Polly don't need no cracker  |
Imagine you spent years on a book (I did, so I don’t need to imagine). Notes. Dead chapters. Arguments with yourself at 2 a.m. You send the manuscript in.
Somebody runs it through a detector. Comes back: 87% AI. What a load of shit.
Now what?
Surprise — that number does not know who wrote the book. It never saw your abandoned drafts. It didn’t sit in the room while you killed a character you loved. It looked at the finished page, compared it to machine patterns, and coughed up a percentage that sounds like science.
PublishingState put it plain this month: publishers need better evidence than an AI score. Been thinking about that all week while the big houses scrambled over disclosure. OUP found people hiding tools because stigma is real. Springer said ask how, impact, risk, oversight. And then there’s this other instinct in the industry — upload the file, wait for the score, and use it as a hangman’s rope.
That’s detector cult. We’ve been building Circle against that too.
A probability is not authorship. Detector companies will tell you that themselves if you read the fine print. Human writing can look “too clean” after edits. Second-language prose trips alarms. Genre formulas look like models. Meanwhile somebody who generated half a book and sanded it smooth might skate. So the honest author gets the scarlet percentage, and the ghost pipeline keeps humming.
Watermarks? Helpful when they exist. Absent when they don’t. No watermark proves nothing. C2PA content credentials? Fancy receipts for a file’s journey — useful, not a mind reader. Still a human has to decide what the trail means.
Here’s the little-world version for GAPOZ and Circle.
We are not going to gate the workshop on a magic percentage. We already said it: disclose by role. What did the tool touch? Who owns the take? Sign the sentence that says the judgment is yours — or you’re still in the draft room with a model.
That’s process evidence. It’s ugly and human and it scales to a small room better than a black-box score. If Circle goes public-facing soon — and yeah, I’m still aiming that direction — the point isn’t to catch you with a detector. It’s to make hiding look pointless because honesty is the cheaper path.
That entire paragraph above was suggested by AI. I like it. I left it just as it was written. Why? Because it says what I wanted to say, and it was a lot faster for AI to do than me. I thought AI was supposed to help us be more productive? To accomplish things that we want to do but didn’t have the time for. To increase our potential and capabilities, not deflate us. People have had secretaries for years and nobody had a problem with that. Now, I wonder how this piece would be scored? I wonder, but I don’t care. I know the work I do. I know what won’t exist at all unless I drive it. And I know what I like. I don’t read things and judge them on who wrote it or how. I guess I’m weird like that.
Big publishers can update warranties and ask for drafts and version history. Fine. Do that. Just don’t pretend a dashboard percentage is the same thing as a person standing behind a page.
An 87% is a signal to look closer. It is not a verdict. And it damn sure isn’t a substitute for a name on the line.
— Richard Crossley
GAPOZ Press · Circle
Also on the beat
- Who wrote the book? — why an AI score can’t settle authorship; process evidence over percentages — https://publishingstate.com/who-wrote-the-book-why-publishers-need-better-evidence-than-an-ai-score/2026/
- YouTube Get Feedback on Shorts — optional pre-publish coach on hook/pacing (U.S. 18+ mobile); suggestions, no earned credit — https://www.searchenginejournal.com/youtube-gives-you-feedback-on-shorts-before-publishing/589088/
- Ask Studio — YouTube’s AI assistant in Studio expanding; channel-specific ideas from your history — https://miraflow.ai/blog/youtube-ask-studio-ai-assistant-creators-explained-2026
- YouTube free AI Remix / Create — Gemini-based remix tools out of beta for more creators; labels and opt-outs in the mix — https://aivideoadvisor.com/youtube-free-ai-remix-create/
- Originality still decides who earns — AI can remove the skill barrier; it doesn’t remove the work of deciding (faceless-channel beat) — https://www.tribuneindia.com/news/business/rudra-pratap-singh-rudy-highlights-why-originality-matters-in-ai-assisted-youtube-content/
How this was made
AI assisted research and a first draft of this Circle View. Richard Crossley reviewed, edited, and is accountable for the judgment and the final take. Disclosure by role — not undetectable.
