How Can You Tell If AI Assistants Are Sending You Visitors?

A customer asks ChatGPT which local bakery does vegan birthday cakes. The answer names you and links to your site. She taps the link and orders. In your analytics, that sale may show up as direct traffic, with no sign an AI assistant was involved.

Most site owners we talk to start AEO work and then ask one fair question. Is any of this sending people to us? It's a reasonable thing to want to know before spending more time on it.

Traditional search gave us clean tools for this. Search Console shows queries and clicks. AI assistants mostly don't. Bing Webmaster Tools now has an AI Performance report that shows how often your pages are cited in Microsoft Copilot and Bing's AI summaries, but it doesn't show clicks, and most other assistants offer nothing like it. So we have to piece the picture together from the analytics we already have. If you're new to the difference between the two disciplines, our SEO vs AEO comparison covers it.

Where does AI traffic show up in your analytics?

When someone clicks a link in a browser, the browser usually tells your site which page they came from. That is the referrer. Analytics tools read it and file the visit under a source.

Referral traffic

When a person clicks a cited link inside an AI assistant's web app, the referrer is often the assistant's own domain. In Google Analytics 4 you'll find these under Reports, then Acquisition, then Traffic acquisition. Switch the table to the Session source dimension to see each domain. Other analytics tools have a similar referrers or sources report.

GA4 also has a built-in AI Assistant channel in its default channel group. Google's help docs say it covers sources like ChatGPT, Gemini, DeepSeek, Copilot and Grok, and that a visit is placed there when its referrer matches Google's list of AI assistants. Claude and Perplexity aren't among Google's examples, so check where they land in your own reports. Clicks from Google's AI Overviews and AI Mode are not in this channel. GA4 counts them as Organic Search.

Domains worth looking for include:

• chatgpt.com (and the older chat.openai.com)
• perplexity.ai
• claude.ai
• gemini.google.com
• copilot.microsoft.com

Treat this as a starting list. New assistants appear and domains change, so scan your full referral list now and then for anything new.

Tagged links

Some assistants add a tracking tag to the links they show. OpenAI's publisher FAQ says ChatGPT automatically adds utm_source=chatgpt.com to the links people click. That tag matters, because many ChatGPT links don't pass a referrer. With the tag, GA4 can still show chatgpt.com as the session source. Without a referrer, the visit may miss both the AI Assistant and Referral channels and end up as Unassigned, so check that row too.

The tag isn't guaranteed either. Independent tracking found it missing from most ChatGPT links for about a week in mid-October 2026 before it came back. If your ChatGPT numbers dipped then while direct traffic rose, that gap is the likely reason.

Direct traffic

Some AI visits carry no referrer. This can happen when someone copies a link out of an answer and pastes it, when a desktop or mobile app opens the link in a way that strips the referrer, or when the link itself is set not to pass one. If there's no tag either, the visit lands in direct traffic with nothing to mark it. There is no reliable way to pull them back out.

How do you set up a simple AI traffic view?

Looking at five domains scattered across a long referral list gets old fast. A single grouped view makes the trend visible in a few seconds.

Create a custom channel group

In GA4, go to Admin, then Data display, then Channel groups, and create a new channel group. GA4 starts you with a copy of the default group. Add a channel named something like "AI assistants" with a condition on Source that matches the domains above. Then reorder the list so your channel sits above Referral and above the built-in AI Assistant channel, because GA4 puts each visit in the first channel it matches.

Google's help docs say custom channel groups can be applied to your reports retroactively, so you can look back at earlier months, not just forward. Standard GA4 properties allow two custom channel groups, so it's worth planning before you use both.

If you use a different analytics tool, look for a way to save a filtered view or segment on referrer domain. The idea is the same.

Use a regular expression if your tool allows it

A single pattern saves you adding domains one at a time. In GA4, choose the "partially matches regex" condition and use a pattern like this. The plain "matches regex" option has to match the whole value, so the partial version is more forgiving:

chatgpt\.com|chat\.openai\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|copilot\.microsoft\.com

Test it against your own referral list before trusting it. A pattern that's slightly off will quietly miss visits.

What should you look at once you can see it?

