What structured data and llms.txt can—and cannot—do

Give markup a specific job

Google says its Search systems do not use llms.txt for visibility or ranking, and does not prescribe a special AI writing format. Google: Generative AI search guidance. Its structured-data policies require accurate, relevant markup consistent with visible content; eligibility does not guarantee a rich result. Google: Structured data policies.

Our rule is to start with the visible facts. Use Organization information for the real publisher, Article information for an authored guide and applicable product or other types when the page supports them. Do not add a fake rating, local address, certification or human author to make a validator look more impressive.

A small example tied to visible facts

Educational example — fictional organization.

The example page visibly says “Example Studio is an independent design business. Visit https://example.com/.” Its minimal organization record could be:

{
  "@context": "https://schema.org",
  "@type": "Organization",
  "name": "Example Studio",
  "url": "https://example.com/"
}

This does not establish eligibility for a particular search enhancement. The point is that each value can be checked against the visible business identity. Add more properties only when they are true, maintained and appropriate to the chosen type.

Validate more than JSON syntax

1. Parse the markup to catch broken syntax.
2. Compare names, URLs, prices and dates with the visible page.
3. Check the documented properties for any search feature you intend to support.
4. Use the provider’s appropriate validation tools and record remaining warnings.
5. Recheck after a price, policy or publishing-date change.

A generic schema validator and a search feature test answer different questions. Neither proves a model cited the content. Keep the original source facts in ordinary readable HTML; the markup should not be the only place a customer can find the information.

If a particular downstream integration uses llms.txt, assess its requirements as that integration’s needs. Maintaining a parallel file with stale prices would undermine the source consistency you were trying to improve. Fix clear content and access problems first.