AI writes the draft.
Proselike makes it sound like you.
Every frontier model now writes in the same flat voice, and readers can tell. Proselike learns your style from things you've already written. It rewrites any AI draft in that style and gets sharper with every edit and every 👍 or 👎 you give it. A stylometry judge scores every output on how much it reads like you and how much it reads like AI.
Use AI, still sound like yourself
For founders, creators, newsletter writers and anyone who drafts with ChatGPT or Claude but publishes under their own name. Paste the AI output and get it back in your rhythm, word choice and habits.
Built from your past writing, then refined by your edits and 👍 or 👎 feedback. Two gauges on everything: reads like you and reads like AI.
Teach it your style →Every client's style, for your whole team
For ghostwriting agencies and founder-brand shops. It's the same style engine with a workspace: one locked style per client, shared across writers, and a “reads like the client” score you can show them.
Client folders, seats and consent attestations on file.
Style Studio →Say it without it pointing back at you
For sources, whistleblowers and activists. Strip your stylistic fingerprint, and don't swap it for the “written by a model” one.
Free score. Pay-per-use credits. No login. Crypto accepted.
Anonymize →How it works
1 · Learn your style
Paste writing you've already done. Our stylometry judge measures your rhythm, punctuation, word choice and habits, about 155 signals in all, and picks the samples that best represent you.
2 · Rewrite any draft
A frontier model rewrites your AI draft using your style profile and samples. The judge scores each draft, and only what's off gets revised, in at most 3 passes. Meaning is checked, and numbers, names and quotes never change.
3 · Learn from your feedback
Edit and accept, 👍 what reads like you, 👎 what doesn't. Accepted text and 👍 passages become examples of your style, and 👎 passages are steered away from. The style keeps getting closer to you.
Your writing stays yours
- Accounts are a random number. No email is required, and we store only a salted hash.
- Your Style is used only for you. We never train shared models on it without opt-in, and you can export it or delete it in one click.
- Drafts are never stored. Only text you accept is kept, to teach your style, and you can turn that off.
- For writing that must not point back at you, Anonymize keeps your fingerprint in your browser only.
- Card or BTC/USDC through a hosted processor.
Read the threat model. It says plainly what we protect against and what we don't. Read the benchmarks, measured on held-out authors and on models the judge never saw in training.
What we don't build: a “humanizer” marketed for slipping past AI detectors. Model-likeness is a writing-quality signal. We don't sell against detectors, and we don't model a real person's voice without their consent.