AI Provenance Mark
Pronounced "ape 'em" — rhymes with shape pum, phonetically ap-um.
A Creative Commons–style transparency mark for AI-assisted content. A small, recognizable badge paired with a QR code lets anyone instantly understand whether and how AI was used to create something — and makes it easy to share the original context or prompt with others.
Open Standard
Vendor-neutral and freely usable. No proprietary apps, no accounts, no tracking.
Declare AI involvement
From full AI generation to full human authorship — every meaningful level of involvement expressible in a single field.
Share your context
Provenance pages include direct links to open the context in Claude, ChatGPT, Gemini, and others — one click to continue the conversation that created the content.
What's New in 1.2
| Feature | Description |
|---|---|
| File integrity hashing | hs and hd store SHA-256 hashes of the content file and Full Context Document in W3C SRI format (sha256-base64url) — algorithm-agile for 30-year durability. Anyone can drag-and-drop or paste a hash (SRI, base64url, or hex — auto-converted) to verify the file is unchanged since marking. |
Content type (type) |
Declares the media type — text, code, image, audio, video, multimodal. Recommended for EU AI Act Article 50 compliance, where disclosure obligations differ by content type. |
Wrote role (role=wrote) |
New role for the most common AI-assisted writing scenario: you wrote the content, AI corrected grammar or clarity, you reviewed and approved. Show mode abbreviation: W. Intellectual authorship is fully human. |
| Automated role | New role=automated for AI-generated content with no human involvement — publishing pipelines, automated systems. Show mode abbreviation: AI. Use org to identify the responsible party. |
| File upload to context | Drop a .txt, .md, .pdf, .docx, or other text file onto the context field in the generator. Text is extracted client-side — nothing is uploaded to any server. |
| JSON-LD output | Each provenance page now emits structured Schema.org JSON-LD in its <head>, making AIPM records discoverable by search engines and AI crawlers without scanning. |
| Copy as citation | One-click button on the provenance page generates a formatted citation string suitable for footnotes, README files, and acknowledgements sections. |
| EU AI Act alignment | The type and org fields support Article 50 disclosure requirements. The provenance page shows an Article 50 notice for AI-generated text, image, audio, and video content. |
How It Works
AIPM combines three elements: a stable visual mark, a QR code, and a human- and machine-readable provenance record. Together they disclose who or what created a piece of content and what role a human played.
- Create a AIPM URL using the QR generator — fill in the model, your role, and prompt/context.
- Download the QR code or copy the badge and embed it in or near your content — the QR code works for print, slides, and images; the badge works for web pages, GitHub, and email.
- Anyone scanning the QR code or clicking the badge sees a clear, human-readable disclosure page showing how the content was created.
A 1.2 Provenance URL looks like this
All metadata is encoded in the URL hash fragment — nothing is sent to any server.
https://aipmq.org/1.2/aipm/#v=1.2&model=Claude+Sonnet+4.6&role=prompted%2Breviewed&date=2026-05-03T14%3A30-04%3A00&ctx=Internal+memo+draft&show=1
View example provenance page →
A 1.2 QR Code looks like this
Generated locally in your browser using the example URL above. The AIPM mark is centered in the QR code.
Scan this code or open the AIPM URL directly →
For digital content — GitHub READMEs, blog posts, web pages — use an AIPM Badge instead:
The badge is a clickable link to the provenance page. Download all badge variants from the Marks & Badges page →
Generate a QR Code Read the 1.2 Specification