What schema.org, JSON-LD, and llms.txt each cover

Article is the schema.org type for an article. inLanguage is the property for the language of the content. JSON-LD 1.1 is a W3C Recommendation dated 16 July 2020. It specifies how to write linked data in JSON. It is not a ranking algorithm.

If Article markup is going to match what a reader can see, the headline, dates, and language should be the ones on the page. A sentence the reader cannot see does not belong in the markup.

The structured data introduction says structured data gives Google explicit clues about the meaning of a page and can enable rich results. It recommends JSON-LD in most cases, and it says not to add structured data about information that is not on the page.

The general guidelines are more direct. Google does not guarantee that structured data will appear in search results, even if the Rich Results Test passes. The markup must represent the main content of the page, and it must not describe content hidden from the reader. The guidelines also say a structured-data manual action affects eligibility for rich results and does not affect how the page ranks in Google web search. The subject of those sentences is Google Search. They do not say ChatGPT or Perplexity cites a page because it has JSON-LD.

The HowTo and FAQ rich result note of 8 August 2023 says FAQ rich results would, from then on, be shown regularly only for well-known, authoritative government and health sites, and not regularly for other sites. The update of 14 September 2023 says How-to rich results are no longer shown on desktop either. The post says unused structured data does not cause problems for Search and also has no visible effect, and that the change should not be treated as a ranking change.

Adding FAQ markup to a marketing page in the hope that a generated answer will cite it goes beyond that post, and it does not match the post’s description of Google Search rich results.

RFC 9309 is the IETF Robots Exclusion Protocol. The abstract says site owners use it to control whether, and how, automatic clients — crawlers — may access the content they serve. It does not define how a generated answer chooses a sentence.

The OpenAI crawler overview and the Perplexity crawler page both use robots.txt user-agent names as switches. The meaning of each switch is the one in their own docs: the search crawler and the training crawler are separate.

The llms.txt page is signed by Jeremy Howard. It proposes an llms.txt file that gives an agent a short briefing and links. The page presents the file as a proposal and says it can sit beside robots.txt and a sitemap because the jobs differ: robots.txt says which automatic access is acceptable, and llms.txt is read when an agent needs background. The page mentions the Well-Known URI pattern in RFC 8615 as another approach, which the proposal does not use.

The proposal page says several labs publish llms.txt for their own developer documentation. OpenAI’s crawler documentation page also says its documentation index is an llms.txt file. That is an index for a docs site. It is not a statement that ChatGPT search reads llms.txt on your site and cites you because of it. None of the OpenAI or Perplexity documents checked for these notes lists llms.txt as an input to search citation. Adoption claims on the proposal page are not a brand result.