iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
To track referrals from ChatGPT, Claude, Gemini and Perplexity, open Reports → Acquisition → Traffic acquisition in GA4 and inspect Session source, Session medium and Session source / medium. Add Landing page plus key-event or revenue metrics to see which assistant sends visitors, where they arrive and what they do. GA4 now also offers an AI Assistant default channel, identified by the exact medium ai-assistant. For historical consistency and sources GA4 does not classify automatically, create a custom channel group that matches the assistant domains you actually observe. This measures attributed, tagged visits that execute your GA4 tag—not every visit influenced by an AI answer.
What does GA4 record when someone clicks an AI assistant link?
GA4 builds attribution from campaign fields, UTM parameters and the document referrer. Those inputs become dimensions such as source, medium and campaign. A single assistant can therefore appear under different labels when one click includes a campaign parameter, another exposes a referrer, and a third supplies neither.
For example, a click may be recorded with a source resembling an assistant domain, with ChatGPT’s documented utm_source=chatgpt.com, or with no usable referral information at all. The report label reflects the data that reached your site and the processing rules active in your GA4 property; it is not a guaranteed census of every person who encountered your content in an assistant.
| GA4 field | What it tells you | Why it can vary |
|---|---|---|
| Session source | The source value assigned to the session | Referrer and campaign processing can produce different source names |
| Session medium | The medium assigned to the session | Google’s AI Assistant channel uses the exact value ai-assistant; other visits may use a referral or campaign medium |
| Session source / medium | The combined diagnostic view | It exposes differences that are hidden when you look at source or medium alone |
| Session campaign | The campaign value attached to the session | Recognized AI Assistant traffic can receive (ai-assistant); UTM campaigns or absent values produce other results |
Start with the combined source/medium dimension rather than assuming every assistant click will use one universal naming convention.
Popular on Amazon
How do I see ChatGPT, Claude, Gemini and Perplexity traffic in GA4?
The fastest check is a Traffic acquisition report filtered or searched for the assistant names, then expanded with landing-page and outcome metrics.
- Sign in to the correct GA4 property and open Reports → Acquisition → Traffic acquisition.
- Set the date range you want to analyze. Record the dates before exporting results, because channel definitions and attribution data can change over time.
- Use the table’s primary dimension selector to choose Session source / medium.
- Search the rows for
chatgpt,claude,geminiandperplexity. Also inspect rows containing the domains or source names you see in your own property; capitalization and host-name formatting are not guaranteed to be identical across visits. - Add Landing page as a secondary dimension. This shows which URL each source opened rather than only how many sessions it generated.
- Keep the standard engagement columns visible, then add your relevant key events and revenue metrics. A high session count is less useful than knowing whether those sessions reached a signup, purchase or other business outcome.
- Compare the assistant rows with Referral, Organic Search and Direct for the same date range. Save the report or export it with the channel-group definition noted in your analysis.
If a name does not appear, do not conclude that the assistant sent no visitors. First check the AI Assistant channel, campaign values and rows that lack an obvious brand name; attribution can be grouped differently when referrer or campaign data is missing.
What is GA4’s AI Assistant default channel?
GA4’s current default channel group includes AI Assistant. Google defines it as the channel by which users arrive from sources such as ChatGPT, Gemini, Deepseek, Copilot or Grok. The rule uses the exact medium ai-assistant; when Google recognizes the referrer as an AI assistant, GA4 sets that medium and uses (ai-assistant) as the campaign value.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Google’s release note dates this capability to May 13, 2026. It is therefore useful for current reporting, but older sessions may have been classified before the rule existed or under a different channel definition. When you compare periods, preserve the date range and channel-group version in your notes.
- In Traffic acquisition, change the primary dimension to Session default channel group.
- Locate AI Assistant and select it to examine sessions, landing pages, key events and revenue.
- Drill into the source/medium dimension for that channel when you need to distinguish ChatGPT from Claude, Gemini, Perplexity or another recognized assistant.
The default channel is a convenient roll-up, not a replacement for source-level diagnosis. Keep the source/medium view available when validating whether a particular assistant is being recognized correctly.
How do I create a custom AI-assistant channel group?
Create a custom channel when you need a consistent historical view, when your property has source names outside the built-in rule, or when you want an explicit list of assistants. Put the custom channel above Referrals so matching sessions are claimed by it before the broader referral rule.
Also popular on Amazon
- Open Admin → Data display → Channel groups.
- Create a new channel group or copy the group you use for reporting, then add a channel named AI assistants.
- Set the matching condition to Session source and use a regular expression that covers the domains and source strings you actually observe.
- Place the AI assistants channel above Referrals in the channel order.
- Save or publish the group only after reviewing the expression against current source rows. An overly broad expression can classify ordinary Google or referral traffic as AI traffic.
