Schema.org vs llms.txt: which AI search signal matters most?
Schema.org. JSON-LD identifies your entities (Organization, Product, Person) per page and powers Google rich results. llms.txt is a proposed site summary that Google Search ignores and that no major AI engine documents using to choose citations. Treat it as optional.
The two signals serve different layers of the AI visibility stack. Treating them as alternatives leads to picking the wrong one and shipping incomplete coverage.
Side-by-side
| Dimension | Schema.org JSON-LD | llms.txt |
|---|---|---|
| Scope | Per-page entity identification | Site-level declaration |
| Format | Embedded JSON-LD in <head> | Standalone markdown file at /llms.txt |
| Primary purpose | Identify entities (Org, Product, Person, etc.) | Summarize the site and link key pages |
| Maturity | Established schema.org vocabulary, widely supported | 2024 proposal (llmstxt.org), not a standard |
| Documented consumers | Google, Bing and other search engines | None documented for citations; Google Search ignores it |
| Update frequency | On content change | On structural change (product launches, etc.) |
When Schema.org matters more
Per-page entity identification and rich results in Google Search (breadcrumb, Article, Product). Google says structured data is not required for its generative AI features, but it remains the documented way to describe your entities to search engines.
When llms.txt is worth publishing
When you want a clean summary for the coding agents and AI tools that fetch it. Google says Google Search ignores it, and no major AI engine documents using it to choose citations (checked September 2026), so treat it as optional infrastructure, not a visibility signal.
The right framing: Schema.org describes each page's entities to search engines; llms.txt is an optional whole-site summary. Neither is a documented citation lever on its own.
How to audit both
Radar runs Schema Audit (per-page JSON-LD coverage) AND llms.txt Validator (site-level file structure) in parallel. The unified report shows where each signal is weak and which to fix first.
Want to take action on this?