Use GA4’s AI Assistant default channel to measure recognized visits from ChatGPT, Gemini, Deepseek, Copilot, Grok and other chatbot sources. In release notes dated May 13, 2026, Google documented that these referrals use medium ai-assistant, channel AI Assistant and campaign (ai-assistant). Then compare that traffic with key events and revenue—not sessions alone. For assistants or link formats GA4 does not recognize, create a custom channel group, tag links you control with UTMs, and reconcile results through the Reporting API or BigQuery.
What GA4’s AI Assistant channel measures
Google defines AI Assistant as the channel through which users arrive from sources such as ChatGPT, Gemini, Deepseek, Copilot or Grok. The default channel group documentation lists the recognized chatbot referrals and distinguishes them from Google AI Overviews and AI Mode, which are excluded from this channel. See Google’s default channel group definitions.
As an Amazon Associate I earn from qualifying purchases.
When GA4 receives a recognizable referrer, the collected traffic is classified as follows:
Free tools Windows power users keep installed
One-click scans. No signup required.
| GA4 field | Value for recognized AI referrals |
|---|---|
| Default channel | AI Assistant |
| Medium | ai-assistant |
| Campaign | (ai-assistant) |
This classification applies to a click that actually reaches your site with a usable referrer. It does not reveal every visit influenced by an AI answer. A user may read an answer, remember your brand, and return later without a referrer; GA4 cannot infer that unclicked influence.
What counts as an observable visit
GA4 can measure only what reaches the property and passes collection, consent and attribution checks. A chatbot must send a browser to your site, the landing request must preserve the relevant referrer or campaign parameters, and the page must load your analytics tag. If any link in that chain fails, the visit can still exist in server logs or a CRM while appearing as direct, referral or unassigned in GA4. Treat the AI Assistant row as measured click traffic, not as a count of every person exposed to an answer.
Step 1: Find AI referrals in the GA4 interface
- Open Reports in the GA4 property.
- Select Acquisition > Traffic acquisition.
- Use the table’s dimension selector to choose a session-scoped dimension such as Session default channel group, Session source, Session medium or Session campaign.
- Filter or search for AI Assistant,
ai-assistantor(ai-assistant). - Set the date range you need and record sessions, engaged sessions, key events and revenue.
Traffic acquisition is designed for session-scoped acquisition questions. Google documents the available dimensions and metrics in the Traffic acquisition report help page. A session source answers “How did this session begin?” It is not the same as the source that receives credit for a conversion under your property’s attribution model.
Use the right columns
Add engagement and business-outcome metrics to the report rather than judging AI traffic by volume. Useful columns include sessions, engaged sessions, engagement rate, key events, session key-event rate, total revenue and ecommerce purchases. If your property has leads rather than ecommerce, mark the lead submission as a key event and use its event count and value.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Keep the date range consistent when comparing assistants. A short spike in sessions can look impressive while producing no qualified leads, whereas a smaller source may generate more revenue per session. Export or record the same dimensions and metrics each period so a change in the interface does not silently change your trend.
Break out individual assistants
After confirming the AI Assistant channel, switch the table to Session source or Session source / medium. This can show separate sources when the referrer is passed distinctly. Keep the channel as your roll-up for trend reporting, and use source-level rows to identify which assistant sends the traffic. Recognition depends on the referrer matching Google’s maintained definitions, so a source may appear as referral or direct when the referrer is missing or changes.
Check the release boundary
Google’s documented change is dated May 13, 2026. When reporting a long period, annotate that date and inspect the data before and after it separately. A sudden increase in the AI Assistant row may reflect the new classification rather than a sudden change in user behavior. Compare the underlying session source and medium rows as well as the default-channel total, and explain any break in the reporting notes.
Step 2: Build a custom AI channel group as a fallback
A custom channel group gives you control when native recognition does not cover an assistant, a referrer format, or a naming convention you need. Google provides a procedure and example rules in its custom channel group documentation.
