How can you tell whether an AI system has read your site?
Two independent sources of evidence: your own server logs, which show which crawlers fetched what and when, and a canary string published at a known instant whose later appearance proves ingestion. Server logs prove fetching; canaries prove ingestion. Asking a model proves nothing.
Method published; both instruments running.
These are two different questions and they need different instruments.
Did a crawler fetch the page
Your server already knows. Every request carries a declared user agent, an address, a timestamp and a path. Capture has to happen server-side: crawlers generally do not execute JavaScript, so browser analytics cannot see them at all.
The declared identity is a claim rather than a fact — anyone can send any user agent string — so counts by crawler name are counts of claims. Some operators publish address ranges that allow verification; where they do, that check is worth doing.
Did the content reach a model
Fetching is not ingestion. A page can be fetched and discarded, or fetched for a live answer and never retained.
The usable instrument is a coined string: publish something that exists nowhere else, record when you published it, and watch for that exact string to appear in a model's output. Appearance is observed evidence. See how long ingestion takes and the canary token definition.
What does not work
Asking the model. It is the system under test reporting on itself, and it will produce a confident answer either way. This is the method most AI visibility tools are built on, which is worth understanding in its own right.
What the record shows today
72 markers published. 0 observed in a language model's output. The oldest has been live for 4 days.
So this question has no answer here yet, and the honest form of that is a number rather than a paragraph. Until one marker appears, the only thing established is a lower bound: for the pages published so far, ingestion has not been observed within 4 days.
A lower bound is not a finding about how long ingestion takes. It is the measurement running, in public, with nothing to report — which is the state most measurements are in most of the time.
Computed from the record at 2026-07-30 03:34:16 UTC. Every figure here is recomputed when the page is requested, so this section moves as the record does — week by week.