Authorship Tags
A small, parseable convention for describing the provenance of published text.
Summary
Authorship Tags are compact strings that communicate how a piece of writing was produced with respect to AI involvement. They are designed to be:
- human-readable
- machine-parseable
- unbranded and portable
- descriptive, not prescriptive
Format
A<level>[-<modifier><modifier>…]
A— fixed prefix (Authorship)<level>— single digit0–3-— optional separator<modifier>— one or more lowercase letters
Examples:
A0 A0-r A1-re A2-c
Levels
A0 — Human Authored
Final text written by a human. No AI-generated prose is included.
Allowed:
- spelling and grammar tools
- search and retrieval
- non-generative tooling
A1 — AI Assisted
AI used during the writing process, such as ideation, outlining, or suggestions. Final text written by a human.
A2 — AI Contributed
AI-generated text appears in the final published work. Human authors curated, edited, or arranged the result.
A3 — AI Generated
The majority of the final text was generated by AI. Human role is primarily selection or arrangement.
Modifiers
Modifiers provide optional detail about how AI was used.
They are:
- lowercase
- appended after
- - unordered
Defined modifiers
r— research / retrievale— editing suggestionst— translationc— code generationd— data analysis
Examples:
A0-r # human-authored, AI used for research A1-re # assisted with research and editing A2-c # AI contributed code/content
Modifiers are optional. Omit them if unnecessary.
Interpretation Rules
-
Final artifact rule The tag reflects what appears in the final published text, not the full drafting process.
-
Survival rule If AI-generated text remains in the final work, the level is at least
A2. -
Replacement rule If AI outputs were fully rewritten and do not appear in the final text, they do not elevate the level beyond
A1. -
Baseline tools Basic spelling and grammar tools do not change
A0.
Usage
Authorship Tags can be placed anywhere appropriate:
Inline:
A0-r
With description:
A1-re — AI Assisted (research, editing)
Footer:
Authorship: A2
Non-goals
Authorship Tags do not:
- evaluate quality
- imply correctness or truth
- prescribe acceptable use of AI
- enforce policy
They describe provenance only.
Examples
| Scenario | Tag |
|---|---|
| Fully human-written essay, AI used for search | A0-r |
| AI helped outline, author wrote everything | A1 |
| AI-generated paragraph included verbatim | A2 |
| AI-generated draft lightly edited by human | A3 |
Status
This is a minimal, evolving convention. Refinements should preserve simplicity and backwards compatibility.