- Open Admin > Data display > Channel groups.
- Copy an existing channel group so the original default group remains available.
- Name the copy clearly, then add a new channel named AI assistants.
- Create a condition based on the relevant traffic-source dimension, normally the session source or source platform used by your property.
- Use a regular expression that matches the assistant domains you want to monitor. Google’s example covers ChatGPT/OpenAI, Gemini, Copilot, Claude and Perplexity patterns; adapt that documented pattern to the sources relevant to your site.
- Move the AI assistants channel above Referrals. Channel rules are evaluated in order, so the specific rule must run before the broad referral rule.
- Save the channel group and select it in acquisition reports.
Analyze the custom group in both User acquisition and Traffic acquisition. User acquisition describes how users were first acquired, while Traffic acquisition describes how sessions began. Keep the regex under change control and update it when assistant URLs or redirect domains change. A custom group changes how reports classify matching data; it does not recreate a referrer that was never sent.
Design and test the regex carefully
Start with the exact hostnames visible in your session-source dimension, then expand only when you can verify another hostname belongs to the assistant. Test uppercase and lowercase behavior, redirect hostnames and any regional domains represented in your data. A rule that is too broad can absorb ordinary referral traffic; a rule that is too narrow leaves assistant visits in Referrals. Record the expression, the date it changed and the person who approved it so future comparisons remain interpretable.
Native channel or custom group?
| Decision factor | Native AI Assistant | Custom channel group |
|---|---|---|
| Coverage | Uses Google’s maintained recognition for documented chatbot referrals. | Can include additional assistants, domains or internal naming rules you specify. |
| Rule control | Google controls the recognition list and definitions. | You control the regular expression, conditions and channel order. |
| Setup effort | Available in properties where the default channel is documented and populated. | Requires copying a group, creating rules and maintaining them. |
| Best use | Consistent reporting of Google-recognized AI referrals. | Coverage gaps, custom governance or a dedicated reporting taxonomy. |
| Historical comparison | Use the documented release date and inspect how older rows were classified. | Keep a versioned record of rule changes and compare periods using the same group definition. |
Step 3: Tag links that you control
If you publish a link in an AI assistant workflow, community answer, newsletter or campaign where you control the destination URL, add UTM parameters. GA4 maps these values into campaign and traffic-source dimensions during collection and processing. Google’s parameter guidance is at Campaign URL Builder and campaign data documentation.
Choose a naming convention
Use lowercase, stable values and document them for everyone who creates links. A practical pattern is:
utm_source: the platform or distribution source, such aschatgptorgemini.utm_medium: a consistent category such asai-assistant.utm_campaign: the initiative or content promotion, such asproduct-guide-2026.utm_content: an optional placement or creative identifier.utm_id: an optional stable campaign ID for joins with other systems.
Do not mix spelling or capitalization for the same source. Decide whether source values identify the assistant, the publishing surface or both, and keep that decision in a written convention. For example, use one source value for the assistant and use utm_content for the specific answer, placement or link position. That preserves clean source reporting while retaining detail for analysis.
Protect parameters through redirects
Test the complete click path, including shorteners, tracking redirects, consent pages and cross-domain transitions. Parameters must arrive at the landing page and be available when GA4 collects the session. Google explains that stripped or unavailable campaign information can leave a visit as (direct) / (none) in its direct-traffic guidance. A redirect that drops the query string is a common cause.
Use a real test click for every new link pattern. Capture the original URL, each intermediate URL and the final landing URL. Confirm that the query string survives, that the expected source and medium appear in DebugView or the acquisition report, and that the destination does not rewrite the values unexpectedly. Shorteners and consent flows should be treated as part of the measurement implementation, not as separate marketing infrastructure.
