Find your AI voice anchor.
AI emulates voices it knows deeply. Paste your writing. We'll find the training-data authors your prose mechanically resembles. Use the blend as an anchor in any agent to write more like yourself.
300 to 8000 words. Use writing that feels characteristic. Essays, posts, or anything you'd want any output to sound like.
Your text is used once to compute your match. Never stored. Never used to train models. Details
Your anchor blend
Authors your prose mechanically resembles, weighted by feature overlap.
Your writing fingerprint
Deterministic measurements of how your prose is constructed. These drive the match.
Use your anchor
Copy the block below. Paste it as the first message in any AI agent, or save it as your project's writing rules.
How to use this block manually
In Claude, ChatGPT, or any chat agent: paste the block above as your first message. The agent will follow it for the conversation.
In OpenWriter: save it as your project's writing rules. Every rewrite anchors to your blend.
Install the authors-voice skill
The anchor alone gets you 60 to 70% of the way to your voice. Install the free skill to add the rest. NEVER rules and sentence rhythm extracted from a growing local corpus on your disk. Pure markdown, works with any AI agent that reads SKILL.md.
After install, tell your agent: "set up my voice match". It walks you through pasting this block plus seeding your corpus. Every future write the agent does for you applies your voice profile automatically.
Plug into your apps
The same anchor system also powers the paid Author's Voice plugin (inline voice edits inside OpenWriter, the free AI writing surface) and the paid API (programmatic voice-matched output for newsletters, blog drafting, social schedulers, and any workflow that needs your voice).
What happens to your text
- Your source text is sent once to compute your match. Not stored persistently.
- The derived result is cached by hash for 24 hours so identical inputs don't re-run. After that, gone.
- Your text is never used to train models. Anthropic's API has data retention controls we honor.
- Shared result pages show only the blend and fingerprint, never your source text.
- No login required. No tracking beyond hashed IP for rate limiting.
Free use is rate-limited to 10 matches per IP per hour. For higher volume and automated workflows, see the API.