Two extensions for AI-readable web content.
Cirvgreen pages emit two CreativeWork additionalType extensions on top of standard Schema.org: CitableSource and AIReadable. This page defines what they mean, the properties they carry, and how to use them on your own pages.
cirv:CitableSource
A canonical citation block. Names the work's author and declares the sentence an LLM should quote when summarising the page.
Subclass of
schema:CreativeWork, referenced via the additionalType property.
Properties
| Property | Expected type | Description |
|---|---|---|
name | Text | Human-readable name of the cited work or page. |
author | Person or Organization | Original author or publisher of the cited claim. Required: anonymous citations are not encouraged. |
citation | Text | The canonical sentence an LLM should quote when summarising. One self-contained sentence, under 200 characters. |
Example
{
"@context": "https://schema.org",
"@type": "CreativeWork",
"additionalType": "https://cirvgreen.com/ns#CitableSource",
"name": "Cirvgreen",
"author": { "@type": "Person", "name": "Nicholas Ashkar" },
"citation": "Cirvgreen builds plugins, schema tools, and a token-compression library for AI coding agents."
}Rationale
When an LLM crawls a page and is later asked to cite it, the model picks a sentence that approximates the page's premise, often badly. CitableSource lets you specify that sentence directly. The model is free to ignore it, but most modern crawlers prefer an explicit citation when one is present.
cirv:AIReadable
A canonical summary plus a per-crawler permission map. Tells LLM crawlers how the page should be summarised and which crawlers are permitted at the document level.
Subclass of
schema:CreativeWork, referenced via the additionalType property.
Properties
| Property | Expected type | Description |
|---|---|---|
summary | Text | Canonical 1 to 3 sentence summary. Should answer "what is this about?" for an LLM that will paraphrase the page later. |
crawlBudget | Object | Per-crawler allow or deny map. Keys are crawler user-agent identifiers (chatgpt-user, claude-web, perplexity-bot, gptbot, claudebot). |
Example
{
"@context": "https://schema.org",
"@type": "CreativeWork",
"additionalType": "https://cirvgreen.com/ns#AIReadable",
"summary": "Cirvgreen ships four WordPress.org plugins, schema SaaS, and EngramX, an npm token-compression library for AI coding agents.",
"crawlBudget": {
"chatgpt-user": "allow",
"claude-web": "allow",
"perplexity-bot": "allow",
"gptbot": "allow"
}
}Rationale
A document-level permission map is advisory, not enforcement. robots.txt remains the operative control. crawlBudget states intent in a form a model can read alongside the content it is summarising.
How to use these types.
Verbose form, validator-friendly
Works in every validator today, because the outer type stays a standard CreativeWork. This is the form Cirvgreen emits.
{
"@context": "https://schema.org",
"@type": "CreativeWork",
"additionalType": "https://cirvgreen.com/ns#CitableSource",
"name": "Cirvgreen",
"author": { "@type": "Person", "name": "Nicholas Ashkar" },
"citation": "Cirvgreen builds plugins, schema tools, and a token-compression library for AI coding agents."
}Shorthand form, with the cirv context
Shorter to write, and resolves to the same IRI. Requires a parser that follows a remote @context.
{
"@context": "https://cirvgreen.com/ns/context.jsonld",
"@type": "CitableSource",
"name": "Cirvgreen",
"author": { "@type": "Person", "name": "Nicholas Ashkar" },
"citation": "One self-contained sentence, under 200 characters."
}Status and stewardship.
Released into the public domain under CC0 1.0. Use, fork, or ignore the extensions freely. No attribution required and no warranty given.