI visitors. Internal UTM links start a new session and overwrite the real source, which destroys the data you are trying to collect. Use UTMs only on links you place on other sites, emails and ads; the UTM Builder and our UTM naming conventions guide show how.
- Counting Google AI Overviews as measurable AI traffic. They are part of organic search in your reports. Do not claim a number you cannot isolate.
- Not filtering spam. Fake referrals that mention AI brands do exist. If a source sends sessions with zero engagement time from one city, exclude it.
- Using a regex that is too broad. A pattern such as
.*ai.*matches unrelated domains (mail, email, and many more). Keep the list explicit.
Turning the Data Into Action
Once you can see which pages earn AI visits, use that to guide your content work:
- Strengthen the pages that already get cited. Update facts, add a short answer at the top, add an FAQ section with real questions, and make sure the page is fast and easy to read. The Keyword Density and Readability Checker flags long sentences that are hard to quote.
- Create more of what works. If definitions and how-to guides earn citations, plan more of those for related questions.
- Make your structure machine-readable. Clean headings, accurate structured data from the FAQ Schema Generator, and an llms.txt file that points to your best pages are low-cost steps. None of them guarantees citations, but they remove obstacles.
- Make the click worth it. AI visitors often arrive with a specific question already half-answered. Give them a clear next step on the page: a tool, a related guide, or a way to contact you.
Quick Setup Checklist
- Filter Traffic acquisition by Session source with the AI regex to see your baseline.
- Create a custom channel group with an AI Assistants channel above Referral.
- Build an exploration of AI landing pages with engagement and key events.
- Mark your important conversions as key events.
- Add a “How did you hear about us?” field to your forms.
- Check robots.txt and server logs for AI crawler access.
- Review the regex and referral list every few months.
If you would like help setting this up, or want to know why AI assistants cite your competitors and not you, request a free SEO audit or get in touch.
Frequently asked questions
Can GA4 show traffic from ChatGPT?
Yes, for clicks that carry a referrer or ChatGPT’s own utm_source=chatgpt.com tag. Those visits show chatgpt.com as the session source. Clicks from apps that strip the referrer appear as Direct and cannot be separated.
How do I see AI Overviews traffic in GA4?
You cannot isolate it. Clicks from Google AI Overviews and AI Mode arrive as google / organic, and Search Console currently includes them in its normal Performance totals without a separate filter.
What regex should I use to track AI referrals in GA4?
Use an explicit list of AI domains, such as .*(chatgpt|openai|perplexity|gemini\.google|copilot\.microsoft|claude\.ai).*, and add new assistants as they appear in your referral report. Avoid broad patterns like .*ai.* that match unrelated sites.
Why is my AI Assistants channel empty?
The most common reason is channel order. GA4 assigns a session to the first channel whose rules it matches, so the AI Assistants channel must sit above Referral in your custom channel group.
Does AI traffic convert well?
It depends on the site and the query. Compare engagement and key event rates for the AI channel with organic search over several months, and be cautious with small sample sizes.
