No setting anywhere grants a citation. What a small site controls is the shape of its passages and its own server logs. What makes a paragraph cheap to quote, how to tell whether assistants are fetching your pages, and why a cited page can still show zero referrals.
You cannot make an assistant cite you, and there is no setting anywhere that grants it. What a small site controls is two things. The first is the shape of its passages: a paragraph that answers one question completely, restates its own subject and carries the specifics is cheap for a model to lift, and a paragraph that depends on the three above it is not. The second is measurement: an assistant that fetches your page leaves a request on your own server, whether or not anybody clicks through afterwards.
A retrieval-based answer is assembled from pages fetched to answer one specific question, and the unit being selected is a passage rather than a whole site. Domain size is not irrelevant — nothing here claims it is — but a small site writing a precise, self-contained answer to a narrow question is competing on something it can actually change, while competing on domain authority is competing on a number that moves over years. The narrower and more specific the question, the smaller the field of pages that answer it properly.
None of this is a trick aimed at models. It is the same discipline that makes a page useful to a person who arrived from a search result and will read exactly one section of it. The overlap is the point: writing for the reader who lands mid-page produces the passage an assistant can quote without repair.
Two measurements answer two different questions, and confusing them produces a wrong conclusion in both directions. Retrieval fetches say whether AI agents are pulling your pages. Referral sessions say whether anyone followed a link out of an answer. A site can have plenty of the first and none of the second, and that combination is ordinary rather than broken.
There is no markup that gets a page into an AI Overview. Eligibility is ordinary indexing plus permission to show a snippet, which makes the question what to write rather than what to add.
A page can be named in an AI answer without ranking for the question, but the case is narrow: the assistant answered from its own weights instead of running a search. Retrieval produces a link. Memory produces a name with nothing behind it.
Market4 turns one release note into a changelog page, a blog post, a mail-out and a week of social posts — and then tells you which of them brought anyone back.
| Measurement | Where it comes from | What it cannot tell you |
|---|---|---|
| Retrieval fetches by AI agents | Your own server's request log for the page | Whether a fetch became a citation |
| Referral sessions from assistants | Analytics, isolated by assistant hostname | Anything about a citation nobody clicked |
| Asking an assistant the question yourself | One session, by hand | Anything repeatable, because answers differ between sessions |
The referral half has a wrinkle worth knowing before you go looking for it. Analytics tools file assistant traffic under the ordinary referral channel, alongside directories and newsletters, so the number that proves an AI citation produced a visitor is invisible in a stock channel breakdown. It has to be pulled out by hostname, and it should always be reported next to total sessions, because a bare count of visits means nothing without a denominator.
Every AI crawler identity is claimed, never verified. A User-Agent is a string the client chooses, anyone can send one that says GPTBot, and unless something performs a reverse lookup on the connecting address, the name in the log is an assertion rather than a fact. Say "claimed" when reporting these numbers. The connecting address is the part that was actually observed.
The second caution is to read the kind of agent rather than its name. An agent that retrieves a page to answer someone's question right now can cite that page, so blocking it is self-inflicted invisibility. An agent collecting text for a model that ships months later gives no link back, so blocking it costs no citations. Some operators document a crawler as doing both, and where that is the case the honest classification is unclear rather than a guess.
Assistants pass a referrer only when somebody follows a link out of the answer. A page that was retrieved, quoted and named in the answer, and then read as part of that answer, leaves no trace in analytics whatsoever. So an empty AI referral report is not evidence that nothing was cited, and a crawler log full of retrieval fetches beside a referral report of nothing is the normal shape of this, not a broken integration.
No. There is no submission form, no setting and no header that grants a citation, and any tool promising one is selling something else. What is under your control is whether the page contains a passage that answers a specific question completely and correctly, and whether the page is reachable and fast for the agent that fetches it. Everything past that is somebody else's ranking decision.
There is no published guarantee that any assistant reads it, so treat it as cheap and unproven rather than as a lever. Writing one costs an hour and breaks nothing. Expecting it to change which pages get quoted is a different matter, and if you add one, keep measuring retrieval fetches the same way you did before, so you can tell whether anything actually changed.
Decide per kind, not per name. Blocking an agent that fetches pages to answer live questions removes you from the answers it produces, which is the opposite of what most small sites want. Blocking an agent that collects training text costs no citations, because that route gives no link back either way. Where an operator documents a crawler as doing both, you are choosing between the two effects, and that is a judgement rather than a default.
Most likely because nobody clicked. Assistants pass a referrer only when a person follows a link out of the answer, so pages that were cited and read inside the answer leave no session behind. Check your server-side crawler data before concluding you are invisible: retrieval fetches with zero referrals is a common and unremarkable pattern.
Each assistant arrives under its own name, obeys the rules written for that name, and takes away whatever the page contained at fetch time. Your own log is the only place all three can be compared.