Google’s example names ChatGPT, Gemini, Microsoft Copilot, Claude and Perplexity. A practical starting expression supplied for that purpose is:
^.*ai|.*.openai.*|.*chatgpt.*|.*gemini.*|.*gpt.*|.*copilot.*|.*perplexity.*|.*google.*bard.*|.*bard.*google.*|.*bard.*|.*.*gemini.*google.*$Treat that expression as a starting point, not a universal standard. Review your own source values before publishing: the .*gpt.* and broad .*ai alternatives can match unrelated names, while assistant hosts may change. Remove terms that create false positives and add only source patterns you can verify in your property.
Custom channel groups do not rewrite the original source, medium or campaign values. They provide another reporting classification, so you can retain the raw diagnostic dimensions while presenting a stable AI-assistant roll-up.
How should I preserve ChatGPT’s UTM parameter?
OpenAI states that ChatGPT automatically adds utm_source=chatgpt.com to referral URLs. That parameter can be used by analytics platforms such as Google Analytics to identify the source. Do not remove it in redirects, link shorteners or canonicalization rules.
- Choose a page that can receive a ChatGPT referral and copy its final URL.
- Follow the link from ChatGPT in a test environment, then inspect the URL at your server, CDN and analytics entry points. Confirm that
utm_source=chatgpt.comsurvives every redirect. - In GA4 DebugView or a near-real-time acquisition report, verify the resulting source and campaign values after the GA4 tag executes.
- Check the same visit in Traffic acquisition after processing completes. Compare the UTM-derived row with any referrer-derived row rather than assuming the labels will be identical.
A redirect that strips query parameters can turn a recognizable campaign into an unattributed or differently classified session. Preserve the parameter while still enforcing your site’s normal security and URL rules.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchHow can I separate Claude, Gemini and Perplexity visits?
Use source-level rows and landing-page breakdowns rather than relying only on the aggregate AI Assistant channel. The channel answers “how much AI-assistant traffic arrived”; source and source/medium answer “which assistant did GA4 identify.”
Also popular on Amazon
Claude
Search Session source / medium for rows containing claude, then add Landing page and key-event metrics. If the source is absent, inspect the AI Assistant channel and any referral rows whose host names you can verify as Claude; the lack of a visible name can reflect missing referral or campaign data.
Gemini
Search for gemini and review source variants rather than merging them manually without documentation. A custom channel expression should match the source values you actually receive, because broad terms can capture unrelated domains.
Perplexity
Search for perplexity in Session source / medium and compare its landing pages and key events with the other assistants. Keep the date range and channel definition alongside any exported result so later reports remain comparable.
Use a consistent comparison view
| Comparison axis | Recommended GA4 field or metric | Question it answers |
|---|---|---|
| Assistant identity | Session source and Session source / medium | Which source label did GA4 receive and assign? |
| Classification | Session default channel group | Was the session grouped as AI Assistant, Referral or another channel? |
| Entry page | Landing page | Which content did each assistant open? |
| Volume | Sessions and engaged sessions | How much traffic arrived and how much met GA4’s engagement definition? |
| Business outcome | Key events and revenue | What actions or value followed the visit? |
Use the same attribution settings, date range and channel group for every assistant in a comparison. Otherwise a change in definitions can look like a change in performance.
Why is my AI traffic showing as Direct or Referral?
GA4 can only attribute a visit when usable campaign or referral information reaches the site and the GA4 tag executes. Google’s documentation defines Direct as traffic for which no referral-source information is available. A missing referrer, stripped query string, privacy behavior, app handoff or tagging failure can therefore move an assistant-influenced visit into Direct or another broad category.
When it appears as Direct
- Check whether redirects preserve query parameters, especially ChatGPT’s
utm_source=chatgpt.com. - Confirm the landing page loads the intended GA4 tag and that consent settings do not prevent the session from being recorded.
- Review whether the click crossed an app, browser or privacy boundary that removed the document referrer.
- Use Landing page and timing patterns as context, but do not relabel Direct sessions as AI traffic without evidence.
When it appears as Referral
- Inspect the exact Session source value and host name before changing channel rules.
- Check whether the source matches your custom AI-assistant regex and whether the AI channel is ordered above Referrals.
- Remember that changing a channel group affects reporting classification, not the underlying event collection.
The visible AI total should therefore be described as attributed AI-assistant traffic, not as the total number of sessions influenced by AI systems.
Also popular on Amazon
What should I measure beyond AI-referral sessions?
Sessions alone can reward an assistant that sends many low-intent visits. Evaluate each source on acquisition, engagement and outcome dimensions, and compare it with Referral, Organic Search and Direct under the same reporting setup.
