How to Tell Whether AI Search Is Actually Sending You Customers
If you have noticed a small, unfamiliar spike in referral traffic from chatgpt.com or perplexity.ai and wondered whether it means anything, you are not alone, and the honest answer is that most businesses cannot yet see the full picture. AI search traffic is real, it is growing, and standard analytics setups undercount it badly. Here is how to actually track it, and how to tell whether it is sending you visits or paying customers.
What counts as AI search traffic?
AI search traffic is any visit that started with someone asking an AI assistant or AI-powered search feature a question and then clicking through to your site. In practice that covers a few different sources:
- Direct clicks from AI assistants, when ChatGPT, Perplexity, Claude, Gemini or Copilot cite your page in an answer and the user taps through.
- Google’s AI Overviews and AI Mode, the generative summaries that now sit above or replace the standard results page for many queries.
- Voice and app-based assistants that read out or link to a source.
It is a different signal to being mentioned by an AI engine without a click, which some people call a “zero-click citation.” That builds awareness but leaves no trail in analytics at all. This guide is about the visits that do click through, and how to stop losing sight of them in your reporting.
Why is AI search traffic so hard to track?
Two structural problems sit underneath every AI traffic report.
Missing referrer headers. A large share of clicks from AI apps, especially inside mobile in-app browsers, do not pass a referrer header to your site. Without a referrer, Google Analytics has nothing to attribute the session to and defaults it to Direct. This is the main reason AI search traffic in most GA4 accounts looks tiny: a meaningful chunk of it is hiding inside your Direct traffic total, indistinguishable from someone typing your URL straight into their browser.
Blended reporting in Search Console. Google rolled out dedicated AI Overviews and AI Mode performance reports in Search Console in 2026, which is useful for understanding visibility. Clicks are a different story. At launch, those reports cover impressions, pages, countries and devices, but not clicks, click-through rate, queries or average position, and AI Overview clicks still sit inside the same Web search totals as ordinary blue-link clicks, with no dimension to split them apart.
Neither of these is a setup mistake on your end. They are limitations in how the platforms currently report the data, and every business is dealing with the same blind spot right now.
How does AI search traffic actually show up in GA4?
Once a click does carry a referrer, it typically lands one of two ways:
- The AI Assistant channel. Google Analytics added a dedicated AI Assistant channel to the GA4 default channel group, which recognises referrer headers from recognised assistants, sets the session medium to
ai-assistant, and groups the visit under its own channel automatically. No configuration is required for the platforms it covers. - Plain Referral traffic. Sources that are not (yet) covered by the native channel typically land as ordinary Referral traffic, with the source set to the platform’s domain, for example
chatgpt.com,perplexity.aiorclaude.ai. If you have never gone looking, this traffic is easy to miss inside a long list of referral sources.
Because coverage of the native channel varies by platform and changes as Google expands it, do not assume it has caught everything. Build your own view alongside it.
Setting up GA4 to track AI search traffic properly
A ten-minute setup gets you a much clearer picture:
- In GA4, go to Admin > Data display > Channel groups and create a custom channel group, for example “AI Search 2026”.
- Add a channel rule that matches session source against a regular expression covering the platforms relevant to you, such as
chatgpt\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|copilot\.microsoft\.com|chat\.openai\.com. - Save it, then view Reports > Acquisition > Traffic acquisition and switch the channel group selector at the top of the table to your new group.
- Add a secondary dimension of session source or full referrer so you can see exactly which assistants are sending visits, not just the total.
- Build a simple exploration that shows this channel against key events (form submissions, calls, bookings) so you are looking at outcomes, not just sessions.
Treat every number this setup produces as a floor, not a ceiling. Because of missing referrer headers, actual AI-driven visits are very likely higher than what you can see, in some cases substantially so. Use the trend over time, not the absolute figure, as your real signal.
What should you do with Google Search Console?
Search Console still earns its place in the process, just for a different job. Use the Search results report’s search type filter to check AI Overviews and AI Mode impressions for your key pages. A rising impression count with flat or falling clicks tells you your content is being read and summarised by AI without being visited, which is useful context even though you cannot isolate the clicks themselves. Pair that with your GA4 AI channel view: impressions from Search Console show where you are being surfaced, and referral sessions from GA4 show where people are actually following through.
How do you tell if AI search is sending you real customers, not just visits?
This is the question that actually matters for the business, and it is where a lot of AI-traffic reporting stops short. A spike in sessions from chatgpt.com means nothing on its own. Check the same things you would check for any other channel:
- Engagement quality. Are AI-referred sessions viewing more than one page, or bouncing immediately? Because an assistant has often already summarised your offer before the click, some AI traffic behaves like a highly qualified, short visit rather than a browsing session, which is not automatically a bad sign.
- Goal completions. Connect the AI channel to your actual conversion events, form fills, calls, bookings, in GA4, the same way you would for SEO or paid traffic.
- Assisted paths. Someone may read your brand in an AI answer, search your name later, and convert through a branded search or direct visit days afterwards. Look at branded search volume alongside AI visibility, not just last-click AI sessions, to catch this.
- A consistent baseline. Track the trend weekly rather than reacting to single-day spikes, since AI platforms change how they cite sources often enough that raw numbers can jump around.
If you are already working on your visibility inside AI answers, our guide on what AI Search Optimisation actually involves and our piece on whether AI Overviews are stealing your traffic cover the other half of this: getting cited in the first place. Measurement only matters once there is something worth measuring.
The bottom line
AI search is already sending Australian businesses real visits and, for many, real enquiries, but almost every standard analytics setup understates it. Build a proper AI channel view in GA4, treat Search Console’s AI impressions as visibility signal rather than a click count, and judge the whole thing on conversions, not session totals. Businesses that measure this properly now will be the ones making confident decisions about it in twelve months, while everyone else is still guessing from a Direct traffic report that quietly includes half their AI-driven visits.
Not sure how AI search is currently showing up in your own analytics, or whether your site is even being cited? Book a free strategy call or explore our AI Search and GEO services and we will show you exactly where you stand.
