How to Track AI Traffic in GA4 (2026)

People are starting to find your website by asking ChatGPT, Gemini, and Perplexity instead of typing a search into Google. The question every business owner is now asking is a fair one: how much of my traffic is that, and is it worth anything?

For a long time the honest answer was that you could not tell. Visits from AI tools got mixed in with everything else in Google Analytics, filed under "Referral" next to links from blogs and forums, or dumped into "Direct" with no source at all. You could see your total traffic. You could not see how much of it came from an AI answer.

That has partly changed. In May 2026, Google Analytics 4 (GA4) added a built-in way to see traffic from the major AI chatbots, with nothing to set up. It is a real improvement, and it has four blind spots that will quietly undercount your AI traffic if you do not know about them.

Here is what the new channel shows, where to find it, the four things it misses, and how to fill each gap.

What GA4 Changed in May 2026

GA4 sorts every visit into a "channel", a plain-English label for how someone got to your site: Organic Search, Direct, Referral, Email, Paid Search, and so on. That list is called the Default Channel Group, and until recently it had no bucket for AI tools.

On 13 May 2026, Google added one. Its release note reads:

Google Analytics now provides a dedicated way to measure and analyze traffic originating from popular AI assistants. You can now identify how users are discovering your site through chatbots like ChatGPT, Gemini, and Claude via a new AI Assistant channel in your Default Channel Group reports.

The new channel is called AI Assistant. You do not have to switch it on, build anything, or edit a setting. It applies to every GA4 property automatically, and it started showing data from the day it launched.

Underneath, GA4 does this: when someone lands on your site and the previous page was one of the AI tools Google recognises, GA4 labels that visit with a traffic source "medium" of ai-assistant. Every visit carrying that label rolls up into the AI Assistant channel in your reports. "Medium" is just GA4's word for the type of source. You will see ai-assistant appear as a value wherever GA4 shows you source and medium.

Where to Find It

Three places, depending on what you want.

The quick check. Open Reports, then Acquisition, then Traffic acquisition. The table is grouped by "Session primary channel group". Look for the AI Assistant row. That is your AI chatbot traffic for the date range, with sessions, engaged sessions, and any key events or revenue you have set up, sitting right next to Organic Search and Direct so you can compare.

The detail. In that same report, change the grouping dropdown from channel to "Session source / medium". Now you can see the split inside the AI Assistant channel: chatgpt.com / ai-assistant, gemini.google.com / ai-assistant, and so on. This tells you which AI tool is actually sending people.

The custom report. If you want landing pages, countries, devices, or conversions broken out by AI source, build a free-form exploration in the Explore section. More on that below.

One caveat. You can set the date range as far back as you like, but GA4 only started recognising AI sources when the channel launched in mid-May 2026, so anything earlier reads as zero even where older AI visits existed. Treat the channel as a "from now on" measure, not a historical one.

What the Channel Counts

Google's channel definition is short:

AI Assistant is the channel by which users arrive at your site from sources like ChatGPT, Gemini, Deepseek, Copilot, or Grok. It excludes Google's AI Overviews and AI Mode.

So the recognised list, as Google currently describes it, is five tools: ChatGPT, Gemini, DeepSeek, Copilot, and Grok. Google maintains this list and has already changed it once. The launch note named Claude as an example; the current definition does not. Because the list can move, the safe approach is to treat Google's definition page as the live source and check it if a tool you care about seems to be missing.

Two things follow from that definition, and both matter enough to have their own sections.

Blind Spot One: Perplexity Is Not on the List

Perplexity is one of the most useful AI referral sources for a lot of businesses, especially in B2B, because people use it specifically to research and compare options before buying. It is not in Google's recognised list, so GA4 does not give its visits the ai-assistant label. They land in Referral instead, mixed in with every other site that links to you. Earning that traffic in the first place is a separate task, covered in getting cited by Perplexity; this section is only about seeing it once it arrives.

You can fix this yourself with a custom channel group, which is a copy of the Default Channel Group that you are allowed to edit.

1. In Admin, under Data display, open Channel groups. Click Create new channel group and give it a name like "Default + AI".
2. Add a new channel above the built-in Referral rule. Name it something like "AI Assistant (extended)".
3. Set its condition to: Source matches regex .*perplexity\.ai.*. A regex is just a pattern; this one matches any source containing perplexity.ai, including www.perplexity.ai.
4. If you want a single AI bucket that also catches the five Google already recognises, extend the pattern: .*(perplexity\.ai|chatgpt\.com|openai\.com|gemini\.google\.com|copilot\.microsoft\.com|grok\.com|deepseek\.com).*
5. Order matters. GA4 checks channel rules top to bottom and stops at the first match, so this rule has to sit above the Referral rule or Perplexity keeps getting filed as Referral.
6. Save. Reports that let you pick a channel group will now offer yours, and it applies to past data too.

Keep the custom group as a supplement, not a replacement. The built-in AI Assistant channel keeps working, and Google updates its recognition list without you having to touch anything.

Blind Spot Two: Google's Own AI Is Filed as Organic Search

When someone clicks a link inside a Google AI Overview or in AI Mode, GA4 does not put that visit in the AI Assistant channel. It puts it in Organic Search. Google's own definition says so directly:

Organic Search is the channel by which users arrive at your site/app via non-ad links in organic-search results, including Google's AI Overviews and AI Mode.

This is not a bug. The web address the visit comes from is google.com/search, the same address as an ordinary blue-link click, because an AI Overview is still a search results page, just laid out differently. GA4 has no way to tell the two apart from the visit alone.

The practical result: the AI Assistant channel only covers the standalone AI apps, not AI inside Google Search. If most of your audience is finding you through AI Overviews rather than through ChatGPT, this channel will look almost empty while your actual AI exposure is high.

