You may already have clients from ChatGPT and no way to see them
Assistant traffic lands in direct because the referrer is often stripped. Three reports and one segment separate it out, and the share of enquiries is bigger than the share of sessions.
Traffic from AI assistants is real, small and badly measured. In most analytics setups it is scattered across direct, referral and nothing at all, which means a business can be getting clients from ChatGPT and not know it. Fixing the measurement is a configuration job, not an SEO job.
Why it hides in direct
An assistant answer is often read inside an app rather than a browser tab, and when the visitor taps a citation the referrer may be stripped or absent. The session then lands in your reports as direct traffic, indistinguishable from somebody typing your domain. Some assistants do pass a referrer, which is why the honest picture needs both a referral view and a landing-page view rather than one number.
Build one segment and one report
- A referral group that collects the assistant hostnames you care about, kept in one place so you can add the next one without rebuilding anything.
- A landing-page report filtered to that group, because which page the assistant cited is the actionable fact, not the total.
- A conversion view on the same segment. Assistant traffic is usually low volume and high intent, and judging it on sessions alone understates it badly.
- Server log review, which is the only place you see the crawlers themselves rather than the humans they send afterwards.
Do not build a strategy on the number yet
For most of the service businesses we work with, assistant referrals are a low single-digit share of sessions and a disproportionate share of qualified enquiries. That is worth measuring and not worth reorganising a content programme around, because the pages that get cited are the same pages that rank: specific, answerable and written by somebody who does the work. The measurement exists so that when the share grows you will know, and so that you can tell which page earned it.
What the crawlers actually fetch, which is a separate question from what the humans do afterwards, is in what an AI crawler receives when it requests your page.
