Don’t invent the rest.
Google Analytics 4 can measure visits attributed to AI services when a referral is passed through. It cannot measure every mention in an AI answer, every recommendation or every visit with missing referral information.
Can Google Analytics 4 show traffic from AI assistants?
Yes, GA4 can report visits that arrive with identifiable referrer information from AI assistants or related services. However, this is only a subset of AI-driven discovery. A user may read an AI answer without clicking, open a link in an app that strips attribution, or return later through direct or branded search.
Because referral data varies by product, device and user journey, no GA4 report can truthfully count every ChatGPT, Gemini or Perplexity mention of your business. The practical objective is to monitor observable visits and conversions while being transparent about what is not measured.
This distinction is important for Kenyan companies investing in SEO and AI search visibility. A rise in AI referral sessions is useful evidence of visits, but it does not prove every generated answer was favorable or that an assistant ranked the company in a stable position.
Track visits, not invisible recommendations
GA4 observes activity on your website. It cannot directly observe every answer shown inside an external AI product.
Set up a clean GA4 measurement foundation
Before building an AI traffic report, verify that your GA4 property is collecting website data reliably. Confirm the measurement ID is installed once, check realtime traffic and make sure major conversion events such as lead-form submissions or confirmed enquiries are configured correctly.
Duplicate Google tags can inflate page views and confuse engagement metrics. Missing consent implementation, blocked scripts or broken form tracking can undercount conversions. Fix these basic issues before analyzing small referral sources.
If you use WordPress with Elementor, WooCommerce or custom forms, test each important conversion action in GA4 DebugView or another appropriate debugging workflow. For lead-generation sites, a genuine form submission is usually more meaningful than a generic button click.
- Confirm GA4 is receiving visits from your live domain.
- Verify the reporting time zone and currency settings.
- Test key lead and purchase events.
- Exclude obvious internal testing traffic where appropriate.
- Check that redirects preserve normal referral information when possible.
Where to find AI referral traffic in GA4
Open Reports and locate the traffic acquisition report. Examine the session source/medium dimension, and search for recognizable referrer domains associated with AI services. The precise menu labels can vary as Google updates its interface.
Common domains you may encounter include chatgpt.com, perplexity.ai or gemini.google.com. These are examples, not an exhaustive or permanent list. Some traffic may appear under other domains, as direct, or under a different channel classification depending on the environment.
For a reusable report, create an exploration or comparison based on session source, then filter for the known sources you actually observe. Keep a log of any filter changes so historical comparisons remain meaningful.
| Dimension | What to inspect | Why it matters |
|---|---|---|
| Session source / medium | Identifiable AI referrers | Shows attributable incoming visits |
| Landing page | Which content receives visits | Identifies useful pages |
| Key events | Enquiries or purchases | Connects traffic to business outcomes |
| Engaged sessions | Meaningful interaction | Adds context beyond raw session counts |
Build a reusable AI-referral exploration
In GA4 Explore, create a free-form exploration and add dimensions such as Session source / medium and Landing page plus query string. Add metrics including Sessions, Engaged sessions and Key events where available in your property.
Apply a session-source filter containing the AI referrers you have observed. Avoid using an unreviewed giant regular expression copied from a blog: domains change, and overly broad patterns can classify unrelated sources as AI traffic.
For each reporting period, export or record the source list, dates and any changes in your filters. A consistent definition is more useful than an impressive-looking chart that silently changes its measurement rules every month.
How to identify ChatGPT, Gemini and Perplexity visits
For ChatGPT, inspect sessions attributed to chatgpt.com and any other source labels that appear in your own GA4 data. For Perplexity, look for perplexity.ai. For Gemini, inspect identifiable Gemini-related referrers, but recognize that traffic from Google’s broader AI experiences may not appear as a neat “Gemini” source.
Do not assume every Google Search visit generated by an AI Overview will have a special AI label in GA4. Search and AI surfaces can share traffic pathways, and the information available to website analytics does not fully describe the experience that preceded the click.
If a client asks for “the number of times ChatGPT recommended us,” answer that GA4 cannot provide that count. You can report attributable referral visits, observed citations from a defined manual sample, and resulting enquiries as separate metrics.
Connect AI referrals to real business outcomes
A Kenyan agency should not evaluate AI traffic only by visits. Ten qualified enquiries from a small audience may be more valuable than hundreds of visits that never reach a service page. Define what a qualified lead means for your business and track it consistently.
For example, a visitor may land on a technical SEO article, read a troubleshooting section and then visit the contact page. The useful journey includes the article, the service information and the enquiry event, not merely the original source label.
Use landing-page and conversion data to decide which articles deserve updates. If an AI-referral article receives visits but no meaningful engagement, improve its answer, examples and next-step guidance rather than immediately adding more promotional links.
A simple monthly AI visibility report for Kenyan businesses
Report the number of identifiable AI-referral sessions, the landing pages involved, key events and any qualified leads. Add a short qualitative note on notable changes, new sources or tracking limitations. Keep Google organic search performance separate so the report does not double-count or mislabel visits.
