All guidesReports & oversight
Read your website traffic
Where visitors to your public site come from, and whether the AI engines are reading it — two different things this page keeps apart on purpose.
- Steps
- 6 · About 5 min
- Who can do this
- Owner
Your public website and these guides are the two things a stranger can find before they ever talk to you. This page says whether anybody is finding them, and how.
Website traffic in the sidebar, under Control & data. Owners and administrators only.
It is recorded on your own server and stored in your own database. No cookies, no visitor identity, nothing sent to any other company. That is not only a privacy position — it is the reason the page can answer the question in step 3, which no ordinary analytics tool can answer at all.
Two kinds of "AI traffic", and why they are not the same
This is the thing to understand before reading anything else on the page, because getting it wrong turns a good number into a misleading one.
Someone arrived from an assistant. A person asked ChatGPT or Perplexity a question, it mentioned you, they clicked. That is a visit. It is a potential customer, and it appears under From AI assistants.
An assistant's robot read your site. No person was involved. A crawler called GPTBot or ClaudeBot fetched your pages so its model can mention you to somebody later. That is visibility, not a visit, and it appears in its own section — Are the AI engines reading us?
The second almost always comes first, often by weeks. A site with hundreds of crawler fetches and no assistant referrals yet is a site that is being read but not yet recommended — which is normal, and is progress.
The page never adds them together. Neither should you.
How people arrived
Four buckets worth knowing:
Search engines — the classic SEO result. Google, Bing and the rest.
AI assistants — the AEO result, broken out by which assistant.
Other sites — somebody linked to you.
Direct or unknown — and this one needs a warning.
"Direct" does not mean "typed the address". Browsers hide where a visitor came from on many kinds of link, and some assistants send nothing at all. Every one of those lands in Direct. So treat it as unknown, and read it as a floor on your other numbers rather than as people who knew your name.
Are the AI engines reading us?
The section that exists because nothing else can show it.
Crawlers do not run JavaScript. Every browser-based analytics tool — Google Analytics included — is therefore blind to them: it can only count visitors whose browser executed a script, and a robot never does. Recording on the server is the only way to see this at all.
Two lists:
Which crawlers came. GPTBot is OpenAI, ClaudeBot is Anthropic, PerplexityBot is Perplexity. Googlebot and Bingbot are the ordinary search crawlers and are counted here too, so you can compare.
What they fetched. This is the useful one. If llms-full.txt and your
guides appear in this list, the material written for answer engines is being
read. If nothing appears for weeks after launch, the usual cause is that the
sitemaps have not been submitted in Search Console yet.
One honest limit: a crawler's name is self-declared. Anything can call itself GPTBot. The figures are reliable enough to act on and not evidence for an argument.
What the rest of the page tells you
Most visited pages — which of your pages actually does the work of bringing people in. Usually not the home page.
Which language they read — for a site published in Arabic, French and English, this is the one that says whether all three were worth it.
Where they are — visitors outside Algeria are worth noticing before you decide the product is only for one market.
By day — two bars per day, visits and crawler fetches. A long flat run means nobody has found you yet; that is a marketing problem, not a bug.
It takes days, not minutes
Recording starts the moment the site is deployed, but real traffic does not. Search engines take days to re-crawl, answer engines take longer, and neither does it because you are watching. If the page is empty on the day you launch, that is the expected result rather than a fault.
Ask the assistant instead
If the AI assistant is switched on, you can ask it directly — "is anyone finding us through ChatGPT?", "is our SEO working?", "is AI reading our site?" — and it reads these same figures, keeps visits and crawler fetches apart the same way, and refuses the same way if you do not have the rights.