The site map for language models.
A plain-text directive at your domain root that tells AI engines what your site is about, which pages matter, and where to look first. The signal-to-noise ratio they wish your homepage had.
Written forChatGPT (GPTBot)Claude (ClaudeBot)Perplexity (PerplexityBot)Google GeminiBing AI
Read full details →Your page, in the format LLMs were trained on.
Clean Markdown stripped of nav, ads, scripts, and CSS. Headings preserved, link targets normalized, content placed in a single retrieval-tight document. The format AI reads first when given a choice.
Written forGPTBotClaudeBotPerplexityBotGeneral LLM systems
Read full details →Structured data, the way Google and AI engines prefer it.
JSON-LD using schema.org vocabulary, page-typed automatically. Geordy detects whether a page is an Article, Product, FAQPage, or Organization and emits the right shape - including required fields most sites silently miss.
Written forGoogle SearchBingVoice assistantsAI citation systemsKnowledge graph systems
Read full details →Page metadata, indentation-clean.
Configuration-style metadata: title, summary, canonical, taxonomy, last-updated. Easy for tooling to ingest, easier still for retrieval pipelines to filter.
Written forChatGPTClaudePerplexityBuild toolsRAG pipelines
Read full details →Continuous discovery for new and updated pages.
An XML feed of every new or recently-updated page on your site, with full content embedded. AI crawlers poll it instead of guessing when to re-crawl.
Written forGPTBotClaudeBotPerplexityBotFeed readersAI training pipelines
Read full details →Identity declaration for installable surfaces.
Declares your site's name, icons, language, description and entry point in one JSON file, at a fixed address, in a shape browsers and app stores already agree on.
Written forChrome and Edge install promptsSafari Add to Home ScreenMicrosoft Store via PWABuilderGoogle Play via Trusted Web Activity
No AI engine publishes that it reads manifest.json.
Read full details →Attribution for the people behind the site.
Names the team, tools and last-updated date behind the site, in plain text, at a conventional address. The one fixed place to point at when somebody asks who maintains this.
Written forPeople who open itAnything following your rel=author link
No AI engine publishes that it reads humans.txt, and this page will not name one that has not.
Read full details →Open Graph metadata, centralized.
All social and AI preview metadata in one place - title, description, image, type. Powers chat surfaces, social embeds, and assistant cards from a single source.
Written forAI chat surfacesSocial platformsPreview generators
Read full details →Dataset discovery, the W3C way.
RDF-based catalog vocabulary for datasets - endpoints, distributions, licenses. AI agents use this when looking for structured data assets, not articles.
Written forGPTBotClaudeBotPerplexityBotLLM crawlers accessing semantic metadata
Read full details →Logic, rules, and inference.
RDF logic-based notation. Lets AI systems reason over relationships in your content rather than just extract them.
Written forReasoning enginesKnowledge graph systemsAI agents using inference
Read full details →RSS, but JSON-native.
A modern alternative to RSS, easier to parse for JSON-first pipelines. Useful for AI agents that consume live syndication without an XML parser.
Written forModern AI agentsDeveloper-built RAG pipelinesJSON-first crawlers
Read full details →Minimal, human-friendly config.
An alternative to YAML with cleaner type rules: a date is a date, not a string that looks like one. The shape to hand your own tooling when it has to read page metadata without guessing.
Written forEvery major language's TOML parserYour own build and content tooling
No AI engine asks for page.toml.
Read full details →Compact RDF triples.
A concise serialization of semantic relationships. The lingua franca of knowledge graphs.
Written forKnowledge graph systemsSemantic search enginesChatGPTClaude
Read full details →Schema definition with validation.
A constraint language for data validation. The file you run your other generated files against, so a broken shape fails your build instead of reaching a crawler.
Written forcue vet, in your CICUE-based tooling: Dagger, Timoni, Holos
Nothing on the AI side reads .cue.
Read full details →Federated identity, in the Fediverse's own shape.
The W3C protocol the Fediverse runs on. Publish your updates as typed, addressable JSON-LD objects and a Mastodon or Threads server can dereference them the way it does any other actor.
Written forMastodonLemmyPeerTubePixelfedThreads
These are the servers that speak the protocol, named because you can go and check them.
Read full details →Compact representation for repeated terms.
A stenography dictionary - shorthand outlines mapped to their full translations. Used to teach AI engines that every variant of a term resolves to the same canonical phrase.
Written forChatGPTClaudePerplexity
Read full details →