Your clients now ask ChatGPT and Gemini which lawyer to call, and a lot of that traffic lands in your reports as “Direct,” unattributed and uncounted. The ground under search is already moving: across thousands of news sites worldwide, Google organic search referrals fell 33% globally and 38% in the US between November 2024 and November 2025. If you can’t see where your visitors come from, you can’t tell what signs cases. This guide shows you how to surface AI search traffic in GA4, what the numbers mean, and where they fall short.
Why Law Firms Should Track AI Traffic in GA4
AI search is starting to send you potential clients, and right now you’re crediting the wrong channel or none at all. When someone asks an assistant for a personal injury lawyer in their city, clicks through, and fills out a form, that session often shows up as “Direct.” You’ll credit the wrong channel, or no channel at all.
Keep the scale honest, though, because your persona will. Chatbot referrals are still small next to search. From search alone, Google delivers about 500 times as many referrals as ChatGPT, and 1,300 times as many once you count Discover. So this is an emerging signal, not a flood. Track it now because it’s growing while the rest of your search traffic shrinks, and you cannot defend or cut a channel you cannot measure. Signed cases are the test, not sessions.
What AI Traffic Looks Like in GA4 by Default
GA4 now has a native “AI Assistant” channel, but it catches less than you’d think. Google added it to the Default Channel Group on May 13, 2026. When a visit arrives from a recognized assistant, GA4 tags the medium as ai-assistant and drops the session into the AI Assistant channel with no setup from you.
Here is the catch. The native channel only names ChatGPT, Gemini, and Claude, and only when the visit carries a clean referrer. Perplexity and others are not confirmed in it. It does not backfill history, so your trend line starts flat at zero on the day it went live for your property. Treat it as a floor, not a full count. The rest of this guide covers how to recover what it misses.
Quick Check: Find AI Traffic GA4 Already Records
Start by seeing what you already have, before you build anything. Open GA4, go to Reports, then Acquisition, then Traffic Acquisition. Switch the primary dimension to Session default channel group. If an “AI Assistant” row shows up in the table, the native channel has reached your property and is already recording sessions.
If you don’t see it yet, the rollout is staged and may land in a few days. Broad availability arrived around early June 2026. Either way, bookmark this report. It is your baseline. Whatever number sits in that AI Assistant row is the minimum amount of AI traffic you’re getting, and the custom setup in the next sections shows how much more is hiding elsewhere.
Build an AI Traffic Exploration Report in GA4
Build an Exploration when you want to dig into AI sessions without changing your reports. An Exploration is a sandbox. It lets you filter and slice traffic on the fly, which is the fastest way to confirm what’s arriving from AI before you commit to a permanent channel group.
Name the Report and Add Dimensions and Metrics
Open Explore, start a blank exploration, and name it something you’ll recognize later, like “AI traffic.” Add the dimensions that matter here: session source, session medium, and landing page. For metrics, pull in sessions, engaged sessions, conversions, and any case-related conversion you’ve set up, such as form submissions or calls.
Filter for AI Referrals with Session Source Regex
Add a filter on session source set to “matches regex,” then paste a pattern covering the AI domains. Something like chatgpt|openai|perplexity|claude|gemini|copilot works as a start. GA4 uses RE2 regex syntax and is case-sensitive, so test your pattern against live rows before you trust it. Regex is fast and flexible, but it can overmatch and pull in unrelated traffic, so check what it catches.
Visualize and Analyze the Results
Set the visualization to a table, with your AI sources as rows and conversions alongside sessions. Now you can see which assistant sends the most traffic and, more importantly, which sends traffic that converts. A channel with few sessions but a high consultation rate is worth more than one with volume and no signed cases.
Save and Share Your Report
Save the exploration so it persists, and share it with anyone on your team who reviews performance. Explorations are per-user by default, so use the share option if your intake or leadership team needs access. This becomes your working view while the permanent channel group, built next, handles standing reports.
Create a Custom AI Traffic Channel Group
Build a custom channel group when you want AI traffic to show up as its own line in every standard report, not just in a sandbox. The native AI Assistant channel is a fine start, but you still build your own for two reasons: it recovers history the native channel left behind, and it covers assistants Google’s list misses.
Add the AI Platform Regex and Reorder Above Referral
Go to Admin, then Data display, then Channel groups, and create a new channel group. Add a channel named “AI Search,” set its condition to source matches regex, and paste your AI domain pattern. Then do the one step the whole setup depends on: drag the AI Search channel above Referral. GA4 evaluates rules top-down, so if Referral sits first, it claims your AI sessions before your rule ever fires.
Use these referral domains in your pattern:
| Assistant | Common referral domains | Notes |
| ChatGPT | chatgpt.com, chat.openai.com | Highest volume for most sites. |
| Perplexity | perplexity.ai, www.perplexity.ai | Not confirmed in GA4’s native channel. |
| Claude | claude.ai | Covered by the native channel when a referrer is present. |
| Google Gemini | gemini.google.com, bard.google.com | bard.google.com is legacy. |
| Copilot | copilot.microsoft.com | Legacy bing.com/chat traffic needs a path-level filter, since GA4’s session source is hostname-only. |
| Grok | grok.com, x.ai | xAI’s chat product. x.ai also hosts API docs and marketing, so it can catch non-chat visits; use grok.com alone if you want tighter precision. |
| Meta AI | meta.ai | Meta’s standalone AI surface. Embedded Meta AI in WhatsApp and Instagram usually passes no referrer, so those visits land in Direct. |
View the AI Channel in Traffic Acquisition Reports
Check your work in the Traffic Acquisition report. Open Reports, then Acquisition, then Traffic Acquisition, and switch the channel group dimension from the default to your custom one. A custom group lives alongside the default; it does not replace it. Your AI Search channel should now appear next to Organic Search, Direct, and Referral, with sessions and conversions attached.
