Two questions get asked as if they were one: can AI engines cite my pages, and are they already doing so? The first is a property of your site, and you can inspect it today. The second is an observation about answers you do not control, on a question set you did not choose, at a moment that has already passed. Treating either as a proxy for the other is where most bad AI-visibility dashboards start.
What is the difference between being citable and being cited?
Citable is a state your page is in: a crawler can reach it, a parser can read it, and it contains something an answer could lift - a fact, a quotation, a direct answer to a question someone asks. Cited is an event: a specific engine, answering a specific question, used your page.
The first can be checked before any answer exists. The second cannot be checked at all until the engine answers, and even then the answer is one sample from a system that may answer differently tomorrow. That asymmetry is the whole reason the two measurements are separate products.
Why is readiness upstream of mentions?
A generative answer is assembled in stages, and a page can drop out at any of them. The scanner measures only the first two; a mention measurement observes only the last.
- Retrieval. The engine or its crawler fetches pages it considers candidates. A robots.txt disallow, a firewall rule or a noindex directive ends the chain here.
- Parsing and selection. The engine reads what it got and decides which passages answer the question. Entity markup, heading structure, answer-first paragraphs and extractable facts are what make that decision possible.
- Generation. The chosen sources are used to write the answer. This is where a mention or a citation can happen.
- Attribution. The engine may name the brand, link the page, or do neither. A mention without a citation is common, and it is why mention tracking and readiness cannot be read off one another.
Readiness is upstream because it removes the reasons not to cite you. It does not create demand, authority, or a question your page answers better than somebody else’s.
Can a page be ready and still go unmentioned?
Yes, routinely. A page can pass every check this scanner runs and still never appear in an answer, for reasons that have nothing to do with its markup.
- Competition. Another page may answer the question better, or be more recent, or be the source the engine retrieves first.
- Query demand. If nobody asks the question your page answers, there is no answer for it to appear in. Readiness cannot create a market.
- Index lag. Crawling, indexing and retrieval are not instant, and a page published last week may simply not be in the pool yet.
- Third-party sourcing. An answer can name your brand from a review site, a forum thread or a news article. Your own page being absent is not necessarily your brand being absent.
- Unattributed use. An engine can reuse your facts without naming the domain. A mention tracker may miss a use that readiness made possible.
Can a page be mentioned without being ready?
Also yes, and the two failure modes should not be confused. The mentions that arrive without a citable page are usually brand mentions from third parties - a review, a comparison, a directory listing. They are evidence that the brand exists in the conversation, not evidence that the site is ready to be the source.
That gap is common and it is not a paradox: the engine is naming what it learned about you elsewhere, while the page that could support the claim is blocked, thin or unparseable. Fixing the page is the only way to move the second kind of mention.
How do I use both together?
Audit first, monitor second. The order is not arbitrary: an audit produces edits, and a monitor produces a number that cannot be acted on until the pages behind it are fixed.
- Fix access first. Unblock the crawlers before measuring anything. A site that fails its own crawl checks is measuring the block, not the content.
- Then fix what a parser reads. Entity markup, question-shaped headings, short opening answers, numerals and quotations. The methodology page publishes the weights so the order of the work is visible rather than guessed.
- Record a baseline. Ask a fixed set of questions by hand, save the answers with the date, and keep the question set unchanged. The self-check guide is the do-it-yourself version of this step.
- Then monitor. Repeat the same questions on a schedule. A mention count is only meaningful against a baseline you defined before you started changing the page.
- When a mention appears, find the cited URL. Audit that page rather than the homepage. The page an engine chose is the one worth making stronger.
What does each measurement actually measure?
The parallel is easiest to hold in one table. The last two rows are the ones the two measurements disagree about, and the disagreement is the point.
| Question | Readiness audit | Mention tracking |
|---|---|---|
| What it measures | The state of your pages: access, structure, and whether an answer could lift a claim from them. | The presence of a brand in AI answers, counted over a set of prompts and a period of time. |
| What it needs | One URL per page, fetched the way a crawler would. No account for this site's audit. | A question set, access to model responses, and a brand definition that includes aliases and misspellings. |
| What it cannot tell you | Whether an engine will actually name or cite you. | Whether the cited page is yours, or whether a page that is never cited could even be crawled. |
| When it is the right tool | Before there is any evidence of citations, and whenever a page changes. | After there is content worth mentioning, to see whether it is surfacing. |
What can this site tell me, and what can it not?
This site runs the readiness audit: 12 weighted dimensions, in which AI crawler access is 16% of the score and the AI context file is 5% - a deliberately small weight for a file whose effect no engine has committed to. It reports a score, a grade and the list of failed checks with the evidence each check saw.
It does not query an engine, so it cannot tell you whether you are mentioned, cited or recommended. That boundary is stated in the scanner’s own about page and demonstrated in the 30-homepage study, where no AI engine was queried and no row is a measurement of what ChatGPT, Claude or Perplexity actually say.
Where should I start?
Run the free audit on the page you most want cited. Fix what it reports, and note which failures were real rather than page-type artefacts - reading a failed check is the guide for that judgement. Then, if you have content worth mentioning, set up a mention measurement against the fixed baseline.
The product comparison is on the LLMention vs DataForSEO’s LLM Mentions API page. The four-measurement map, including rank tracking and crawler logs, is on the comparison hub.