Does schema markup or llms.txt improve AI visibility?
They help at the margins but they are hygiene, not growth. Schema.org markup makes content easier for engines to parse, and some studies associate it with higher citation rates, but Google has stated structured data is not required to appear in AI features, and that it does not use llms.txt. Neither is a substitute for earned media and citable content.
Schema.org: useful, not decisive
Structured data (Article, FAQPage, Organization, Product) helps an engine understand and confidently quote your content, and some analyses report meaningfully higher citation rates for pages with rich schema. But it's a readability aid, not a ranking lever: Google's documentation states schema is not a requirement for AI features. Treat it as table-stakes that removes friction: implement it, then move on to the work that actually grows presence.
llms.txt: unproven
llms.txt is a proposed file for telling LLMs how to use your site. As of now Google has said it does not use it, and there is no evidence the major engines weight it. It's harmless to publish and good developer hygiene, but it will not move your AI visibility on its own; don't let it crowd out the levers that do.
Where the leverage actually is
The KDD 2024 "GEO" paper found that adding citations, quotations and statistics raised a page's share of an answer it was already in by about 27% to 41%, in a simulated engine; later benchmarks found most rewrite methods do not hold across domains. Earned media and authentic third-party corroboration are where the larger gains live. Schema and llms.txt make you legible; content and citations make you citable.
Frequently asked
Should we still add schema markup?
Yes, it's cheap, it removes parsing friction, and it can help an engine quote you accurately. Just don't expect it alone to grow your citations; treat it as foundation, not strategy.
Is llms.txt worth implementing?
It's fine to publish as developer hygiene, but there's no evidence the major AI engines use it today (Google has said it does not). Don't prioritise it over content and earned media.
What lifts AI citation rates the most?
Research points to citation-grade content (answer-first, with statistics and quotes) and earned media across independent sources, not markup. In a simulated engine, adding stats and citations raised a page's share of an answer it was already in by about 27% to 41%.
Published by New Map, the AI visibility platform. Updated 2026-10-15.