A raw visit count is the least useful number here. The pages and the behavior behind it tell you more.

Landing pages

In the Traffic acquisition report, add "Landing page + query string" as a secondary dimension and look at your AI assistants channel. This shows which of your pages assistants are sending people to. Often it's a handful of pages that answer a specific question clearly. Those pages are your best evidence of what works on your site, and a good model for the next page you write.

Engagement and key events

Compare engaged sessions and key events (the actions you've marked as important, such as a form submission) for AI visitors against your other channels. Someone who clicked through from an AI answer has often already read a summary of you. They may arrive further along in their decision. Or they may bounce because the answer gave them enough. Your own numbers will tell you which, and we'd rather you trust them than any general claim.

Trend over time

Look at the line month by month, not day by day. Volumes on small sites are often low, and a single week can swing for reasons that have nothing to do with your content. Note the dates when you made AEO changes so you can line them up against the trend later.

What can't analytics tell you?

This part matters, because it's easy to read too much into a small number.

• Citations without clicks. An assistant can name your business and the person never clicks. They might call you, search your name later or walk in. Analytics sees none of that as AI-driven.
• Hidden direct traffic. As covered above, some real AI visits are filed as direct. Your AI channel is a minimum.
• Google's own AI answers. Clicks from AI Overviews and AI Mode are counted as Organic Search in GA4, so they won't appear in your AI channel.
• Branded search after an AI answer. A person reads about you in an assistant, then searches your name on Google. That visit shows up as organic search, and the AI answer gets no credit.
• What the assistant actually said. Referral data tells you someone came from chatgpt.com. It doesn't tell you the question they asked or how you were described.

To fill that last gap, ask the assistants yourself. Pick the five or ten questions a customer would ask before choosing someone like you. Ask them in two or three assistants every month and write down whether you're mentioned and what's said. It's manual, and the answers vary from one session to the next, but it shows you something analytics can't.

You can also simply ask new customers how they found you, and include "an AI assistant like ChatGPT" as an option. It's low tech, and it catches the visits your tools miss.

Is AI traffic worth tracking if the numbers are small?

Yes, for two reasons. The first is that the trend matters more than today's total. A channel that's small now but growing is worth knowing about early. The second is that the landing pages tell you which content gets picked up, and that guides your next piece of work.

Small numbers also keep expectations honest. AEO improves your chances of being read and cited. It doesn't guarantee either, and a measurement setup that admits its blind spots helps you make calm decisions instead of chasing every swing.

Frequently Asked Questions

Why does ChatGPT traffic show up as direct traffic sometimes?

Some clicks arrive with no referrer and no tracking tag. This happens when a link is copied and pasted, when an app opens it in a way that drops the referrer, or when a link is set not to pass a referrer and the tag is missing. Your analytics then can't tell where the visit came from, so it files it as direct.

Which AI assistant domains should I track?

Start with chatgpt.com, chat.openai.com, perplexity.ai, claude.ai, gemini.google.com and copilot.microsoft.com. Then scan your full referral list every month or so. New assistants launch and domains change, so your own data is the best guide to what to add.

Can I see which questions people asked the AI before visiting my site?

Not from your analytics. Referral data shows the source domain and the page the visitor landed on. It doesn't include the question or the assistant's answer. Bing Webmaster Tools' AI Performance report shows grounding queries, the search phrases Copilot used to find pages it cited, but those aren't the person's own words. To learn how you're described, ask the assistants your customers' common questions yourself and record the results.

How long after AEO changes should I expect AI referral traffic to change?

There's no fixed timeline. Assistants recrawl and update on their own schedules, and some don't publish them. Record the date of each change and compare month-over-month trends, rather than expecting a jump within days.

Does a low AI referral count mean my AEO work isn't working?

Not necessarily. Your count misses citations without clicks, visits filed as direct, and people who searched your name after reading an AI answer. Look at the trend, the landing pages and what assistants say when you ask them directly.

Want to check your own site?

Run the free AEO audit to see your score and get ready-to-paste fixes. Using WordPress? AEO Bob Lite is free on WordPress.org. Rather have it done for you? Our done-for-you fixes start at $497.