PLGen is a plain-text syntax for declaring the provenance of AI-assisted content. It is human-readable, tool-parseable, and implementation-agnostic. Any tool — a CLI, a website, a browser extension — can read or write a valid .plgen file using this specification.
Think of it like Markdown: a simple set of rules that does useful work across many contexts without requiring special software to understand.
.plgenkey: value# are comments and are ignored@provenance directive must appear on line 1| Field | Type | Description | |
|---|---|---|---|
@provenance |
directive | Version declaration. Must be line 1. | Required |
author |
string | Full name of the human accountable for the content. | Required |
date |
YYYY-MM-DD | Date of completion or publication. | Required |
human |
integer 0–100 | Percentage of human contribution. Must sum to 100 with ai. |
Required |
ai |
integer 0–100 | Percentage of AI contribution. Must sum to 100 with human. |
Required |
tools |
string | AI tool(s) used. Comma-separated if multiple. | Optional |
shared-by |
string | Name of person republishing content they did not originally create. | Optional |
updated |
YYYY-MM-DD | Date of most recent substantive revision. | Optional |
note |
string | Single plain-language sentence describing the collaboration. | Optional |
ref |
URL or citation | Source referenced during creation. Repeatable for multiple sources. | Optional |
# PLGen v1.0 — provenance-label.org/spec
@provenance 1.0
author: Shelton Davis
date: 2026-02-17
human: 65
ai: 35
tools: Claude Sonnet 4.6
note: Concept and direction human. Draft and structure AI. Edited for voice.
ref: https://provenance-label.org/spec
Any valid implementation must support at minimum the --short and --long output formats. --json is recommended for tool integrations.
Single-line badge. For inline use, footers, social posts.
Full disclosure block. For articles, documentation, reports.
Machine-readable export for tool integrations and APIs.
PL v1.0 | Shelton Davis | 2026-02-17 | Human 65% · AI 35% · Claude Sonnet 4.6
─────────────────────────────────────
PROVENANCE LABEL v1.0
─────────────────────────────────────
Author: Shelton Davis
Date: 2026-02-17
Human: 65%
AI: 35%
Tools: Claude Sonnet 4.6
Note: Concept and direction human.
Draft and structure AI.
Edited for voice.
─────────────────────────────────────
provenance-label.org/spec
{
"plgen": "1.0",
"author": "Shelton Davis",
"date": "2026-02-17",
"human": 65,
"ai": 35,
"tools": "Claude Sonnet 4.6",
"note": "Concept and direction human. Draft and structure AI. Edited for voice."
}
@provenance must appear on line 1 of the filehuman + ai must equal exactly 100date and updated must be valid YYYY-MM-DD formatnote must be a single line (no newlines)ref may appear multiple times for multiple sourcesSubmitted labels will receive a permanent short ID — for example, PL-000312 — that can be embedded in content. Anyone can look up a label ID at provenance-label.org/r/PL-000312 to verify the original, timestamped disclosure.
The @provenance directive specifies the spec version. Major version changes indicate breaking format changes. Minor version changes are backwards-compatible additions. Implementations should specify the version they support.
This specification is licensed CC BY 4.0. Anyone can build implementations — CLI tools, website generators, browser extensions, GitHub Actions, editor plugins — as long as they reference this spec. Implementation code is licensed MIT.
To propose changes to the spec, open an issue on the GitHub repository.