Learn how to track AI traffic in GA4 using the new AI Assistant channel. Find out what it captures, what it misses, and the 5-step setup that gives you
Track ai traffic ga4 matters more than most guides admit. If you opened your GA4 Traffic Acquisition report sometime after June 2026 and spotted a channel called "AI Assistants" that nobody on your team set up — that's not a glitch. Google quietly shipped a dedicated channel to track AI traffic in GA4 on May 13, 2026, and it's the most important new row in your acquisition report since Organic Social got its own lane years ago.
This piece walks through exactly what the channel captures, where it falls short (and it does fall short), and the five steps to get your reporting properly configured before the data you need isn't there when you want it.
Tracking AI visibility has been a patchwork exercise for most of 2024 and 2025 — custom regex, messy referral filters, educated guesses. The new channel changes that baseline.
When GA4 detects a visit whose referrer matches a recognized AI assistant, it automatically does three things:
ai-assistant(ai-assistant)No configuration required. No regex to write. No custom channel group to build first. If you have a GA4 property, you already have this channel. Broad availability across properties landed around June 7, 2026, so if you haven't checked recently, open your Traffic Acquisition report now.
At launch, Google named ChatGPT, Gemini, and Claude as recognized sources. By June, the live documentation had expanded that list to include DeepSeek, Copilot, and Grok — so the recognized set is a moving target. Check Google's Help Center definition page before locking in any reporting that names specific platforms.
Getting eyes on the data takes about 30 seconds:
ai-assistant to isolate and verify those sessions.If the channel shows zero sessions, your property may still be rolling out, or you genuinely haven't received referrer-tagged AI traffic yet. Either way, the channel is there, and it starts counting the moment a qualifying visit arrives.
The native channel is a genuine step forward, but it's a floor, not a ceiling. There are four gaps you need to understand before building reporting or telling a client what AI is sending them.
GA4 doesn't reclassify historical data. Any AI traffic that arrived before the rollout date stays buried in Referral or Direct — and comparing this channel month-over-month across that cutoff will look distorted. Strict year-over-year comparisons for this channel are effectively impossible until May 2027.
Perplexity passes a perplexity.ai referrer header, which GA4 files under Referral, not the AI Assistant channel. It's still measurable — you can find it in your Referral report — but it won't appear in your AI Assistants count automatically. Given that Perplexity is widely considered one of the higher-intent AI traffic sources, this is a gap worth closing manually with a custom channel group rule.
Clicks from AI Overviews and Google's AI Mode carry a google.com referrer identical to a normal organic result. GA4 has no way to distinguish them from standard clicks, so they stay inside Organic Search. For that data, the Search Console Generative AI Performance report (launched June 3, 2026) is the right tool — it shows impressions from AI Overviews and AI Mode, broken down by page, country, and device.
AI traffic only enters the channel when GA4 can read a referrer header. In-app browsers, native mobile apps, and copy-pasted links all strip that header. One server-log study found that 56 Gemini-on-iOS visits produced just 5 GA4 referrals — roughly 9% of the actual total. Industry estimates put the under-count from referrer-less AI sessions at somewhere between 35–70% of real volume, depending on how much of your audience uses AI tools in mobile contexts.
Treat every number in the AI Assistants channel as a lower bound. It tells you at least this much traffic came from AI. The real number is higher.
Open Traffic Acquisition and confirm the AI Assistants row is visible. Then create a GA4 annotation — May 13, 2026: AI Assistant channel rollout — so anyone reviewing historical trend lines has context for why the channel appears to start from zero on that date. Annotations take two minutes and prevent confusion for every future report.
Go to Admin → Data display → Channel groups → Add new channel group. Create a rule that catches perplexity.ai as a source, plus any other AI platforms you want to track together (Poe, You.com, and others that aren't in Google's recognized list). GA4 allows a maximum of two custom channel groups per property, so plan accordingly. The custom group can also be applied retroactively, recovering some of the pre-May traffic that's currently sitting in Referral.
Standard reports give you a high-level view. For real analysis, go to Explore → Blank and build a free-form report using:
This lets you see which pages AI assistants are sending people to, and whether those people are converting. Landing page data is where the real intelligence lives — if ChatGPT is sending 200 sessions a month to one product page, that page is worth prioritizing in your content strategy.
Session counts are fine for a status update. Revenue and conversion data are what justify content investment. Make sure your key events — purchases, add-to-carts, email signups, whatever matters to your store — are firing correctly on pages receiving AI traffic. Filter your conversion reports by the ai-assistant medium to see the channel's contribution in dollar terms. Some data suggests AI-referred visitors convert at a significantly higher rate than organic search visitors, so even small session counts can carry real value.
