How to track visits from ChatGPT and other AI assistants

Updated 25 September 2026 · 6 min read

Why is AI-assistant traffic hard to see?

AI assistants such as ChatGPT, Perplexity, Microsoft Copilot, Gemini and Claude now include links to sources in many answers. When someone clicks one, the visit can reach your site in several ways:

• With a referrer. The browser tells your site which domain sent the visitor, for example chatgpt.com.
• With a UTM tag. Some assistants add a tag such as utm_source=chatgpt.com to the link.
• With nothing. Clicks from apps, or links copied and pasted, often carry no referrer at all.

Most analytics tools have no built-in idea of what an AI assistant is, so the first two cases usually end up in "referral" and the third in "direct". The traffic is there, but it is scattered.

How do you give AI assistants their own channel?

1. List the assistant domains

   Collect the referrer domains AI assistants send traffic from, such as chatgpt.com, perplexity.ai, copilot.microsoft.com, gemini.google.com and claude.ai.

2. Create an AI channel

   Add a channel rule in your analytics that groups those referrers, and any links tagged with an AI source, into one channel.

3. Tag the links you control

   Where you share links in places AI tools read or surface, use consistent UTM parameters so the visits are recognisable.

4. Watch landing pages

   Look at which pages AI visitors land on, because they show which of your content assistants are citing.

5. Track visibility as well as visits

   Check whether your brand appears in AI answers for your key queries, since many answers are read without a click.

6. Review the list regularly

   New assistants and domains appear often, so update your channel rules every few months.

Which referrer domains should you include?

A starting list for a channel rule might include the following. Domains change, so check your own referral report for others.

Be careful with broad rules. A pattern that matches any Google or Microsoft domain will sweep ordinary search traffic into your AI channel and distort both.

What can you learn from AI-assistant visits?

Landing pages are the most useful signal. They show which of your pages assistants are choosing to cite, which often differs from the pages that rank best in traditional search. Compare engagement and conversions for AI visitors with search visitors, but keep sample sizes in mind: for many sites this channel is still small, and a handful of visits can swing the figures.

If you attribute sales and deals across several touches, an AI visit is simply another touch. It may introduce a customer who later returns through search or a newsletter, which a last-click report would miss. See marketing attribution models explained.

What about the answers that never produce a click?

Many people read an AI answer and never visit a source. Visit tracking cannot see that, so it is worth measuring visibility too: for the questions your customers ask, does the assistant mention your brand, and in which languages? This is where generative engine optimisation comes in. Pages that answer one question clearly near the top, state facts accurately and use descriptive headings are easier for assistants to understand and cite.

How does consent apply to AI-assistant traffic?

Exactly as it does to any other channel. Recognising that a visit came from an assistant uses the referrer or the link tag, but joining that visit to later ones and to a conversion still depends on the visitor's analytics consent. See why analytics undercounts after cookie consent.