No one can reliably tell whether a piece of writing came from a person or a machine.
People have tried. The detectors that promise to flag AI-generated text get it wrong often enough that no serious institution will treat their output as evidence. They tag a human’s essay as machine work. They wave a machine-written ad copy through as the real thing. Academia stopped trusting them a while ago. And here’s the uncomfortable part: when telling the two apart becomes impossible, the burden of proof quietly turns around. The question is no longer whether someone can prove you used AI. It’s whether you can prove you didn’t.
This May, that question grew teeth.
The Academy laid down new rules for the 2027 Oscars: only human performances can compete for the acting awards, only human-written scripts can compete for the writing awards, and the Academy reserves the right to investigate whether any film leaned on generative AI. Its president, Lynette Howell Taylor, put it plainly. “Human beings must be at the center of the creative work.”
There’s a blank space in that rule. What does “human-written” actually mean? A screenwriter who used AI to look something up, smooth out one line of dialogue, run a spell check — does that count? The Academy didn’t say. It set the gate and kept the right to ask questions later. For the writer who spent three years on a script and typed every word of it by hand, that blank space is the awkward part. Your conscience is clean. But how do you show it?
This reaches well past Hollywood. Literary prizes have begun asking winners to attest to how the work was made. Publishers are writing AI clauses into contracts. Grant committees, submission portals, freelance clients — one after another, they’re putting the same question on the table. Did a person write this? Can you prove it?
Proving a negative is a losing game
The trouble starts with the shape of the claim. It’s a negative.
“I didn’t have AI write this for me” — you can’t establish that by inspecting the finished product. A clean, polished draft and an AI-generated draft that a person tidied up look identical lying side by side on a desk. The final file doesn’t talk. You can stare at that last PDF all day and learn nothing about where it came from.
What talks is the process.
Think about how a painting proves whose hand made it. Not the finished canvas — anyone can copy a finished canvas. What proves it is the stack of things behind it: the loose sketches, the composition reworked again and again, the dated photographs from the studio, the receipts for paint. No single item is a smoking gun. Pile them together and they tell a story that’s very hard to fake — that this painting grew under one person’s hand, stroke by stroke, over time.
Writing has a stack like that too. It’s called version history.
You never write a book in one pass. Chapter one got torn down and restarted four times. The ending of chapter seven didn’t sit right until the twelfth attempt. There were afternoons when the words poured out and a whole week when you moved three hundred of them and called it progress. That crooked, uneven, heavily revised trail is precisely the thing a generative model can’t manufacture. The machine hands you a result with no process behind it — a finished object, no months-long curve of human struggle underneath.
Saving the process was always part of the job
This is what Slima has done since day one — back when nobody imagined it would one day double as evidence.
Before every major change, the system saves a snapshot automatically. Want to mark a milestone yourself? Save one by hand. Every version is kept on its own, every version can be restored, and the whole book’s growth is recorded along a single timeline. We built this so you’d feel free to make big, reckless edits — wreck a chapter, hit one button, and you’re back. But that same timeline now happens to be the most natural evidence that a person wrote the book.
One more thing makes the record sturdier. Slima’s AI works as a coach and nothing more. It reads your draft, asks you questions, points out the blind spots you can’t see on your own — but it doesn’t decide what your story is about, and it doesn’t stuff chapters of prose into your book. The words you wrote are yours. That sounds like a marketing line. It’s actually a design choice that changes how much your evidence is worth.
So we did something direct: we let you export that timeline as a document. The creation timeline, the way your word count climbed, the snapshots you saved by hand, the number of times the AI was involved. Laid out, for whoever needs to see it. You can sign your name at the end, add a statement, and spell out that the book is yours and what role the AI played in your process.
What it proves, and what it doesn’t
Let me be careful here, because overclaiming would hurt you more than it helps.
An Authorship Record is strong evidence. It is not proof beyond doubt.
What it can show: that the book passed through real time, through a real sequence of revisions, and grew into its current shape one piece at a time. What it can’t do is just as concrete. The timestamps come from our servers, not a third-party notary. It can’t read the intent in your head. And it won’t authenticate, line by line, exactly whose hand wrote each sentence. Any honest proof marks its own edges, and our report is written that way — it doesn’t inflate a thing on your behalf.
That’s enough. Most matters in a courtroom aren’t settled by one decisive, knockout fact. They’re settled by a stack of mutually reinforcing things that would be very hard to fake all at once, assembled into a plausible account. The Authorship Record does exactly that. It doesn’t hand you a verdict that says “certified true.” It hands you a credible, checkable, hard-to-invent account of how the work came to be.
The people this actually protects
It’s worth sitting with the question of who this mechanism really shields.
At a glance it looks like one more shackle. Writing a book used to mean writing the book well; now you have to keep receipts too. But turn it around. In a world where nobody can tell the real from the fake, the person who loses most is the one who actually spent three years and wrote every word by hand — because he has no better way than anyone else to prove he’s clean.
The Authorship Record tips that scale back. For the first time, “I really did write this” carries weight you can put on the table. For the corner-cutters, it does nothing — no process, no record to conjure from thin air. For the writer who put in the work, it turns those few hundred hours you already spent into something nobody can forge.
That blank space is going to hang there a while. The industry will argue for a long time about where the line for “human-written” sits. But before the dust settles, there’s one solid thing you can do right now: keep a good record of how you wrote this book.
Want to see how the document is actually generated, signed, and handed over? Here’s the step-by-step walkthrough: Generate your Authorship Record.