Echo

What this actually checks

Nine things, weighted by how much they affect whether an answer engine can use your content. The heaviest one by far is whether your body text exists in the HTML the server sends, because that is the difference between being quotable and being invisible.

The rest: which LLM crawlers your robots.txt allows or blocks, one by one; whether Googlebot and Bingbot are allowed; whether robots.txt exists and declares a sitemap; whether the sitemap resolves and how many URLs it holds; whether llms.txt and llms-full.txt are present; whether the page has a title and meta description; whether the heading structure is usable; and what structured data types you publish.

Why JavaScript rendering is the one that matters most

Google renders JavaScript. It has for years, on a second pass. LLM crawlers largely do not. GPTBot, ClaudeBot and PerplexityBot mostly fetch the HTML and read what is there.

So a single page application can rank acceptably in Google and be completely absent from AI answers. We have audited sites that carefully allowed every LLM crawler in robots.txt, and then served those crawlers a 3 KB shell with no headings and no body text. The invitation was sent and there was nothing in the room.

If this check fails for you, it is the only one worth fixing first. The others are hygiene by comparison.

Do you actually need an llms.txt?

It is a proposed convention, not a standard anyone is required to honour, and no major provider has committed to reading it. We check for it because it costs nothing, it is a clear signal of intent, and the sites investing in answer-engine visibility tend to have one. Treat a missing llms.txt as a nudge, not an emergency. Treat client-side rendered body content as an emergency.

What to do with a low score

Work in the order the tool gives you, because the weights are not arbitrary. Rendering first, then crawler permissions, then structure, then the optional signals. Fixing structured data while your content is invisible to crawlers is effort spent on the wrong layer.

If the fix is rendering and your site is a single page application, that is usually a prerendering or server rendering change rather than a rewrite. It is a well understood piece of work with a known shape, and it is the kind of thing we do when we build production AI, end to end.