Step 4: Measure conversions with the correct attribution scope
Define the business result before evaluating the channel. In GA4, mark the event representing that result as a key event—for example, a completed signup, qualified lead submission, purchase or subscription start. Record the same date range for AI sessions, key events and revenue.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Clear out junk files and repair common Windows errors3Scan for outdated or missing drivers - takes under a minuteSession acquisition versus conversion credit
Session-scoped dimensions describe the source that started a session. Event-scoped traffic-source dimensions and the property’s attribution model assign credit for a key event. These are different questions:
- Session question: How many sessions began after an AI assistant click, and how engaged were they?
- Conversion question: Which sources receive credit for the key event under the selected attribution model?
Do not add session-source conversions to event-attributed conversions as if they were the same measure. Use the scope that matches the decision, and state it in dashboards and reports. Google describes these scopes and attribution dimensions in the acquisition reporting documentation and the campaign and attribution guidance.
For a campaign review, you might report sessions and engaged-session rate by session source, then report key-event credit by the event-scoped dimensions and selected attribution model. Those figures can legitimately differ. Label tables as “session acquisition” or “event attribution” rather than using a generic “AI conversions” heading.
Evaluate quality, not just volume
For each AI source or channel, calculate the relationship between sessions and outcomes: engaged-session rate, key-event rate, revenue per session and revenue per user where those metrics fit your implementation. Segment by landing page, device, geography and new versus returning users only when those dimensions are collected consistently. A high session count with no meaningful key events may indicate curiosity traffic, weak landing-page relevance or broken conversion tracking.
Validate the key event itself before judging an assistant. Confirm that the event fires once at the intended completion step, carries a value when appropriate, and is not blocked on one domain but allowed on another. Compare GA4 purchase or lead totals with the order system or CRM for the same period; a channel cannot be evaluated accurately when the underlying outcome is incomplete.
Step 5: Reconcile GA4 with the Reporting API and BigQuery
The GA4 interface is useful for fast investigation, but exported data is better when you need repeatable joins to CRM records, order systems or lead-quality outcomes.
Reporting API
The Google Analytics Data API exposes first-user, session and event traffic-source dimensions. Use it to automate scheduled extracts, apply the same date filters every day and compare channel performance with another system. The available schema is documented at the Reporting API schema reference.
Choose dimensions that match the question: first-user fields for the user’s initial acquisition, session fields for the current visit, and event fields for key-event analysis. Keep the query’s date range, timezone, filters and attribution-related settings in version control. An automated extract is useful only when the definition stays stable enough to compare one run with the next.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →BigQuery export
BigQuery export provides traffic-attribution data at user, session and event levels, including session last-click manual campaign fields. This level of detail lets you join a session or event to a CRM lead, subscription or order ID and investigate disagreements with the GA4 interface. See Google’s BigQuery traffic-attribution documentation.
- Choose a stable join key available in both systems, such as a transaction ID or internally generated lead ID.
- Extract the relevant user, session and event traffic-source fields for the same timezone and date window.
- Group by your native AI Assistant channel, custom channel, source and campaign values.
- Reconcile totals with GA4 reports, allowing for reporting delays, consent-related gaps and different attribution rules.
- Document the exact scope and attribution model beside every KPI.
Use the interface for quick questions, the Reporting API for repeatable scheduled summaries, and BigQuery when row-level joins or investigation are required. They are complementary views of collected GA4 data, not independent counts that should be summed together.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Why AI traffic appears as referral, direct or unassigned
Referral instead of AI Assistant
The referrer may not match Google’s maintained recognition list, or your custom channel rule may not match the actual hostname. In a custom group, check the regular expression, the source dimension selected in the rule and the channel’s position above Referrals.
Direct / (none)
Google defines (direct) source and (none) medium as traffic without a clear referral source. This can result from privacy controls, copied or typed URLs, apps that do not pass referrers, redirects that remove UTMs, shorteners, offline links, consent choices or cross-domain configuration problems. Confirm the final landing URL and inspect every redirect for preserved query parameters.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Unassigned or unexpectedly low volume
Check that the property is receiving the expected page_view and session data, that consent settings allow analytics storage where legally appropriate, and that cross-domain links are configured consistently. Compare the raw landing URL, source and medium values in an export with what the interface displays. Recognition cannot recover a source that was never transmitted.
