llms.txt adoption among sites that rank for SEO queries

The sites in this sample sell SEO and AI search services or write about them, and several popular SEO plugins and documentation platforms generate llms.txt automatically. Adoption therefore reflects tooling defaults and an industry experimenting with AI search tactics. It should not be read as evidence that the file improves visibility; no major AI search provider has said it uses llms.txt to choose sources.

The llms.txt proposal asks for a Markdown file that begins with an H1 naming the site, a short blockquote summary, and sections of curated links. 71 files at the path returned text that did not start with a heading, so they do not follow the format and are less likely to be useful to the agents and tools that read llms.txt.

The median conforming file carried 42 links and about 10 KB of text. Many appeared to list most of a site's pages rather than a curated selection. Very long files work against the proposal's purpose, which is to point a model to the few pages that best explain a site. If you publish one, curate it and keep descriptions factual.

We publish both llms.txt and llms-full.txt, generated at build time from the same registry as our sitemap, so they include only released pages. We treat them as a convenience for agents and developers, not as a ranking tactic. Our free llms.txt generator builds a starting file from a sitemap.

The sample is biased toward the SEO industry and is not representative of the web. We checked only the root path, not alternative locations. A file existing does not mean any AI system reads it. Some sites may return the same content for every path, which our HTML check would not always catch for plain text responses. SerionFlow is the wrong tool for checking llms.txt across many sites; this was a one-off study, not a product feature.