How to track ChatGPT traffic in GA4, step by step

This walkthrough uses GA4's standard reports and explorations, with nothing to install. We describe each screen in words rather than with screenshots, and we only name menus and fields that Google documents. Google renames things now and then, so if a label differs slightly, the help pages linked in each step are the reference.

Step 1: Find the AI Assistant channel

On May 13, 2026, Google added an AI Assistant channel to GA4's default channel group. In Google's words, "a new 'ai-assistant' value is automatically assigned when the referrer matches a recognized AI Assistant", and those visits get the campaign name "(ai-assistant)". Google's channel definitions put a visit in the AI Assistant channel when its medium exactly matches ai-assistant. You don't have to set anything up.

1. In the left menu, open Reports, then Acquisition, then Traffic acquisition.
2. The table's first column is the dimension. Make sure it reads Session default channel group.
3. Find the AI Assistant row. Set the date range in the top right to at least 28 days; most small sites see a handful of these visits a month, not dozens.

Google hasn't published the full list of assistants it recognises, and the help page doesn't say whether the channel applies to data from before May 2026. If you want to compare with earlier months, step 4 covers a custom channel group, which does apply to past data.

Step 2: See which assistant sent each visit

The channel row lumps every assistant together. To split it:

1. In the same Traffic acquisition table, open the dimension dropdown in the first column header and choose Session source.
2. Use the search box above the table to type chatgpt, then perplexity, gemini, claude and copilot in turn.

ChatGPT visits can arrive two ways. Most carry a referrer, and ChatGPT also tags links: OpenAI's publishers FAQ says ChatGPT "automatically includes" utm_source=chatgpt.com in referral URLs. Either way the source reads chatgpt.com. Google's help pages don't say how a visit with a utm_source tag but no medium is classified, so it may not always land in the AI Assistant channel. Searching by source, as above, catches it either way.

Step 3: Build a reusable AI traffic exploration

The standard report answers "how many". An exploration answers "which pages, and did they convert", and you only build it once.

1. Open Explore in the left menu and start a Free form exploration. Name it "AI assistant traffic".
2. In the Variables column, add these dimensions: Session source, Session default channel group and Landing page + query string. Add these metrics: Sessions, Engaged sessions and Key events.
3. Drag Session source into Rows and the three metrics into Values.
4. Under Filters, add Session source, choose matches regex, and paste:

chatgpt\.com|openai\.com|perplexity\.ai|gemini\.google\.com|bard\.google\.com|copilot\.microsoft\.com|claude\.ai|chat\.mistral\.ai|chat\.deepseek\.com|meta\.ai|grok\.com

This is our list of the domains these assistants send visitors from, not an official one. Check it against your own Session source report and add anything new you see there.

5. Add Landing page + query string to Rows under Session source. You now see which pages each assistant sends people to. This is the most useful view in the whole exercise: it shows which of your pages assistants trust enough to link.

Comparing Key events per session with your Organic Search row tells you whether AI visitors are worth more or less to you than search visitors. On a small site the samples are small, so read it over a quarter rather than a month.

Step 4 (optional): A custom channel group for past months

The built-in channel only covers what Google recognises, and possibly only data from its launch. A custom channel group fixes both, because Google says "custom channel groups can be applied to your reports retroactively."

1. Open Admin (the gear icon), and under Data display choose Channel groups. You need the Editor role or above on the property.
2. Create a new channel group (a standard property can have two custom groups). It starts as a copy of the default channels.
3. Add a channel named "AI assistants". Set its condition to Source, matches regex, and paste the regex from step 3.
4. Move the new channel above Referral. Order matters: Google's help page says traffic goes into "the first channel whose definition it matches given the current order of channels in the group." Below Referral, ChatGPT visits would be counted as referrals first.
5. Save. In Traffic acquisition, switch the dimension to your new channel group.

Google's help page includes its own example regex for an AI assistants channel. Whichever you use, update it when a new assistant starts sending you visits.

Why your number is lower than reality

GA4 can only label a visit that tells it where it came from. Three things hide AI visits:

• No referrer and no tag. Some apps and in-app browsers open links without a referrer. If the UTM tag is missing too, the visit lands in Direct. Clickport, an analytics company, reported that the share of identified AI visits arriving with a source tag but no referrer rose from 35.7% in April to 49.7% in October 2026 in its own data (Clickport), which shows how much depends on the tag.
• The October 2026 tracking gap. From about October 8 to 15, 2026, after the GPT-6 rollout, links inside ChatGPT answers lost both the UTM tag and the referrer (links in the Sources panel reportedly kept the tag), so many of those clicks showed as Direct. A fix began rolling out on October 15 (Search Engine Roundtable, Search Engine Journal). Add an annotation for that week so a dip doesn't get misread later.
• Google's own AI features. Clicks from AI Overviews and AI Mode come through as ordinary Google organic traffic. GA4 can't separate them. Search Console's generative AI reports, available to all sites since August 31, 2026, show how often your pages appear in those features, but not clicks.

None of this makes the number useless. It makes it a floor. Track the trend, and don't compare it directly with someone else's figure, which will be affected by the same gaps in different amounts.

What to expect, and what to do with it

For most small businesses the number is small. Across the 12 small-business sites in our Observatory, AI assistants sent a median of 0.03% of all visits between August 9 and October 30, 2026, while AI crawlers fetched those sites about 120 times for every visit an assistant sent. A low count in GA4 doesn't mean assistants ignore you; it means most people read the answer and don't click.

So use GA4 for what it's good at:

• Which pages get AI visits. Strengthen those pages first: clear facts, prices, service areas, answers to the questions customers ask.
• Whether AI visitors convert. If they do, that's the business case for spending more time on AI visibility.
• Whether a change moved anything. Compare the quarter before and after a site change, not week to week.

For whether assistants name you, which GA4 can't show, use the free prompt method in our guide to tracking AI visibility, or a tracker from our reviews. And if AI visits are near zero, check first that AI crawlers can read your site with the AI access checker.

Sources

Google Analytics help: default channel group (AI Assistant channel, May 13, 2026), channel definitions and custom channel groups. OpenAI's publishers and developers FAQ for the utm_source=chatgpt.com tag. Coverage of the October 2026 tracking gap from Search Engine Roundtable and Search Engine Journal. All read on November 1, 2026. The regex is ours.