How to Track AI Traffic in GA4 (ChatGPT, Perplexity, Claude, Gemini)

To track AI traffic in GA4, start with Traffic acquisition, inspect Session default channel group, and compare the assistant sources in Session source / medium. GA4 now has an AI Assistant channel, so a custom channel group is an optional reporting tool rather than a prerequisite.

This guide shows how to check ChatGPT, Perplexity, Claude, Gemini, and other assistant sources, build a precise source filter, and review landing pages in an Exploration. It also explains what these reports cannot tell you: a recorded AI source does not prove a human recommendation, identify the prompt, or count every visit influenced by an AI answer.

Documentation and product domains checked on 11 October 2026. Examples below illustrate configurations, not results from a live analytics property.

What Counts as AI Traffic in Google Analytics?

For website reporting, distinguish three activities: a person clicking a website link in an assistant answer, an automated visit carrying an assistant referrer, and a crawler fetching content. Keep them separate when evaluating acquisition. A crawler request may never execute your GA4 tag, while a source label alone cannot establish who generated a recorded session.

Google’s current rules classify recognized assistant referrers under AI Assistant, with medium ai-assistant and campaign (ai-assistant). Non-ad clicks from Google’s AI Overviews and AI Mode belong to Organic Search. Older data, unrecognized sources, or manually tagged links may have different values. See Google’s default channel definitions.

Do not filter only for / referral when checking AI traffic. Inspect the source and channel together, and compare equivalent dates before interpreting a change.

Find AI Traffic in GA4: The Quick Check

1. Open the correct GA4 property and choose the reporting dates.
2. Go to Reports → Acquisition → Traffic acquisition. If the report is absent, an Editor or Administrator can add it to the reporting collection.
3. Choose Session default channel group and look for AI Assistant.
4. Switch to Session source / medium, or add it as a secondary dimension. Search for an observed source such as chatgpt.com or perplexity.ai.
5. Record the sessions, engaged sessions, and key events for each source. Review the landing pages using the Exploration below.

Use session-level dimensions for session acquisition. User acquisition answers a different question about first-time acquisition, so its totals should not be treated as an interchangeable version of this report. Google’s Traffic acquisition instructions explain the report and its navigation.

If the AI Assistant row is empty, check individual sources before concluding there are no AI visits. A missing row can reflect your dates, the collected source values, or a lack of matching sessions. It is not evidence that your site has never appeared in an AI answer.

Assistant Domains to Check in Your Reports

The table lists assistant web hosts to investigate. These are not a guaranteed list of emitted referrers: apps, redirects, link tags, and browser policies can change what reaches your site. Confirm the actual Session source and Page referrer in your own data before extending a rule. Google’s recognition list can also differ from your custom list.

Provider links in the first column establish product hosts, not a promise about every outbound click. A source is a processed attribution value; Page referrer is an event-level URL when collected. A tagged source such as chatgpt is not the same string as chatgpt.com. Add non-domain aliases only after checking their origin. Google’s page_referrer reference and campaign tagging guide explain these inputs.

A Copyable Regex for Assistant Source Domains

Use this starter pattern with Source → matches regex in a custom channel, or Session source → matches regex in an Exploration. It matches exact domain boundaries and subdomains of the listed assistant hosts.

^([a-z0-9-]+\.)*(chatgpt\.com|chat\.openai\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|copilot\.com|copilot\.microsoft\.com|chat\.deepseek\.com|grok\.com|meta\.ai|chat\.mistral\.ai)$

Select and copy the entire line, including ^ and $. This is our conservative starter definition, not Google’s complete assistant recognition list. It intentionally excludes generic google.com, bing.com, openai.com, x.com, and facebook.com.

It matches chatgpt.com and www.perplexity.ai, but excludes notchatgpt.com, chatgpt.com.example.org, and a full URL such as https://chatgpt.com/. Do not paste it unchanged into a Page referrer URL filter. GA4 uses full regex matching by default; case sensitivity can differ by context, so inspect your source strings and test your condition. See Google’s regex reference.

Create an Optional “AI Assistants” Channel Group

A custom definition is useful when you need a maintained source roster, historical comparison, or a grouping specific to your organization. Creating it does not recover lost referrers or validate the visits.

1. With Editor or Administrator access, open Admin → Data display → Channel groups.
2. Select Create new channel group, using the default group’s channels as the starting point. Name the group AI source review.
3. Choose Add new channel; name the channel AI Assistants.
4. Add a condition group, choose Source, select matches regex, and paste the domain pattern above.
5. Select Save channel. Choose Reorder, move your new channel above the existing AI Assistant and Referral channels, and click Apply.
6. Select Save group. In Traffic acquisition, choose the session dimension for your new group and inspect its AI Assistants row.

The first matching channel wins. Custom group definitions can reclassify available historical report data; the primary channel group has different forward-looking behavior. Keep the primary group unchanged for this exercise. Google’s custom channel group guide describes ordering, permissions, and limits.

The source-only rule is deliberately easy to audit. If you need only certain mediums, add an explicit medium condition after inspecting real data. Do not require referral alone, because that would exclude sources recorded with ai-assistant. If paid links share the same sources, define how your paid channels should take precedence and document that choice.

Build an AI Traffic Exploration by Landing Page

Use this suggested table to answer two practical questions: which assistant sources are recorded, and which entry pages receive those sessions?

