{
 "question": "How many top websites answer an AI agent's request for Markdown, advertise a Markdown version, or describe themselves with JSON-LD on the homepage?",
 "population": "Tranco top-10,000 domains with a live homepage (n=5,902); HTML-only analyses use the 5,710 homepages returned as HTML.",
 "inclusion": "Homepage requested with Accept: text/markdown, text/html;q=0.9, */*;q=0.8. Markdown negotiation = response Content-Type text/markdown.",
 "exclusion": "Homepages that failed or returned 400+.",
 "calculations": [
  "JSON-LD types collected recursively (including @graph).",
  "Organization = Organization, Corporation, NewsMediaOrganization, LocalBusiness, OnlineStore, EducationalOrganization, GovernmentOrganization or NGO.",
  "Sentence and name matching are lexical proxies (word trigrams and capitalized phrases), not semantic verification; no human or model coder judged meaning.",
  "Representation check (version 1.1, 28 September 2026): each of the 191 homepages that served Markdown on 26 September was requested four more times, twice asking for HTML and twice for Markdown, in a seeded random order. A pair counts as the same content only if every number, date, price, heading, link and name-like phrase of each version appears in the other, and at least 95% of its sentences do. Only items present in both requests of the same version are compared, so content that changes between requests is not counted as a difference.",
  "Sentence and name matching are lexical proxies (word trigrams and capitalized phrases), not semantic verification; no human or model coder judged meaning."
 ],
 "limitations": [
  "Homepages over 600 KB were re-fetched in full (790 of 807 succeeded); for the other 17 only the first 600 KB was searched for JSON-LD.",
  "Only the homepage was tested.",
  "Server header used as a hint of the serving platform; it can be masked.",
  "The analysis with menus, headers and footers removed was added after seeing the results and is exploratory.",
  "The representation check compares what a site sends; it does not show whether any AI system requests the Markdown or uses it in an answer.",
  "The analysis with menus, headers and footers removed was added after seeing the results and is exploratory."
 ],
 "update": "quarterly",
 "@context": "https://schema.org",
 "@type": "Dataset",
 "name": "Do websites serve Markdown to AI agents? 2026 data",
 "version": "1.1",
 "dateCreated": "2026-09-26",
 "creator": {
  "@type": "Organization",
  "name": "Underneath",
  "url": "https://underneath.agency"
 },
 "url": "https://underneath.agency/research/agent-readable-web-study",
 "temporalCoverage": "2026-09-26",
 "isAccessibleForFree": true,
 "license": "https://creativecommons.org/licenses/by/4.0/",
 "code": "cite/pipeline/ (collection, validation and analysis scripts)",
 "variableMeasured": [
  "sample",
  "s3_md_negotiation",
  "s3_md_servers",
  "s3_md_with_vary_accept",
  "s3_html_n",
  "s3_alt_markdown_link",
  "s3_capped_refetched",
  "s3_jsonld",
  "s3_note",
  "s3_capped_pages",
  "s3_jsonld_by_tier",
  "s3_derived",
  "s3_md_vs_html",
  "s3_capped_not_refetched",
  "s3_representation_check"
 ],
 "dateModified": "2026-09-28"
}