How to Track AI Search Traffic (ChatGPT, Perplexity) 2026

Open Google Analytics and search for “ChatGPT” in your traffic sources. For most SMB sites, it returns almost nothing — not because AI isn’t sending traffic, but because standard analytics was never built to see it. Learning how to track AI search traffic properly is now as basic as setting up conversion tracking was in 2015, and most businesses still haven’t done it.

Google Analytics 4 classifies traffic by reading a referrer header — the URL of the page a visitor clicked from. Browsers and web apps send that header reliably. AI assistants often don’t. ChatGPT’s in-app browser, Perplexity’s answer citations, and Gemini’s app links frequently strip or truncate the referrer before the click ever reaches your site, so GA4 has nothing to attribute the session to and dumps it into “Direct.”

That’s the core problem with trying to track AI search traffic using out-of-the-box reports: the traffic is real, it’s growing fast, and your dashboard is telling you it doesn’t exist. According to industry referral data, AI-driven sessions grew roughly 700% across 2025, with ChatGPT alone responsible for the majority share of that volume. If your reporting still shows “AI: 0 sessions,” that’s a measurement gap, not a performance one.

This matters commercially, not just academically. If a prospect asks ChatGPT “best [your service] near me,” gets your business named, and clicks through to book a call — and your CRM shows that lead as “Direct / (none)” — you will systematically undercount the channel that’s actually influencing pipeline. Fixing that is the whole point of this guide, and it pairs directly with the work in our piece on AEO vs SEO across 500 sites of real data, which covers why these engines behave differently from classic Google search in the first place.

Not every AI referral looks the same in your reports. Each assistant sends its own domain pattern when the referrer does survive, so the first step in any AI referral traffic analytics setup is knowing exactly what to filter for.

A practical note on volume before you build anything: these sources will not arrive evenly. ChatGPT sends the overwhelming majority of AI referral traffic for most SMB sites right now, so if your channel group shows nothing but ChatGPT for the first month, that’s expected, not broken. Perplexity tends to follow at a distant second, with Gemini, Copilot and Claude contributing smaller, more volatile numbers that spike around specific product launches or model updates rather than growing steadily. Don’t judge the setup as a failure because four of your five rows show near-zero sessions in week one — check back monthly, because the mix shifts as adoption of each assistant changes.

The last row is important: AI Overviews live inside regular Google search results, so they never show up as a separate referrer at all — they’re folded into your normal “Google / organic” line. You can only isolate their effect through query-level trends in Search Console, which is a different measurement problem to the one this guide solves. For that side of the picture, our post on recovering organic clicks lost to AI Overviews is the better companion read.

The fastest way to track ChatGPT traffic in GA4 is a custom channel group, not a saved filter — filters only apply to the report you build them in, while a channel group reclassifies sessions everywhere, including Explorations and any Looker Studio dashboard pulling from the property.

1. In GA4, go to Admin → Data display → Channel groups.
2. Click Create custom channel group and name it “AI Assistants.”
3. Add a new channel rule where Session source contains any of: chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, claude.ai.
4. Drag the “AI Assistants” rule above “Direct” in the evaluation order — GA4 processes rules top-down, and Direct will swallow the session first if it sits higher.
5. Save, then apply the same channel group as a secondary dimension on your acquisition reports going forward.

Even a perfectly built channel group won’t catch every session — a meaningful share of AI referral traffic will still land as Direct because the app-side referrer was stripped before your channel rule ever saw it. The workaround is behavioural, not source-based.

Segment your Direct traffic by landing page, device, and session characteristics, then look for the signature: short, specific landing pages (not your homepage), unusually low bounce on first visit, and sudden step-changes rather than gradual growth. Organic Direct traffic to a homepage grows slowly with brand awareness; AI-driven Direct traffic tends to spike on a single deep page right after that page gets cited in an answer, then holds at a new baseline. A homepage-to-inner-page ratio that shifts noticeably in Direct traffic, with no matching change in your ad spend or PR, is close to a reliable tell that AI search visibility tracking needs a second layer beyond GA4 alone.

Cross-reference against your CRM too. If sales suddenly logs several leads with source “unknown” or “direct visit,” ask on the next call how they found you — “I asked ChatGPT” is now a common enough answer that it’s worth adding as a dropdown option in your lead form.

GA4 can only ever show you traffic that clicks through. It cannot show you the times your business is named inside an AI answer and the user never clicks anything at all — which, per recent AI search share data, is still the majority of AI-answer exposure. That’s a visibility gap, not just an attribution gap, and it needs a different kind of tool: AI-visibility platforms that run your target queries through ChatGPT, Perplexity and Gemini on a schedule and log whether — and how — you were mentioned.

Used together, GA4’s AI Assistants channel group tells you what AI traffic converts, and a visibility tool tells you how often you’re shown but not clicked. Growth100X builds both into one reporting stack for clients, alongside the practical fixes covered in how to get cited by ChatGPT and Perplexity — measurement without a plan to improve the number is just a more precise way of watching it stay flat.

One technical piece worth pairing with your tracking setup: giving AI crawlers a clean, structured map of your site via an llms.txt file. It won’t move your GA4 numbers by itself, but it’s part of the same AI-readability foundation that makes citation — and therefore trackable referral traffic — more likely in the first place.

There’s also a reporting habit worth building alongside the tooling: a monthly one-page summary that puts AI Assistants traffic next to Organic, Paid and Direct in the same table, rather than buried three clicks deep in a GA4 Exploration nobody opens. Once decision-makers can see the AI Assistants row move month over month, tracking stops being an IT afterthought and starts influencing where content and PR effort actually goes — which is the entire reason to measure a channel in the first place. We build exactly this kind of reporting into ongoing engagements, drawing on the same case-study work visible in our client results.