Free tools Windows power users keep installed
One-click scans. No signup required.
- Sessions: the volume of recorded visits attributed to the source or channel.
- Engaged sessions: the subset that meets GA4’s engagement criteria, useful for separating quick bounces from substantive visits.
- Landing pages: the pages that assistants expose to users, which can reveal content gaps or unexpectedly strong entry points.
- Key events: the actions your organization has marked as meaningful, such as a signup or lead submission.
- Revenue: the monetary outcome available in your implementation, interpreted with the same attribution model and date range.
Export source, medium, landing page, sessions, engaged sessions, key events and revenue together when stakeholders need to reproduce the comparison. Include the channel-group definition and the report dates in the export name or accompanying notes.
There is no authoritative cross-site percentage or conversion benchmark in the official material for these assistants. Avoid declaring that one assistant “converts better” from a small or differently classified sample; report your property’s observed counts and the measurement limitations instead.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Can GA4 measure clicks from Google AI Overviews or AI Mode?
Not through the AI Assistant default channel. Google’s documentation explicitly excludes Google AI Overviews and AI Mode from that channel. Their visibility and search-result interactions belong in Search Console’s Generative AI performance report.
Search Console’s report covers AI Overviews and AI Mode with impressions broken down by pages, countries, devices and dates. Its methodology counts clicks on links in those surfaces as Search clicks. GA4 and Search Console therefore answer different questions:
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →| Tool | Primary question | What to use it for |
|---|---|---|
| GA4 | What happened after an attributed click reached my site? | Sessions, landing pages, engagement, key events and revenue |
| Search Console Generative AI performance report | How visible were my pages in Google’s generative-search surfaces? | Impressions and clicks for AI Overviews and AI Mode by page, country, device and date |
Do not add Search Console impressions to GA4 sessions or treat a Search click as an AI Assistant referral. Join the reports by page and date only when you clearly label the different measurement scopes.
Also popular on Amazon
How do I validate an AI-referral implementation?
Validation should prove both collection and classification. Run these checks whenever you add a channel, change redirects or notice a sudden shift in source names.
- Verify the tag: confirm the destination page executes the GA4 tag and that your consent configuration permits measurement in the test case.
- Verify the URL: inspect the final address after every redirect and confirm expected UTM parameters remain intact.
- Verify raw dimensions: find the test session in source, medium, source/medium and campaign views rather than checking only the default channel.
- Verify the custom rule: compare the exact source string with your regular expression and confirm the AI assistants channel is above Referrals.
- Verify the outcome: trigger a test key event only in a controlled environment, then confirm that the event and its value are associated with the expected session.
- Document the setup: record the test date, property, channel-group version, source values observed and any redirect behavior.
Allow for normal processing delays before judging a production change. If the source appears in raw dimensions but not in the custom channel, the issue is usually rule scope or channel order; if it appears nowhere, investigate the referrer, query string, tag and consent path first.
What are the practical limits of AI-referral reporting?
GA4 reports visits that arrive with usable attribution data and execute the GA4 tag. It cannot reconstruct every AI-influenced journey when a user remembers a recommendation and later types the URL, uses a bookmark or searches independently. Those sessions may legitimately appear as Direct, Organic Search or another channel.
Recommended Free Tools
Assistant interfaces and browsers can also change how referrers and query strings are passed. A source pattern that works today should be reviewed against current rows after a platform changes its linking behavior. Keep raw source/medium reporting alongside any custom roll-up so you can detect those changes instead of hiding them behind one aggregate number.
Finally, avoid cross-site conversion claims that the available data cannot support. GA4 can show your site’s attributed sessions and outcomes; it does not provide a universal market share, traffic percentage or conversion benchmark for ChatGPT, Claude, Gemini or Perplexity.
Frequently Asked Questions
Will every ChatGPT click appear with the same source in GA4?
No. ChatGPT may add utm_source=chatgpt.com, but the final GA4 label also depends on redirects, referrer availability, campaign processing and whether the GA4 tag runs. Inspect source, medium and campaign together for the session.
Can I change an already collected session from Referral to AI Assistant?
A channel-group rule changes how eligible data is classified in reports; it does not rewrite the original events or restore referral information that never reached your site. Keep the raw source and medium dimensions when you need an audit trail.
Should I merge AI Assistant traffic with Organic Search?
Keep them separate. AI Assistant describes an assistant-originated visit recognized by GA4, while Organic Search describes a search-engine result click; combining them obscures how the visitor arrived and makes landing-page and outcome comparisons harder.
How long should I keep a custom AI channel definition?
Keep a dated record of each expression and its channel order, then review it whenever new source rows appear or an assistant changes its linking behavior. This preserves historical context when you compare reports created under different definitions.
