The writer stays the author.
You can feel it in a paragraph: the over-explained lesson, the tidy ending, the story that never touches the real world. BylineProof breaks the shape the only honest way: it works like an editor, not an author. It measures your voice from your real writing, asks for the story only you can supply, and drafts from nothing else. You bring the story. It never invents one.
A Claude plugin from Diakryon
Your voice. Your story. Then the plan. Then the review. In that order, on purpose.
Setup takes five minutes and three to five samples of your real writing. From those it builds a measured profile: your rhythm, your phrases, your habits on the page. Not a style dropdown. Yours.
Ask it to make something up and it stops and asks you back: what actually happened? A name, a date, a place. It works like the editor good publishers assign. A missing detail is a question for you, never an invention.
Humanizer workflows start after the draft: generate first, scrub second. BylineProof never scrubs. It plans every draft around your approved voice profile, then shows you exactly where the result still drifts. Every flag ends with a choice, and KEEP AS WRITTEN is always one of them.
Paragraph by paragraph, against the public research into AI writing patterns, and against your profile, so a choice you made on purpose is never reported as a defect.
Not a score. A conversation about your page, one flag at a time, with you holding the pen.
The first time I read my own chapter back, I didn't recognize the person talking. The sentences were fine. That was the problem.
In the end, the experience taught me a valuable lesson about perseverance, and I knew that whatever challenges lay ahead, I would face them with newfound confidence.
My editor circled one line and wrote two words in the margin: whose voice?
A lesson stated outright, a future pre-solved. Your profile says otherwise: your samples end on an open question or a hard image, not a moral. Want to close the way you actually close?
Interface shown for illustration of the review flow.
AI providers now mark generated text.
Products that erase the trail are the reason the marks exist. BylineProof does not remove AI-provenance marks. Not quietly, not on request, not ever.
BylineProof documents your process (your direction, your review, your decisions) so the mark tells the truth and so do you.
If your process is honest, you don't need the mark removed. That's the whole point of BylineProof.
These tools reduce recognizable AI patterns. They do not guarantee any AI-detector score. Detection tools have real error rates, and human writers use many of these patterns too. The only durable test is the one BylineProof ends every review with: would a stranger reading this trust the person who wrote it?
BylineProof is available now on Gumroad: $69 one time, and every update is free.
Claude. BylineProof installs as a Claude plugin (Customize, then Plugins) and works in Cowork with a connected folder or in Claude Code. Type /bylineproof to start. Setup runs once.
It drafts with you, from your voice profile and the story you give it. Ask it to make something up and it asks you what actually happened instead.
No. BylineProof never removes AI-provenance marks, and it doesn't promise any detector score. The honest stance and the honest limits above say why.
Always. Every flag ends with a choice, and KEEP AS WRITTEN is one of them, every time.
$69 one time. Every update free. Install it in Claude, type /bylineproof, and setup runs once.
BylineProof is a Diakryon product. Diakryon is the product studio of Dennis Winten, who also runs Bible Morning, and a trade name of Bible Morning LLC. More about Diakryon. Questions: info@diakryon.com
Also from Diakryon: PRIME, turning Claude into an assistant with a great memory.