A sudden change after a site release
Compare the timing of the change with analytics-tag deployments, consent-banner updates, redirect changes and cross-domain configuration. Test a controlled URL from each important assistant and inspect the resulting source, medium and campaign. If only one hostname changes, review that assistant’s redirect behavior rather than changing every channel rule.
A practical measurement routine
- Weekly: Review AI Assistant sessions, engaged sessions, key events and revenue by source and landing page.
- After a taxonomy change: Test a tagged URL, verify its source, medium and campaign, and check the custom channel ordering.
- Monthly: Compare GA4 totals with CRM-qualified leads or completed orders using API or BigQuery extracts.
- When assistant traffic spikes: Inspect referrer values, redirect behavior, consent rates and the pages receiving visits before interpreting the increase as demand.
- After an assistant changes domains: Check source rows, update the documented regex if needed, and record the effective date so trend reports explain the classification change.
What GA4 can—and cannot—tell you
GA4 measures observable clicks and assigns key-event credit according to its configured collection and attribution rules. It cannot prove that an AI answer influenced a person who never clicked, cannot identify an assistant when the referrer was stripped, and cannot make offline or consent-blocked activity visible. Treat AI referral reports as a measured subset of AI-influenced demand, not a census of every conversation.
The most defensible report therefore shows the native AI Assistant channel where available, a documented custom channel for gaps, tagged links for controlled placements, and conversion results labeled by their session or event attribution scope.
Frequently Asked Questions
Does GA4 automatically track ChatGPT traffic?
GA4 documents an AI Assistant default channel for recognized chatbot referrals, including ChatGPT. Traffic is classified only when a usable referrer reaches your property; otherwise it may appear as referral or direct/(none).
How do I see AI referrals in GA4?
Open Reports > Acquisition > Traffic acquisition, choose a session source, medium, campaign or default-channel dimension, and filter for AI Assistant, ai-assistant or (ai-assistant).
Which assistants are included in GA4’s native channel?
Google lists sources such as ChatGPT, Gemini, Deepseek, Copilot and Grok. Google’s documentation excludes Google AI Overviews and AI Mode from the AI Assistant channel.
Why is an AI visit showing as direct traffic?
The referrer or UTM parameters may have been removed by an app, privacy control, redirect, shortener, consent choice or cross-domain issue. Verify the final URL and every redirect in the click path.
Recommended Free Tools
Should I use session source or event attribution for AI conversions?
Use session source to describe how a session began. Use event-scoped dimensions and your property’s attribution model to analyze credit for key events; the two scopes answer different questions.
When should I create a custom channel group?
Create one when native recognition does not cover an assistant or referrer format, or when you need your own naming and governance. Match the actual source hostname, place the rule above Referrals and maintain the regex as URLs change.
Do UTM parameters change the AI Assistant channel?
UTMs populate campaign and traffic-source dimensions for links you control. Use a documented source, medium and campaign convention, and preserve the query string through redirects. Native classification and your custom taxonomy should be analyzed deliberately rather than mixed without labels.
Can BigQuery identify every AI-influenced conversion?
No. BigQuery provides user-, session- and event-level attribution for data GA4 collected, but it cannot recover visits with no referrer, stripped UTMs, blocked consent or no click.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallWhy do session conversions and attributed key events differ?
Session dimensions describe how the visit began, while event-scoped dimensions and the property’s attribution model assign credit for a key event. Different scopes can produce different totals without either report being incorrect.
The Bottom Line
Measure AI referrals in layers: use GA4’s native AI Assistant channel, add a documented custom channel for coverage gaps, preserve UTMs on links you control, and report key events and revenue with the scope and attribution model clearly labeled. Validate redirects, consent and cross-domain tracking, then use the Reporting API or BigQuery when you need repeatable reconciliation with CRM or order data.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
