The seven things an answer engine needs before it can quote you

Why seven, and why these seven?

Because these are the preconditions, not the ranking factors. Every check below measures whether an engine CAN fetch, parse and quote your page. None of them measures whether it should — that depends on what your pages say, who says it better, and how a given model weighs a thousand signals it has never published.

That is a much smaller claim than most of this category makes, and it is the only one anybody can support. No answer engine publishes its ranking function, so a method claiming to have reverse-engineered one is guessing with confidence. What is knowable is on this page.

The seven checks, and what each is worth

One hundred points, fixed and public. The engine makes four requests — your page, robots.txt, llms.txt, sitemap.xml — runs no browser, executes no JavaScript and gives up on a page that takes longer than twelve seconds, because the crawlers give up too.

• AI crawler access — 20 points. Reads robots.txt, resolved the way each crawler resolves it.
• Structured data — 20 points. Reads every application/ld+json block in the page source.
• Answer-shaped content — 20 points. Reads question headings, lists, tables and any FAQ section.
• Content extractability — 17 points. Reads the text present in the raw HTML, before any JavaScript runs.
• Metadata and identity — 10 points. Reads the title, meta description, Open Graph set and canonical link.
• llms.txt — 8 points. Reads /llms.txt, and whether it is genuinely plain text.
• Sitemap — 5 points. Reads sitemap.xml, or a Sitemap: line inside robots.txt.

What order should you fix them in?

Not the order of the points. Points measure what a check is worth; the sequence below is about what unblocks what, and following it means you never do work that a later fix invalidates.

Start with crawler access, because a root-level block makes the other six moot — a page an engine may not fetch cannot be scored by anything downstream. Then extractability, because structure describing text that never arrives describes nothing; if your pages assemble themselves in the browser, fix that before you mark anything up.

Then take metadata and the sitemap together. They are fifteen points, they are an afternoon, and they are the cheapest points on the list — do them while you are already in the head tag. Structured data comes next: one Organization block, validated, then the type that matches what each page actually is.

Answer shape comes second-to-last precisely because it is the expensive one. It is writing, not configuration: reading your headings, finding the nouns, rewriting them as the questions a buyer would type, and moving the answer from the fourth paragraph to the first. It is also the one that improves the page for people, which is the only reason worth doing it. Leave llms.txt for last — it is the cheapest to add and the least certain to be read, so it is a fine thing to finish on and a bad thing to start on.

What does a hundred out of a hundred not mean?

It does not mean an engine will recommend you. A perfect score removes the reasons you were being skipped; it does not supply a reason to be chosen.

It also scores one page at one moment. A homepage scoring a hundred can sit on top of documentation scoring twelve, and the pages a buyer's question resolves to are usually the deep ones. Run it against the pages you want quoted, not only the one you are proud of.