Analyze AI Visitor Quality: Landing Pages, Engagement and Conversions
Volume tells you nothing until you know whether these visitors become clients. Review your AI Search channel by landing page, and line it up with engaged sessions and conversions. Which pages are AI assistants sending people to? Are those visitors booking consultations, or bouncing?
This is where the signed-case question gets answered. Watch whether AI-referred visitors convert at a different rate than your other channels, and which practice-area pages pull the traffic that books consults. That tells you what content the assistants trust, and where to build more. Keep the context in mind: these are still small numbers compared to search, so judge them by conversion quality, not raw volume.
Estimate the Hidden AI Traffic in Direct
A large share of your real AI traffic is sitting in “Direct,” and no setup fully recovers it. Most assistants either send no referrer or use inconsistent hostnames, so those sessions never match your regex and land in Direct instead. In-app browsers, mobile apps, and copy-pasted links make it worse.
You can estimate the gap, not erase it. Watch for Direct traffic that behaves like AI traffic: landing deep on practice-area pages rather than your homepage, arriving after a jump in AI impressions in Search Console, converting at rates unlike your usual Direct. None of this is exact. Treat your AI numbers as a floor, and tell anyone who asks, rather than reporting a precise figure you cannot defend.
Connect Google Search Console AI Mode Data and Set Up Alerts
Search Console shows where you appear in AI search, but it will not tell you who clicked. Google launched its Generative AI performance reports on June 3, 2026, as a dedicated view inside the Performance tab. It covers AI Overviews and AI Mode, with data starting mid-May 2026 and no backfill before that.
Read it for what it is: impressions, not traffic. The report shows impressions, pages, countries, devices, and dates, and nothing about clicks or CTR. Neither the API nor the BigQuery export exposes this data at query level. That gap matters because AI surfaces are where clicks are lost: when an AI summary appears on a Google result, people click a traditional link 8% of the time, versus 15% when there’s no summary, and only 1% click a source inside the summary. So use Search Console to see which pages AI surfaces trust, then cross-reference against the real sessions and conversions in GA4. Set up alerts on sharp impression changes, and pair every impression trend with the traffic data, never on its own.
Automating AI Traffic Tracking and Keeping Your Setup Current
This setup goes stale fast, so plan to maintain it. New assistants launch, referral domains change, and Google updates its native classifier without much notice. Keep a running list of AI referral domains and update your regex whenever a new platform shows up in your data.
One maintenance task people miss: audit your other custom channel groups, not just the AI one. A referral rule you wrote last year might now be catching sessions that Google reclassifies as AI Assistant, which skews both channels. Review the setup each quarter. For top-level reporting, lean on the native channel; for depth and history, lean on your custom group. Run both.
See What AI Search Sends Your Law Firm: Get a Free GA4 AI-Traffic Tracking Review
Most firms are flying blind on AI traffic, and the ones who fix it first will know which channels sign cases before their competitors do. We’ll audit your GA4 setup, show you what AI search is actually sending you, and tell you straight whether it’s converting into signed cases or just padding your session count.
Here’s how we work. The methods we use come from primary research across 1,300+ firms and 15,000+ sites, and our founder, Jason Hennessey, wrote a book on law firm SEO. We’ve helped generate $2B+ in leads and verdicts for clients, and we report in signed cases and revenue, not rankings or traffic. There are no gotcha contracts. You own your GA4 property, your data, and everything we build, even if you walk. If an agency has burned you with vanity metrics, this is the opposite. Ask us to show you the cases.
Frequently Asked Questions
Does GA4 track AI traffic automatically?
Partly, since May 13, 2026. GA4 now has a native “AI Assistant” channel that catches traffic from ChatGPT, Gemini, and Claude when the visit carries a clean referrer. It needs no setup. But it misses assistants not on Google’s list, sessions that arrive without a referrer, and doesn’t backfill history. So “automatically” is true for a slice of your AI traffic, not all of it.
Can I separate ChatGPT, Claude and Gemini traffic from each other in GA4?
Yes, with a custom channel group or an exploration using a source regex. Build a pattern that names each assistant’s referral domains, and GA4 will split them into separate rows you can compare. This is the only way to see, say, whether Perplexity visitors convert better than ChatGPT visitors, since the native channel lumps its covered assistants together and leaves others out entirely.
Why does so much AI-referred traffic show up as Direct in GA4?
Because many assistants send no referrer, or an inconsistent one. When a session arrives with no referrer data GA4 can read, it defaults to Direct. In-app browsers, mobile apps, and copy-pasted links add to the pile. As a result, a real portion of your AI traffic gets buried in Direct, and no configuration pulls it all back out.
Do I need a third-party tool to track AI traffic in GA4?
Not for the basics. The native channel plus a custom channel group and an exploration will cover most of what you need, for free. Third-party tools mainly help at the edges: recovering referrer-less sessions, tracking visibility inside AI answers, or stitching AI traffic to signed cases across your intake system. Start with the GA4 setup here, then add a tool only if a specific gap justifies it.