1. Open Explore → Free form. Name the Exploration AI sources and landing pages, select dates, and use the Table visualization.
2. Import dimensions: Session source, Session medium, Session source / medium, Session default channel group, and Landing page + query string.
3. Import metrics: Sessions, Engaged sessions, Engagement rate, and Key events. Add revenue only when your implementation collects it meaningfully.
4. Put Session source / medium and Landing page + query string in Rows; put your chosen metrics in Values.
5. Add a filter on Session source → matches regex using the pattern above. Save this as your audited source view.
6. Duplicate the tab, remove that source filter, and filter Session default channel group → exactly matches → AI Assistant to compare Google’s grouping.

Use two tabs rather than applying both filters to one table: Exploration filters combine with AND, which could hide sources present in only one definition. Google’s Free-form Exploration documentation explains rows, values, and filters.

Sort by Sessions to find entry pages, then examine meaningful key events. A large automated test, or repeated visits to one page, can dominate the totals without demonstrating customer demand. Keep test activity separate before comparing assistant performance with search or social acquisition.

Why AI Visits Can Appear as Direct

GA4 reports (direct) / (none) when it lacks usable source information. Someone may copy a link into a new tab, open it from an app that does not pass a referrer, or reach your site through a redirect that loses tags. Google’s Direct traffic explanation covers missing attribution and redirect issues.

Link-level noreferrer and referrer policies can also suppress the referring origin. A policy that sends only the origin may preserve the domain while hiding the conversation path. These behaviors are documented in the MDN Referrer-Policy reference.

Do not relabel every Direct session as AI traffic. Without a captured referrer or trustworthy campaign tags, its origin is unknown. A custom channel rule cannot reconstruct a missing source. When you control a genuine shared link, use consistent campaign tags and verify they survive redirects; avoid rewriting unknown traffic just to make the AI report larger.

Check That Your Tracking Setup Works

Use a test property or a clearly isolated destination when possible. Write down the test time, browser or app, initial link, final landing URL, and expected source. Then follow the same route a visitor would use.

1. Check collection. Confirm the correct Google tag runs on the destination and records a page view. Use Realtime or DebugView to inspect collection; Google provides tag troubleshooting guidance.
2. Check the link path. Open an actual citation link from an assistant’s web answer. Inspect the landing URL for preserved campaign tags and, when available, the page_referrer parameter on the collected page view.
3. Check processed acquisition. Revisit Traffic acquisition and the Exploration using the same dates and property timezone. Google notes that processing can take 24–48 hours and recent data can change; see data freshness.
4. Check classification. Compare the actual source and medium with the pattern, then compare the custom channel with the default AI Assistant channel. Investigate differences instead of forcing the totals to match.

Use a clean browser session for a controlled check, while following your site’s consent choices. One successful click confirms one collection path; it does not validate every assistant, browser, app, or returning-session attribution path. Realtime activity alone is not a final session-acquisition verdict.

For DebugView, enable debug mode for your test device with Google Tag Assistant or preview mode, then select that device in Admin → Data display → DebugView. Google’s DebugView instructions explain device selection and privacy controls that can prevent events from appearing.

Useful negative checks include confirming that ordinary Google Search, Facebook, and lookalike domain names stay outside the source pattern. Record any genuine new source before editing the list, then retain the previous definition and the change date for comparison.

Troubleshoot Missing or Inconsistent AI Traffic

• No AI Assistant row: inspect raw session sources and remove restrictive medium filters. Check your dates and whether any matching sessions were collected.
• Source exists but your custom channel is empty: check exact strings, condition scope, extra medium rules, and channel order. Domain-only patterns do not match full URLs or arbitrary UTM aliases.
• Source appears as Direct: check redirects, tags, app behavior, and referrer policy. Keep the attribution unknown when the source was not captured.
• Explore and Reports differ: align dates, dimensions, filters, and definitions. Inspect the data quality indicator and your Exploration retention window before comparing totals. Google explains differences between reports and Explorations.
• Controlled tests disappear from reports: review internal or developer traffic filters with your property administrator. An active exclusion can remove incoming test data; DebugView collection does not necessarily mean that data enters reports. See Google’s data filter documentation.
• Visits are high but business outcomes are weak: review the actual landing pages and independently verified leads or sales. A source label and engagement count do not prove recommendation quality.

AI Traffic Analytics FAQs

Can I track ChatGPT traffic without creating a custom channel?

Start with the default AI Assistant channel and inspect individual session sources. Create a custom group when you need a different maintained definition or an audited historical comparison.

Does a ChatGPT source prove that a real person clicked an answer?

No. Attribution describes the recorded source; it does not authenticate a human visitor or identify a recommendation. Separate natural citation clicks, controlled tests, and automated campaigns when interpreting results.

Can I identify the exact prompt or AI answer in GA4?

A source domain does not provide that information. A collected referrer may contain only an origin, and GA4 source reporting is not a record of the assistant’s conversation.

Why are assistant sources missing from my regex report?

Check the received string first. A full referrer URL, a non-domain UTM alias, or an unlisted host will not match this domain-only pattern. Investigate the origin before changing the rule.

Should I count purchased test visits as earned AI traffic?

Keep them separate. Purchased visits can check a configured reporting path, while earned acquisition analysis should describe genuine audience activity and its independently verified outcomes.

Validate Your AI Channel with a Controlled Test

A manual click from an assistant answer is a useful first test. For repeatable source-rule checks, another option is a small campaign to test your AI channel with referrer-tagged visits.

Seovisitor’s AI Traffic service supplies automated visits carrying selected AI referrers. Such a test can help check collection and classification. It does not demonstrate that real assistant users recommended your site, create citations, improve AI search visibility, or validate the browser behavior of every assistant app. Keep its test window and destination separate from your earned AI acquisition analysis.