If you also monitor a small set of AI prompts manually, record the prompt wording, date, location or account context and the answer observed. Treat these observations as a sample rather than a definitive visibility rank. AI answers can vary between users and sessions.
Combine the report with relevant local service content and broader business performance. Search visibility matters because it supports real enquiries, not because a dashboard contains a large number.
| Monthly report line | Example interpretation | Do not claim |
|---|---|---|
| AI-attributed sessions | Some visits came from identified assistant referrers | Every AI mention generated a visit |
| AI landing pages | These pages attracted attributable visits | These are the only cited pages |
| Key events | Some visitors completed tracked actions | Every event is a paying customer |
| Prompt observations | These answers appeared in a defined sample | Permanent rankings across all users |
Why AI traffic may appear as Direct or disappear entirely
Some apps open links in embedded browsers or remove referral details. Privacy settings, browser restrictions and tracking prevention can also affect what GA4 records. A user who reads an AI answer and later types your URL manually may be recorded as direct traffic.
GA4 attribution is useful but not omniscient. Even when a source appears, it does not necessarily reveal the exact prompt or the wording of the recommendation. Avoid trying to infer personal user queries from anonymous analytics data.
If your website is slow or scripts fail to load, analytics may undercount visits from every channel. Maintaining a reliable WordPress website is part of sound measurement.
Common dashboard mistakes to avoid
Do not create a custom “AI channel” that mixes Google organic search, all direct visits and a handful of AI referrers, then present the combined total as verified AI traffic. That makes the number larger but less useful.
Do not rely on source/medium alone to judge content quality. A single technical guide may assist a later conversion through a different channel. Likewise, a spike from an assistant may be temporary and may not repeat next month.
Finally, review event naming and consent settings before comparing periods. If your form event was broken last month, the conversion trend may reflect measurement changes rather than actual lead performance.
How to create a report that a business owner can actually use
Start with a one-page monthly summary rather than a wall of charts. State the reporting dates, the source filters used and the number of attributable sessions from the AI-related referrers you have observed. Then show the landing pages and any meaningful conversion events. A short explanation of uncertainty should be included directly beside the figures.
Next, identify a decision. If one technical article attracts engaged visitors but produces few enquiries, the action may be to improve the next-step guidance. If a service page attracts relevant traffic but people leave quickly, investigate clarity, mobile usability and the offer itself. If the source count is tiny, avoid declaring a trend from a handful of visits.
Compare with Google organic search and direct enquiries without adding all channels together under the label “AI visibility.” A lead may have several touchpoints, and GA4 attribution settings influence which channel receives credit. A useful report explains these boundaries instead of hiding them.
For a company selling websites or online services, link the report to actual commercial outcomes. A digital marketing strategy should connect visibility, landing-page quality and customer enquiries, rather than rewarding whichever channel produces the most impressive-looking graph.
What to do when referral counts suddenly rise or fall
First check for measurement changes. Did the Google tag move, did consent settings change, did a redirect get added or did the reporting filter change? A sudden shift in attribution may be caused by technical changes rather than user demand.
Then examine landing pages, session engagement and key events. If the same page suddenly receives many short visits from one source, inspect whether the traffic appears genuine and whether it matches the content. If a valuable article loses traffic, check whether it remains available, indexed and up to date.
Do not automatically rewrite a successful article after one quiet week. AI products change their answers and links frequently, and small data sets are volatile. Record the observation, check whether the pattern persists and compare it with other business indicators before acting.
Finally, keep the underlying website dependable. Tracking depends on pages loading and events firing correctly. If site errors or script conflicts are suspected, a website revamp and maintenance review may resolve the measurement problem more effectively than adjusting dashboard filters.
Frequently asked questions
Can GA4 track traffic from ChatGPT?
Yes, GA4 can show visits with identifiable ChatGPT-related referral sources, but it cannot count every ChatGPT mention or answer view.
Why is ChatGPT traffic showing as Direct in Google Analytics?
Some apps, browsers or redirects do not preserve referral information. GA4 may classify such visits as Direct or another channel.
How do I see Perplexity referrals in GA4?
Inspect the traffic acquisition report or an exploration using session source/medium, and filter for observed Perplexity-related referrers such as perplexity.ai.
Does Google AI Overviews traffic appear separately in GA4?
Not as a complete universal AI Overviews referral category. Google Search and AI features can share traffic pathways, so do not assume a separate source label.
Can I track how often Gemini recommends my business?
Not comprehensively through GA4. You can monitor identifiable referrals and record observations from defined manual tests, but not all generated answers.
Which GA4 metric matters most for AI traffic?
For business decisions, pair attributable sessions with meaningful key events and qualified enquiries rather than relying on visits alone.
How often should I report AI search referrals?
A monthly report is a practical starting point for most small businesses. Use a consistent source definition and explain the measurement limits.
