AI Authorship Disclosure

How I label the degree of AI involvement in my writing.

Placeholder — Will to write

This section will explain the philosophy behind the disclosure system. Key themes to cover:

  • Why transparency about AI involvement matters (especially on a site about AI collaboration)
  • The spectrum isn't about quality — all tiers are valid approaches
  • "Why not improve something if it's trivial to do so?" — the self-referential angle
  • What this means for readers and how to interpret the labels

The Three Tiers

Written by me

AI used for research or proofreading only

Shaped by me

AI drafted or structured; I substantially rewrote

Curated by me

AI wrote; I reviewed and approved

✎Written by me

The words are mine. AI may have helped with research — finding sources, summarising papers, answering factual questions — or with proofreading and catching errors. But the structure, argument, and prose are human-authored.

Example: An essay where I wrote every draft, but asked Claude to check for logical gaps or suggest sources.

⇄Shaped by me

A genuine collaboration. The AI may have produced a first draft, or structured the argument, or written sections that I then substantially rewrote. The final piece reflects both minds, with me as the shaping intelligence deciding what stays, what goes, and what gets rewritten.

Example: I outlined the key points, an AI drafted the sections, I rewrote 60% of it, we iterated until it was right.

◈Curated by me

The AI wrote this. I prompted, guided, reviewed, and approved it — but the prose is substantially machine-generated. I stand behind the content (I wouldn't publish it otherwise), but the voice is the AI's.

Example: The Disposability Problem — an essay Claude wrote that articulates something I believe but couldn't have written as well.

Component Variants

The disclosure appears in different forms depending on context:

Full variant — for article headers

<AIDisclosure tier="shaped" model="Claude Opus 4.5" />

Full with note — for additional context

<AIDisclosure
  tier="curated"
  model="Claude Sonnet 4.5"
  note="This essay emerged from a conversation..."
/>

Compact variant — for shorter pieces

<AIDisclosure tier="written" variant="compact" />

Inline variant — for bylines or metadata

Published January 15, 2026 · Shaped by me

Published January 15, 2026 · <AIDisclosure tier="shaped" variant="inline" />

All tiers (inline)

Written by meShaped by meCurated by me

All tiers (compact)

Why This Matters

Placeholder — Will to write

This section could cover the deeper philosophical ground:

  • The tension between authenticity and improvement — if AI can make an argument clearer, why wouldn't you use it?
  • What "voice" means when authorship is distributed
  • How this connects to The Disposability Problem — treating AI as collaborators rather than tools
  • The reader's right to know, and what they do with it

Questions about this system? Get in touch.