AI
Join 500+ brands growing with Passionfruit!
Referral traffic from AI search engines has grown substantially, yet most marketers have no visibility into how much traffic comes from ChatGPT, Perplexity, Claude, and Gemini.
GA4 categorizes generative AI traffic as referral, not organic search. Without proper configuration, AI traffic gets lumped into generic referral buckets where you can't analyze it properly.
Method 1: Quick Check via Traffic Acquisition Report
The fastest way to see AI traffic:
Go to Reports > Acquisition > Traffic acquisition
Change primary dimension to Session source/medium
Use the search box and type: chatgpt, copilot, perplexity, gemini, claude
You'll get immediate visibility into AI referrals, but you'll need to check manually each time.
Method 2: Create Custom Segments with Regex
For ongoing analysis, build a regex-based segment:
Open GA4 and click Explore
Start a new blank exploration
Add dimension: Session Source/Medium
Add metrics: Sessions, Engaged Sessions, Key Events
Create new custom segment named "AI Sources"
Add condition: Session Source matches regex
Use this regex pattern to capture AI traffic:
Save the segment for quick access in future explorations.
Method 3: Create a Custom Channel Group (Recommended)
Custom channel groups integrate AI as a distinct category across all acquisition reports, treating it with the same importance as Organic Search or Paid Social.
Steps to create:
Go to Admin > Data display > Channel groups
Click "Create new channel group"
Add a new channel called "AI Search"
Define the channel using source matching for AI platforms
Reorder channels to place AI Search above Referral
Important: GA4 sorts traffic into channels from top to bottom. Drag AI Search above the Referral channel so AI traffic doesn't get miscategorized.
Custom channel groups work retroactively, applying to historical data as well.
Setting Up Custom Dimensions
For deeper analysis, create custom dimensions to track:
AI platform (ChatGPT, Perplexity, etc.)
Content type receiving AI traffic
User behavior patterns from AI referrals
Register these as event-scoped dimensions in GA4 Admin > Custom definitions.
Custom dimensions let you answer questions like "which content types perform best with AI traffic" and "how does AI user behavior differ from organic."
UTM Taxonomy for AI Campaigns
If you're creating content specifically for AI citation, use consistent UTM parameters:
Apply these to any links you can control, such as links in your own content that AI might cite.
Google Analytics 4 now ships with a native AI Assistant default channel group that automatically separates traffic from recognized AI chatbots like ChatGPT, Gemini, and Claude. The update went live in the Analytics Help Center on May 13, 2026, and is rolling out to all properties over a few weeks. Property owners no longer need a custom regex channel group to see AI assistant traffic broken out from referrals, though the native channel only captures sessions that arrive with a clean referrer header. Roughly 60 to 80% of actual AI-driven visits carry that header, which means the new channel measures the floor of your AI traffic, not the ceiling.
The piece below covers what the native channel does, what it does not yet solve, when to keep your custom regex backup, how to read the numbers without overreading them, and where GA4 fits in the broader AI search measurement stack. Most marketers will need 15 minutes to verify the channel is live in their property and another hour to set up the supporting reporting that turns the data into decisions.
What the new AI Assistant channel actually does
Google Analytics 4's May 13, 2026 update changes three traffic source dimensions automatically whenever GA4 detects a referrer matching a recognized AI assistant. The medium dimension is assigned the value "ai-assistant," the channel group dimension is grouped under a new "AI Assistant" channel inside Default Channel Group reports, and the campaign dimension receives the reserved label "(ai-assistant)." All three changes happen without any configuration on the property owner's side.
Google has named ChatGPT, Gemini, and Claude as examples of recognized referrers in its Help Center documentation. The company has not published the full list, and the August 2025 custom channel group guidance named five platforms (ChatGPT, Gemini, Microsoft Copilot, Claude, and Perplexity), so additional surfaces are likely covered but unconfirmed. The Default Channel Group definitions page has not yet been updated with the AI Assistant entry as of mid-May 2026, so the full technical definition is not publicly documented.
The rollout follows the 2022 pattern when Google added the "cross-network" default channel group for Performance Max and Smart Shopping traffic, moving sessions from a generic bucket into a dedicated channel without manual setup. Properties may see the AI Assistant channel appear anywhere from a few days to several weeks after the May 13 release. The change is not retroactive: sessions that arrived before May 13 stay in whatever channel they were originally classified under.
How to confirm the AI Assistant channel is live in your property
Three steps to verify the channel is active in your GA4 property. Each takes about a minute.
First, open GA4 and go to Reports, then Acquisition, then Traffic Acquisition. Switch the primary dimension to Session Default Channel Group. If "AI Assistant" appears in the table, the rollout has reached your property.
Second, if you do not see the channel, switch the dimension to Session Source / Medium and look for entries with medium set to "ai-assistant." The dimension-level change rolls out alongside the channel grouping, so seeing one without the other usually means a partial rollout.
Third, if neither appears, run a manual test. Ask ChatGPT, Gemini, or Claude a query that prompts a citation to your site, click the citation, and check the Realtime report in GA4 within 30 minutes. If the session shows up under the AI Assistant channel, the rollout has hit your property. If it shows up under Referral or Direct, the rollout is still pending, and your custom channel group (if you have one) is doing the work.
What the native channel does not solve
Three structural gaps remain even after the May 13 update. Treating GA4's AI Assistant channel as a complete measurement solution overstates what the data actually captures.
The first gap is the referrer-stripping problem. Sessions that arrive without a referrer header still land in the Direct channel. The pattern shows up in three common contexts: AI mobile apps and certain in-app browsers, copy-paste link sharing, and embedded browser surfaces that strip referrer data by design. Industry analysis suggests roughly 60 to 80% of AI-originated visits carry clean referrer headers, which means 20 to 40% of your actual AI traffic continues to appear as Direct.
The second gap is the citation-without-click problem. When an AI assistant recommends your brand by name and the user then searches for you directly on Google five minutes later, the resulting visit lands in GA4 as Organic Search. The AI's role in the journey is invisible to the attribution system. The only secondary signal is a rise in branded search volume that does not correlate with other organic activity. AI assistant traffic measures clicks from cited links, not the broader influence of being cited at all.
The third gap is platform coverage. Google has confirmed ChatGPT, Gemini, and Claude. Perplexity, Microsoft Copilot, You.com, Meta AI, DeepSeek, Grok, and smaller AI surfaces may or may not be on the recognized referrer list. Until Google publishes the full list, your custom regex backup remains the only reliable way to catch traffic from platforms outside the named three.
Should you keep your custom regex channel group?
If you already built a custom channel group with regex patterns to track AI traffic, the May 13 update raises a clean question: keep it, retire it, or run both? The answer depends on three factors.
Keep the custom regex if you are tracking platforms beyond the three Google has named. Perplexity, Copilot, You.com, Meta AI, and smaller surfaces are not confirmed in the native channel, and the custom regex remains the only certain way to capture them in a single grouping. Most agencies and B2B SaaS marketing teams should keep the custom group active for this reason alone.
Retire the custom regex only if your tracking only ever covered ChatGPT, Gemini, and Claude, and your team prefers cleaner dashboards over coverage breadth. Even then, document the retirement decision with an annotation in GA4 so the historical step-change in your reports is explainable later.
Run both if you have dashboards or Looker Studio reports that already pivot on the custom channel group. Renaming Google's native group to match your custom group's name is also an option, though that creates two competing numbers in dashboards if anyone forgets which is which. A cleaner approach is to keep the custom group active and rename it to something like "AI Assistant (Custom Extended)" so the distinction is obvious.
Whichever option you choose, add a property-level annotation on the date the AI Assistant channel appears in your reports. Three months from now, an analyst comparing year-over-year acquisition data will see a step-change in Organic Search or Referral as AI traffic moves into the new channel. Without the annotation, that change reads as a campaign result instead of a categorization shift.
Setting up custom dimensions and segments to enrich AI traffic data
The native channel separates AI traffic from referrals, but it does not by itself answer the questions marketers actually ask. Three add-on configurations turn the raw channel data into actionable reporting.
