Visitors from ChatGPT: how to spot them, measure them and decide if they matter

To find a tool, compare two options or look up a concept, most people used to open a search engine. Now a growing share ask ChatGPT, Perplexity, Claude or Gemini first. The AI includes a few links in its answer, the user clicks one, and they arrive on your site.

This traffic is growing fast, yet in many teams' reports it either doesn't exist or is buried among referrals where nobody can see it.

Where it hides in your reports

When someone clicks through from an AI product, the browser usually passes the referring domain — chatgpt.com, perplexity.ai and so on. If your analytics doesn't recognise those domains specifically, they're treated as ordinary websites and filed under "referral" or "other", mixed in with forums and partner links.

Some AI products also open links inside their apps without a referrer, and those visits count as direct.

We give AI products their own channel: AI assistants. In the Sources report it sits alongside organic search, social and paid ads, with its own visitors, sessions and conversions.

How AI visitors differ

From what we've seen, visitors from AI assistants tend to share a few traits (the proportions vary a lot by site):

• Clearer intent. They've already asked and compared inside the AI, and often arrive to confirm something specific — a price, a feature, how setup works.
• Concentrated landing pages. AI tends to cite a handful of pages: comparisons, pricing, an article that explains one concept well, help documentation.
• Possibly higher conversion. Because intent is clearer, on some B2B and tool sites AI visitors sign up at a higher rate than organic search visitors. That's not a law — check your own data.

So rather than AI's share of total traffic (usually still small), look at its conversion rate and the pages it cites.

How to measure it

1. Volume and trend

Open the AI sources tab in the Sources report to see each product separately — how many visits ChatGPT, Perplexity, Claude, Gemini and Copilot each sent. Watch the monthly trend to see whether it's growing.

2. Conversion

Add your key goals — sign-up, demo request, purchase — and compare AI's conversion rate with other channels. If it's clearly higher, take it seriously even at low volume.

3. Cited pages

AI visibility has a "cited pages" table: the landing pages AI assistants sent visits to. It's valuable because it shows which of your content AI considers worth recommending.

A common surprise: the most-cited page is rarely the home page you spend the most effort promoting. More often it's a help article that explains one problem particularly clearly, or an honest comparison table.

Reading and returning are two different things

One point causes confusion. AI companies send crawlers to read your site, for training or for answering questions in real time. How many pages the crawlers read and how many people the AI sends back are separate numbers:

• "AI sources" in the Sources report counts people who clicked a link to your site.
• The Bots report counts crawlers downloading your pages.

They're often out of proportion. One AI company's crawler reading thousands of your pages in a month while sending almost nobody is common. Only by looking at both do you learn whether being read is paying off. We go into this in How many of your pages AI crawlers read.

Is it worth investing in?

Our view: look at the data first. A few tests:

• Have AI visitors grown steadily over the last three months?
• Is their conversion rate at least as good as organic search?
• Are the cited pages related to your core business?

Three yeses and it's worth the effort: keep the most-cited pages in good shape, keep their information accurate (especially pricing and features), and write clear, direct answers to the questions people actually ask.

If volume is still small and conversion average, keep an eye on it and check the trend once a month. There's no need to overhaul your content strategy for a channel that's still tiny.

An easily missed benefit

Visitors from AI have often already read a description of you inside the AI. If that description is wrong — an old price, a feature that's gone — they arrive confused and may leave straight away.

So the cited pages table has a second use: see which pages AI is citing and make sure what's on them is current. It may be the cheapest, fastest-acting thing on your list.

Related articles

AI search & GEO·Oct 9, 2026·4 min read

How many of your pages AI crawlers read — and why it doesn't match the visitors they send

One AI company read 8,000 of your pages last month and sent 40 people back. Is that normal? How to read the ratio between reading and returning, and what you can do about it.

Read article

Product notes·Oct 16, 2026·3 min read

Why bot traffic gets its own report

By default we take identified bots out of your human reports rather than dropping them or mixing them in. The three reasons behind that, and what it means for your numbers.

Read article

Playbooks·Oct 14, 2026·5 min read

Checkout drop-off: a seven-step investigation from data to fix

Plenty of add-to-carts, not many payments. A checklist to work through in order — rule out technical problems first, then find which step, which visitors and what cause.

Read article