Can an answer engine cite your site?

An answer engine can only quote what it can read. Most of the crawlers feeding them do not run JavaScript, so what gets stored for your URL is whatever your server sent in the first response.

What a crawler actually receives

Open your own page, view source, and read what comes back. That text, before any script runs, is the whole of what most crawlers keep. A single-page app often sends a document containing one empty element and a script tag, and a browser fills it in a moment later. A crawler that does not run scripts never sees that moment.

This is the single most expensive thing a site can get wrong here, because it is not a matter of degree. A page that arrives empty is not quoted less often. It is not quoted.

The four things that stop a page being quoted outright

These are definitional rather than measured. Each one makes citation impossible whatever else is true of the page, so no amount of good structured data underneath changes the answer.

• The page says noindex, so it is asking not to be in an index at all.
• The served document carries no readable text without JavaScript.
• The server answers a crawler user agent with 403 while serving a browser normally.
• The canonical tag points at a different host, handing the page away.

What is not a barrier, despite the advice

Citecheck was calibrated against pages answer engines demonstrably quote every day: Wikipedia, MDN, and the Python, React, nginx and Postgres references. The rule was that such a page must come out in the top band, because nothing is in fact standing in its way.

That corpus overturned most of a first cut made from intuition. Five of those six ship no JSON-LD at all. Four carry no meta description. Three have no canonical, and two have no h1. None of those can be a barrier to citation when the most-quoted references on the web are missing them.

So best practice is not the same as a barrier, and only the barriers are scored heavily. An absence is a choice and costs nothing. A broken attempt is a defect and keeps its weight: a canonical that will not parse, JSON-LD that will not read, an hreflang no engine recognises. Somebody meant those to work, and they do not.

robots.txt, sitemaps and llms.txt

A crawler reads robots.txt before it reads a page. A mistyped field there is skipped in silence, so the rule its author wrote was never in force and nothing anywhere reports it. Naming your sitemap in robots.txt costs one line and is the difference between a crawler finding it and guessing.

llms.txt is a newer convention: a plain-text guide written for agents rather than browsers. Publishing one while separately telling most AI crawlers not to fetch anything is a contradiction, and one of the two files is stale.

Blocking AI crawlers is a decision, not a defect

Nothing here scores a site for turning an agent away. That is a choice a site is entitled to make, and charging for it would be charging somebody for meaning what they said. What is worth reporting is two files the same site serves giving opposite instructions.