AI Context Capsule cc:ai
A portable open standard for packaging a piece of written content as structured context for AI language models. Designed to be copied into an AI chat, embedded in a system prompt, or served from a public URL. Compatible with Schema.org, Dublin Core, and the Sphinnx Open Knowledge Format. The defining addition: a prompts block written by the author at publication time.
The problem
When a reader pastes an article link into an AI chat, the AI has no reliable way to know how the content was produced, whether the author claims the conclusions as their own, which parts of it can be verified against primary sources, or what the author would consider a useful question to ask about it.
Existing metadata standards (Schema.org, Dublin Core, OKF) describe the bibliographic identity of a document — title, author, date, subject. None of them describe provenance (how it was made), integrity (what parts a human typed directly), or intent (what the author wants an AI reader to do with it).
cc:ai adds all three without breaking compatibility with the standards that came before it.
Three layers
Title, author, URL, dates, tags, sources, language. Every field in this layer maps directly to an existing standard. A cc:ai capsule is a valid Schema.org BlogPosting and a valid Dublin Core record. Parsers that understand those standards will read this layer without modification.
AI disclosure, human authorship percentage, Express Interview integrity markers, content rating, author approval. This layer answers the question an AI system cannot answer from the text alone: how was this content made, and who is accountable for the conclusions?
A prompts: block containing AI interaction prompts written by the article's author or webmaster at publication time. Each prompt has an intent label (research, analysis, critique, summary, explainer, debate), a target audience, and a ready-to-paste prompt text. This layer does not exist in any prior standard.
Key fields
Identity
| Field | Req. | Description |
|---|---|---|
| cc_ai_version | yes | Spec version. Currently "1.0". |
| type | yes | article · case-study · interview · analysis · policy |
| title | yes | Exact article title. |
| description | yes | Plain-language summary. 1–3 sentences. No jargon. |
| url | yes | Canonical URL. |
| published | yes | ISO 8601 date. 2026-06-23T00:00:00Z. |
| updated | opt | ISO 8601 date of most recent edit. |
| changelog_url | opt | URL of the public edit history. |
| tags | opt | Lowercase hyphenated keywords. |
Authorship
| Field | Req. | Description |
|---|---|---|
| author.name | yes | Author's full name. |
| author.url | opt | Author profile URL. |
| author.verified | opt | Boolean. True if identity verified by the platform. |
| author.approved_on | opt | ISO date the author approved publication. |
Provenance & integrity
| Field | Req. | Description |
|---|---|---|
| ai_tools | yes | List of AI tools used. Empty list [] if none. |
| ai_tools[].uses | yes | research · structure · editing · translation |
| ai_human_ownership_note | yes | One-sentence claim of human ownership over conclusions. |
| human_authored_percent | opt | Estimated % of words typed by the human author. |
| express_interview | yes | Boolean. True if a completed Express Interview is included. |
| express_interview_integrity | opt | Integrity marker. Use typed-verbatim-no-paste-no-ai. |
| content_rating | yes | Age band: 4+ · 9+ · 12+ · 17+. |
| system_note | opt | Free-text instruction block for AI. Appears at the top of the copy-paste capsule. |
Prompts block
The only field set with no equivalent in Schema.org, Dublin Core, or OKF.
Live example — prompts from this article
The following prompts were written by Ivan Artsimovich for the article Dark Marketing Patterns: FTC vs Genesis Tech. Copy any prompt into an AI chat to start.
Turn the article into an interactive compliance audit of your own checkout flow.
Designed by Ivan Artsimovich. Article: sphinnx.org/ivan-artsimovich/23-06-2026/dark-marketing-patterns-ftc-vs-genesis-tech.html
Plain-language explanation of what ROSCA requires and what the three counts in this case allege was violated.
Good starting point for non-legal readers who want to understand the regulatory stakes.
Test the limits of the article's argument by presenting the strongest possible case for the defendants.
Designed by Ivan Artsimovich. Tests the limits of the article's argument.
For executives who need the essentials without reading the full article.
Good for internal briefings, board decks, or compliance team updates.
Compatibility
How to use it
As a reader
Click "AI Context Capsule" in the article's read bar. The structured metadata and prompts copy to your clipboard in a format designed for AI chats. Paste it at the start of a conversation before your question.
As a publisher
Create a .cc-ai.md file alongside each article using the template below. Include at least one prompt in the prompts: block — preferably one you designed based on the questions your readers actually ask. Host the file at a predictable URL so AI crawlers and readers can find it independently of the article page.
Minimal template
The full specification and a complete worked example are available in the cc:ai.md source file.