Observed arrival · 2026-09-21
TextProv, provenance marks for plain text
An open protocol for labeling passages as human-written, AI-generated, mixed, edited, or unknown, with those labels traveling inside the text.
- For
- Editors and developers handling mixed-origin text
- Worth noticing
- Labels are embedded in plain text and can persist through copy, paste, and storage when receiving software preserves the marks.
Field notes
TextProv encodes provenance labels into text rather than storing them only in application metadata, allowing compatible tools to carry disclosures through ordinary text handling. The homepage includes a browser reader, encoder, marked samples, conformance material, and links to Python, JavaScript, and Ruby implementations. Its Draft v0.1 specification explicitly limits the system: labels record claims about origin, but do not detect AI writing, authenticate sources, or establish truth.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue