To track AI traffic in GA4, use the built-in AI Assistants channel, which now groups visits from ChatGPT, Gemini, Copilot, Grok, and DeepSeek. It misses Perplexity and Google's AI Overviews, so add a custom channel group with a regex rule to capture the rest and stop AI referrals from hiding in Direct and Referral.
Most teams already have AI traffic arriving. The question is whether their analytics can see it. This guide shows what GA4 catches on its own, where the gaps are, and how to build a channel that reports AI visits and the leads behind them.
AI referrals are now a channel worth measuring
AI-referred traffic has grown into a real acquisition source, and it converts. Adobe's Digital Insights team reports that traffic from AI sources to U.S. retail sites grew 393% year over year in the first quarter of 2026, and that in March 2026 this traffic converted 42% better than non-AI traffic, a reversal from a year earlier when it converted 38% worse (Adobe, April 2026).
A channel that grows triple digits and outperforms your baseline deserves a line in the report. The problem is attribution: without setup, these visits scatter across Direct, Referral, and Organic Search, so the growth stays invisible to the people deciding where budget goes. Getting cited by AI is the front half of this work, covered in our answer engine optimization playbook. Measuring the visits and revenue that follow is the back half, and it starts in GA4. Teams that already treat AI presence as a KPI usually pair traffic data with a share-of-LLM citation metric.
GA4's AI Assistants channel and what it covers
GA4 ships with a native AI Assistants channel, so the most common sources are grouped for you with no setup. Google's own documentation defines it as the channel by which users arrive from sources like ChatGPT, Gemini, Deepseek, Copilot, or Grok, and notes that it excludes Google's AI Overviews and AI Mode (Google Analytics Help).
You will find the channel in the Acquisition reports and anywhere the Session default channel group dimension appears. Under the hood, GA4 assigns a visit to this channel when the referrer matches its list of AI assistants and sets the medium to "ai-assistant." That covers the largest assistants by volume, which is enough to start trending the data this week. It does not cover everything, and the gaps matter.
Where the rest of your AI traffic hides
Three buckets absorb the AI visits the native channel misses, and each needs a different fix. Perplexity is not on Google's AI Assistants list, so its referrals land in Referral alongside blogs and news sites. Google AI Overviews and AI Mode are counted as Organic Search, because the click still originates on a Google results page (Google Analytics Help). And a large share of assistant traffic arrives with no referrer at all, which GA4 files under Direct.
The Direct bucket is the stubborn one. GA4 treats a session as Direct when there is no source and no medium to read, which is exactly what happens when an AI app opens a link without passing a referrer (Google Analytics Help). You cannot fully eliminate this, but you can pull the identifiable AI sources out of Referral and into a channel of their own. The same scattering is why measuring AI's real influence needs more than one report, a point we make in GEO vs SEO.
Build a custom channel group to catch every AI source
A custom channel group lets you define an AI rule that captures Perplexity and any assistant Google has not added yet. Default channel groups cannot be edited, so Google's supported path is to create a custom group that runs alongside the default (Google Analytics Help). Build it once and every future report can filter on it.
- In GA4, open Admin → Data display → Channel groups and click Create new channel group.
- Name it something clear, such as "AI sources," and add a channel named "AI assistants."
- Add a condition on Source set to matches regex, and enter a pattern covering the assistants you care about, for example
chatgpt|openai|perplexity|gemini|copilot|claude|grok|deepseek. - Drag the AI rule above the Referral rule. Order decides the match, so if Referral sits first, Perplexity is scored as Referral before your rule ever runs.
- Save the group. It runs alongside the default grouping and can be applied to your existing reports.
Rule order is the single most common reason a freshly built AI group still shows traffic under Referral, so check it before you assume the regex is wrong.
Turn AI sessions into a lead metric you can report
Traffic counts are the start; the number a stakeholder cares about is leads. Because the AI Assistants channel and your custom group are standard dimensions, you can cross them with your key events (form submits, demo requests, calls) in any GA4 exploration or report (Google Analytics Help). Build a free-form exploration with your AI channel as the row and your conversion events as the values, and you have an AI-sourced pipeline view.
Segment it the way you would any channel: sessions, engagement rate, key events, and revenue if you track it. That turns "AI is a thing now" into a defensible figure you can put next to paid and organic. Reporting AI-sourced pipeline this way is exactly the kind of measurement layer our Next Best Action program is built around, and the tracking setup itself is standard work for our digital marketing and technology team.
What GA4 will not tell you, and where to cross-check
GA4 has real limits with AI traffic, so treat it as one input, not the whole picture. A custom channel group can reclassify data GA4 already collected, yet it cannot recover referrals that never carried a referrer; those were logged as Direct and stay there. GA4 also only holds data from the point its tag went live, and it measures clicks to your site, not the far larger number of times an AI answer names your brand without sending a click.
Close those gaps with other sources. Google Search Console shows the queries and pages that AI Overviews and AI Mode pull from, since that traffic is Organic Search. Server logs reveal which AI crawlers actually fetch your pages. And citation tracking measures how often models name you at all, the metric behind our AEO work. If wiring this together across tools is more than your team wants to own, we can set it up.