Until Perplexity lands in the AI Assistant channel natively, check your Referral report monthly and filter for perplexity.ai as the source. Add it to your custom channel group so it's grouped with your other AI traffic in custom reports. Also scan for other AI tools you didn't expect — the assistant landscape is changing fast, and new referrers appear without announcement.
The first 30 days of data in this channel sets your baseline, and it's worth documenting before you make any decisions from it. Here's what to record:
Don't benchmark AI traffic against organic search volume yet — that comparison will be misleading given the scale difference. Instead, watch the quality of AI sessions. High engagement rate and conversion rate with low session count is the signal that AI mentions of your store are converting well even when traffic is still modest.
This is also the right time to audit which pages are receiving AI traffic and ask whether they're set up to convert. A page getting cited by ChatGPT that has no clear call to action is leaving money on the table. For help building pages that AI tools are more likely to cite and recommend, check out how to get cited by AI search engines.
Once you have two full months of clean data, the number to track is AI Assistants sessions as a percentage of total non-paid traffic. For most ecommerce stores right now, this is still a small figure — under 3% for most sites. That percentage is expected to grow consistently through late 2026 and into 2027 as AI tools send more traffic, particularly for product discovery and comparison queries.
If your AI traffic percentage is flat or declining while competitors are growing, that's a content and citation signal, not a technical one. It means AI tools aren't recommending your pages. The answer is usually structural: your content doesn't answer questions directly enough to be quoted, or your product pages lack the kind of factual, structured detail that AI systems pull from. A related read: the ten-minute edit that turns an AI draft into a citable source.
No — at least not yet. The native channel fills in recognized sources automatically, but a custom regex group still earns its place for three reasons:
Run both in parallel. Let the native channel handle your top-level reporting, and keep the custom group for digging into the full picture. For a broader look at what AI tracking tools and methods exist beyond GA4, see our guide on 6 ways to track AI visibility in 2026.
It's worth saying plainly: this channel measures clicks, not influence. If ChatGPT recommends your product and the user buys it after copying the URL and navigating directly, GA4 won't know. If Google AI Mode cites your blog post and 500 people read that answer without clicking, GA4 won't know. The click-through is a useful signal, but AI's actual effect on your business is much larger than what any referrer-based tool can measure.
That's not a reason to ignore the channel — it's a reason to set expectations correctly when you report on it. The AI Assistants channel gives you a partial, forward-only signal. It's better than nothing, and meaningfully better than where we were six months ago. But it's not complete, and building strategy purely on what it shows you will lead to under-investment in content that AI already loves and you can't currently see.
Turn the channel on, annotate the date, build the Exploration, and start accumulating data. The stores that have a clean three-month baseline in October will be the ones who know what to do when the holiday season shifts in 2026 — and they'll have real numbers to back it up.
No. Perplexity passes a perplexity.ai referrer header that GA4 classifies as Referral, not AI Assistant. To group Perplexity with your other AI traffic, create a custom channel group in Admin → Data display → Channel groups with a source rule for perplexity.ai.
No. The channel only counts sessions forward from May 13, 2026. Any AI traffic that arrived before that date stays in Referral or Direct and will not be reclassified. This makes year-over-year comparisons for this channel impossible until May 2027.
No. Clicks from AI Overviews and AI Mode carry a standard google.com referrer and are classified as Organic Search. For AI Overview impression data, use the Search Console Generative AI Performance report, which launched in June 2026.
Either your property hasn't received referrer-tagged AI traffic yet, or the rollout hadn't reached your property by the time you checked. Broad availability landed around June 7, 2026. If you still see nothing, confirm the channel exists in your Default Channel Group list and check whether AI assistants are actually linking to your site.
No — keep both running. The native channel doesn't include Perplexity and has a growing but unpublished list of recognized sources. A custom regex group applies retroactively, captures sources Google omits, and gives you a stable definition you control. Use the native channel for top-line reporting and the custom group for deep analysis.
Industry estimates put the under-count at 35–70% of real AI-referred volume, depending on how your audience accesses AI tools. Visits from in-app browsers, native mobile apps, and copy-pasted links all strip the referrer header and land in Direct. One server-log study found only about 9% of Gemini-on-iOS visits showed up as GA4 referrals.
The playbook
The playbook first, then a few short emails to help you publish your first great article, then just the occasional update. No spam, and you can unsubscribe anytime.