The first is page-level analysis. Build an Explore report in GA4 that filters to the AI Assistant channel and adds Landing Page as a secondary dimension. The output shows which pages on your site are pulling AI referrals, which tells you which content is being cited (and therefore which content to invest in further).
The second is source-level breakdown within the AI Assistant channel. Use Session Source as a secondary dimension inside the AI Assistant channel filter to see how much traffic each AI platform sends individually. One assistant often converts at a meaningfully higher rate than the others, which tells you where to deepen content investment. Early B2B benchmarks show Perplexity converting at a higher rate than ChatGPT in research-heavy categories, though the pattern varies by industry.
The third is event tracking for high-intent AI behaviors. If your site has key events configured (signups, demo requests, purchases, phone clicks, form fills), conversions flow through the AI Assistant channel automatically. Build a conversions-by-channel comparison view that places AI Assistant alongside Organic Search and Paid Search. Cited brands inside AI Overviews earn approximately 120% more organic clicks per impression than uncited brands, per Seer Interactive's March 2026 update, and the same citation premium tends to show up in downstream conversion data once attribution flows through.
For UTM-based campaign-level tracking on links you control, apply consistent UTM parameters (utm_source=ai_assistant, utm_medium=ai_assistant, utm_campaign=[campaign_name]) to any AI-targeted links you place yourself. The native channel does not affect UTM parameter handling, so anything you tag manually still flows through as expected.
How to read AI traffic without overreading it
The most common mistake with the new AI Assistant channel is treating its numbers as a real-time KPI. Three framing shifts produce better decisions.
AI Assistant traffic is a lagging indicator, not a leading one. Sessions appearing in the channel today reflect content that was cited by an AI assistant in the past, which in turn reflects content investment from weeks or months earlier. Optimizing for "AI Assistant sessions this week" is optimizing on data with a 30 to 90 day lag.
Citation share is the leading indicator. The work that drives future AI Assistant traffic is whether your content is being cited inside the AI's answer in the first place, before any click happens. Tools that track citation share across ChatGPT, Perplexity, Gemini, and Claude (Passionfruit Labs, Profound, Otterly, and similar platforms) measure the upstream signal that determines what shows up in GA4 later. Without citation tracking running alongside GA4, you are measuring outcomes without measuring the inputs that produce them.
Conversion quality matters more than volume. AI-driven visits remain a small share of total traffic for most sites, often well under 5%. February 2026 research indicated ChatGPT sends roughly 190 times less website traffic than Google, despite handling around 12% of Google's query volume. The volume gap is real, but conversion quality is consistently higher: visitors arriving from AI assistants typically convert at 1.5x to 5x the rate of organic search visitors, depending on industry. The Washington Post reported AI traffic converted at 4 to 5 times the rate of traditional search visitors per Karl Wells, the publication's chief revenue officer, in reporting by Digiday. The right way to use the AI Assistant channel is not to maximize traffic volume but to verify that high-quality citation work is producing high-quality landed traffic.
Where GA4 fits in the broader AI search measurement stack
GA4's AI Assistant channel is one layer of a multi-layer measurement stack. Treating it as the complete picture misses three other measurement surfaces marketers should run in parallel.
Google Search Console measures impressions and clicks for content surfaced in Google AI Overviews and AI Mode, which is a Google-specific measurement that GA4 does not break out separately. AI Overview traffic appears in GA4 as google / organic, indistinguishable from classic search traffic. Search Console is where the AI Overview side of the story lives, and the data should be reviewed alongside GA4 weekly.
Microsoft Clarity introduced dedicated AI channel groups in December 2025 and tracks behavioral signals (session recordings, heatmaps, scroll depth) for AI-driven visitors that GA4 does not capture. For teams investing seriously in AEO and GEO, running Clarity alongside GA4 reveals how AI-referred visitors behave on the site, not just how many of them arrive.
Citation tracking platforms close the loop on the upstream signal. Passionfruit Labs, Profound, Otterly, and similar tools measure how often your brand is mentioned or cited in AI answers across the major surfaces, which is the variable that determines what GA4 reports later. Our comparison of AEO and GEO tracking tools for B2B SaaS and the equivalent breakdown for ecommerce walk through the platforms in detail.
