---
title: "What an AI citation candidate actually means"
url: https://hostmy.blog/docs/rankready/what-citation-candidate-means/
date: 2026-08-24
modified: 2026-08-25
lang: en
author: "Aditya Sharma"
description: "The honest reading of AI crawler data: a hit proves your page was fetched, not that it was quoted in an answer."
word_count: 861
---

# What an AI citation candidate actually means

This page exists to stop you drawing a conclusion the data does not support. RankReady can show you that AI crawlers fetched your pages. It cannot show you that an AI product quoted them, and neither can anything else running on your own server. The difference matters, and it is worth being blunt about it.

## The one sentence version

**A crawler hit proves your page was fetched. It does not prove your page was quoted.**

That is the whole limit. Everything below explains why, and what to do with the data you do have.

## What a candidate means

"Candidate" is the accurate word, and it is doing real work. A page that has been fetched by an AI crawler is a page that is eligible to be used: it was reachable, it was allowed by your rules, it was requested by a named agent, and your server returned it successfully. That is a precondition for ever being used in an answer.

A precondition is not an outcome. A page can be fetched a thousand times and never appear in an answer. A page can be fetched once and be used. From your side of the connection, those two look identical.

## Why your server cannot know

Everything after the response leaves your infrastructure. Whether a model used your page as context, whether an answer drew on it, whether it was named as a source, whether a user saw that name: all of that happens inside a product you do not run, with no callback to you. There is no header, log line or API on your side that reports it.

Any WordPress plugin, including this one, sees exactly one thing: requests arriving and responses going out. Any claim beyond that is inference dressed as measurement.

## What you can state as fact

- Named AI crawlers requested these URLs.
- Your server returned these status codes.
- Your `robots.txt` declares these permissions, and you can read the file to confirm it.
- Your pages publish this structured data, verifiable in Google's Rich Results Test.
- Your machine-readable files exist and return the content you expect at `/llms.txt`, `/llms-full.txt`, `/{post-slug}.md` and `/okf/`.
- People arrived on your site from an AI product, where the visit carried a referrer.

## What you cannot state as fact

- That an assistant cited you.
- How many times you were cited.
- That being fetched more often means being cited more often.
- That a change you made caused an increase in anything downstream.
- Any share, percentage or multiplier about citations. There is no source for those numbers on your site.

## Where referrals sit

AI referral tracking is the strongest signal available to you, and it is still not proof of citation. A referral means a person clicked a link to your site from an AI product's interface. That is a human action you can observe, which puts it well above a crawler hit.

Its limits are real too. A visit only carries attribution when the referrer is sent, and it is not always sent. Referrals count clicks, not mentions: an answer that names your site without linking it produces no referral at all. So referrals undercount by an amount you cannot measure.

The honest ordering, weakest to strongest:

| Signal | What it proves |
| ------ | -------------- |
| Endpoint or schema exists | Your content is published in a parseable form |
| Crawler hit | An agent fetched the page |
| Repeated crawler hits | An agent keeps fetching the page |
| AI referral | A person clicked through from that product |

Nothing on that table is a citation count. There is no fifth row.

## Why the discipline pays

Two reasons, both practical.

First, decisions. If you treat crawler volume as a success metric, you will optimise for being fetched more often, which is not the same as being useful. You will also read normal variation as a result.

Second, credibility. Reporting "GPTBot fetched 40 of our pages this month, and 12 visits arrived from AI products" is a statement you can defend line by line. Reporting "we were cited 40 times" is a statement you cannot support, and one informed question will expose that.

![The RankReady Insights tab in wp-admin showing crawler and referral data](https://hostmy.blog/wp-content/uploads/2026/08/rr-insights-v2-scaled.png)Insights reports what your server actually saw: requests, and referred visits.

## What to do with candidates

Treat the list as a working queue, not a scoreboard.

- **Check coverage.** Are the pages that matter to you being fetched at all? If your most important page never appears, that is a concrete gap worth investigating.
- **Check the response.** Errors and redirects served to crawlers are fixable and unambiguous.
- **Check the content.** Open the fetched URL's [Markdown version](https://hostmy.blog/docs/rankready/wordpress-markdown-for-ai/) and read it. If the page reads as thin or unclear there, it will read that way to anything parsing it.
- **Check your rules.** Confirm your `robots.txt` block says what you intended.

All four are real improvements you can verify. None of them requires believing anything about what happened inside someone else's product.

## The standard to hold

Say "fetched" when you mean fetched. Say "referred" when you mean referred. Do not say "cited" unless you are looking at an answer with your name in it, which is something you find by asking the product a question yourself, not something your server can report. Keeping those words separate is the whole discipline, and it costs nothing.