To see the Google AI side, you need a different tool. Google Search Console added a generative AI report in June 2026 that shows how often your links appear inside AI Overviews and AI Mode. It has its own limits, chiefly that it shows appearances and not clicks, but it is the only view of that surface you are going to get. Run the two together: Search Console for how often Google's AI shows your link, GA4 for the traffic the standalone apps send.

Blind Spot Three: App Visits Often Land in Direct

Plenty of ChatGPT use happens in the desktop app and the phone app, not in a browser tab. When someone taps a link in the app, the visit often reaches your site with no source information attached at all. There is nothing for GA4 to match against, so it files the visit as Direct, the bucket for people who typed your address or used a bookmark.

This is a known limitation of how apps and browsers pass along source data, and there is no setting that fixes it. What you can do is watch for the symptom. If your Direct traffic has climbed steadily over the last year with no matching change in brand awareness or email volume, some of that is very likely unattributed AI traffic. It will not show in the AI Assistant channel, and no amount of channel-group editing will pull it back out, because the information genuinely is not there.

The one place you get full control is links you post yourself, covered further down.

Blind Spot Four: It Tells You They Arrived, Not What They Asked

The AI Assistant channel shows sessions, engagement, and conversions. It does not show the question someone asked the AI before clicking through to you. There is no prompt data, no equivalent of a keyword report.

That is the same gap the Search Console AI report has, and it matters for the same reason: knowing which questions send you traffic is how you would decide what to write next. Neither tool gives you that. The closest you can get is to look at your landing pages within the AI Assistant channel. If AI tools keep sending people to one specific guide or comparison page, that page is answering a question the AI thinks you answer well, and that is a signal worth acting on even without the exact wording.

Build a Report You Will Actually Use

The default table is fine for a glance. For anything real, build an exploration once and reuse it.

In Explore, start a blank exploration and set it up like this:

• Dimensions: Session source / medium, Landing page + query string, Session primary channel group, Country, Device category.
• Metrics: Sessions, Engaged sessions, Average engagement time, Key events, Total revenue.
• Filter: Session primary channel group exactly matches AI Assistant. If you built the custom group above, filter on that instead so Perplexity is included.
• Rows: Landing page, so you see which pages AI tools favour.

Save it. Now you have a single screen that answers the questions that matter: how much AI traffic, from which tools, to which pages, from where, and whether it converts. Check it monthly. AI referral patterns are still shifting fast, and a page that gets no AI traffic today can start getting a lot within a quarter.

Pay most attention to the engagement and conversion columns. AI traffic often arrives already informed, because the person has read a summary of you before clicking, so it can convert differently from cold search traffic. Whether that difference is good or bad for your site is the actual question, and it is one only your own numbers can answer.

UTM Tags for Links You Control

Some of your AI-adjacent traffic comes from links you place yourself: a citation in a piece of research, a link in a community answer, a reference doc you publish for AI tools to read. For those, you do not have to rely on GA4 guessing the source.

Add UTM tags to the address. UTM tags are labels you bolt onto the end of a link that tell GA4 exactly where it came from. A link like:

https://www.example.com/guide?utm_source=perplexity&utm_medium=ai-assistant&utm_campaign=research-citation

arrives in GA4 already labelled, and it does not matter whether the source information survives the trip. This is the only way to get a reliable source out of app clicks and other cases where that information goes missing, so use it anywhere you are placing a link on purpose.

Common Mistakes

Treating the AI Assistant channel as your whole AI picture. It covers five standalone apps. It does not cover Google's AI, and it undercounts app traffic that arrives as Direct. Read it as a floor, not a total.

Comparing the channel number to an AI visibility tool and expecting a match. A visibility tool runs its own set of prompts and counts how often you are mentioned. GA4 counts humans who actually clicked. These measure different things and will never line up. If you want to understand what the paid visibility tools actually do, that is a separate question from what GA4 shows.

Adding a Perplexity rule below the Referral rule. GA4 stops at the first matching rule. Anything you add to catch AI sources has to sit above Referral or it never fires.

Panicking about a Direct traffic increase, or ignoring it. Neither. A steady Direct climb with no other explanation is a hint that unattributed AI traffic is growing. It is a reason to set up UTM tags on your own links and to lean on Search Console for the Google AI side, not a number to try to reverse-engineer.

Optimising for AI traffic before checking whether it converts. More AI referrals is not automatically a win. Build the report, watch the conversion column for a couple of months, then decide how much effort the channel deserves.

How to Check Your Setup in Five Minutes

1. Open Reports, then Acquisition, then Traffic acquisition. Set the date range to the last 90 days. Confirm you can see an AI Assistant row. If it is missing entirely, you may simply have no recognised AI traffic yet, which is common for smaller or newer sites.
2. Switch the grouping to Session source / medium and look for values ending in / ai-assistant. Note which AI tools appear.
3. Check your Referral channel for perplexity.ai. If it is there and you care about Perplexity, build the custom channel group.
4. Look at your Direct channel trend over the last 12 months. A steady unexplained rise is your cue to add UTM tags to any links you place yourself.
5. Open Search Console and check whether the generative AI report is available on your property. That covers the Google AI traffic GA4 files as Organic Search.
6. Build and save the Explore report above so you are not rebuilding it every month.

None of this earns you more AI traffic. It just lets you see what you are already getting, which is the first step before deciding whether it is worth chasing. What actually moves the number is whether an AI system can reach your pages, make sense of them, and trust the answer enough to pass a visitor to you. That gap between being technically ready and actually getting mentioned is worth understanding on its own, and readiness versus visibility walks through it.

Seeing the traffic is the easy part. Whether AI systems can actually read your pages, identify your business, and trust it enough to recommend you is what decides how big that channel gets. Run a free AI readiness check to see what they see.