A working 2026 AI search measurement stack typically includes: GA4 for landed traffic, Search Console for AI Overview impressions, Microsoft Clarity for behavioral signals, and a dedicated citation platform for the upstream visibility data. No single layer tells the full story, and most teams that ask "is our GEO investment paying off" are usually missing two or three layers when they ask the question.
Make AI traffic part of the conversation, not just a row in a report
The May 13, 2026 update is a real improvement, and any marketing team running classic GA4 reports will see cleaner AI attribution within the next few weeks. The harder work is upstream of the GA4 channel: getting cited in the first place. AI Assistant traffic is what shows up after the citation work is done, and a low number in the channel almost always traces back to a content and entity authority gap that no analytics configuration can fix.
If your AI Assistant channel is showing minimal traffic and your branded search and direct traffic look flat, the actual question is whether your brand is being cited inside the AI answers your buyers see, not whether GA4 is tracking the clicks correctly. The cleanest first step is a current-state audit that maps how your content is performing across ChatGPT, Perplexity, Gemini, AI Overviews, and Claude. Look at how Passionfruit's GEO service approaches the audit on top of a solid SEO foundation, see the citation tracking inside Passionfruit Labs, and talk to the team before the next budget cycle locks in.
Common Tracking Issues
Missing traffic
Some AI tools use embedded browsers that appear as direct traffic. You won't capture 100% of AI referrals no matter how well you configure tracking.
Misattribution
Without custom channel groups, AI traffic gets lumped into referrals alongside unrelated sources. Separation requires manual configuration.
Query blindness
Unlike traditional search, GA4 doesn't provide search query data from AI tools. Track landing pages instead to understand which content attracts AI-driven traffic.
Sample Dashboard Elements
Build a dashboard showing:
AI traffic trend (line chart, weekly)
AI traffic by platform (pie chart)
Top landing pages from AI (table)
AI vs organic traffic comparison (bar chart)
Conversion rate: AI traffic vs other channels
Compare AI traffic conversion rates against other channels to understand quality. Early data suggests AI-referred visitors often have higher engagement because they've already researched options.
What the Data Tells You
Once tracking is configured, use the data to inform your revenue-focused SEO strategy:
Content optimization priorities
Pages receiving AI traffic deserve optimization attention. Add FAQ schema, improve extractability, and strengthen entity signals on content AI already cites.
Platform prioritization
If ChatGPT sends 10x more traffic than Perplexity, prioritize optimization for ChatGPT's citation patterns. Platform preferences vary by industry.
Conversion analysis
AI traffic that converts at higher rates indicates high-intent queries. Double down on content capturing those queries.
Connecting GA4 to Revenue
Traffic metrics alone don't prove AI SEO value. Connect GA4 to revenue:
Set up conversion tracking for key events (form fills, purchases, demos)
Create segments for AI-referred users
Compare conversion rates and revenue per user
Calculate AI traffic's contribution to total revenue
Build a simple model: AI sessions × conversion rate × average order value = revenue attribution.
Proper GA4 configuration transforms AI traffic from invisible to actionable. Start with custom channel groups, then layer in dimensions and segments as your AI SEO program matures.
FAQs
Why does some AI traffic show as Direct?
AI assistants often open links in ways that strip referrer data. Mobile apps and embedded browsers commonly cause the issue. You can't fully eliminate direct traffic misattribution.
How accurate is AI traffic tracking?
Expect to capture 60-80% of actual AI referrals. Some will always appear as direct or get miscategorized. Focus on trends rather than absolute numbers.
Should I track AI Overviews separately from AI assistants?
Yes. Google AI Overviews require JavaScript to capture accurately and behave differently from ChatGPT or Perplexity referrals. Track them separately when possible.
What's a good benchmark for AI traffic percentage?
Most sites see 1-5% of traffic from AI sources currently. Sites with strong AI search visibility may see 10%+ in research-heavy categories.
