Paste a draft below. ZeroLLM marks the sentences that read as machine-written while you work, then rewrites only what needs it.
Marks appear once there are 30 words.
Your draft stays one document the whole time. ZeroLLM reads every sentence, marks the ones that sound machine-made, and rewrites only those, in the voice you choose.
Paste a draft or start typing. Once there are 30 words, every sentence gets a score while you write.
Moreover, leveraging asynchronous communication can unlock significant productivity gains across distributed teams.
Sentences that read as machine-made get a highlighter pass. Stock phrases get a wavy underline that explains itself.
Moreover, leveraging asynchronous communication can unlock significant productivity gains across distributed teams. 86% AI-like
Rewrite the flagged sentences one at a time or all at once, in the voice you pick. Every pass is kept, so you can compare or step back.
Moreover, leveraging asynchronous communication can unlock significant productivity gains across distributed teams. Writing things down and letting people reply on their own time does more than another meeting ever will. 12% AI-like
Three things decide a score, and the app shows you each of them, down to the sentence and the phrase.
A neural classifier fine-tuned on writing from the models people actually use today, alongside human writing on the same topics. It's retrained as new models ship.
Stock phrases, transition habits and suspiciously even sentence lengths. They nudge the score and, more usefully, explain it.
Every sentence gets its own score, so you see which lines read as machine-made instead of one number for the whole page.
The same detector and rewriter, over HTTP. Free accounts get API keys and the same monthly allowance as the website. Responses include the overall reading and a score for every sentence.
Read the API docscurl https://zerollm.app/api/v1/detect \
-H "Authorization: Bearer zl_live_…" \
-H "Content-Type: application/json" \
-d '{"text": "Paste at least 30 words of the text you want checked."}'{
"score": 0.82,
"label": "ai",
"confidence": "high",
"sentences": [
{ "start": 0, "end": 118, "score": 0.91 },
{ "start": 119, "end": 204, "score": 0.34 }
]
}Checking runs on our own hardware, so the allowance is generous. Rewriting runs on a language model, and that's the budget that matters.
$0No account needed
$0Sign in with Google
Coming soonPricing to come
Something else? Write to hello@radsoftinc.com.
ZeroLLM is in beta, and every score is an estimate. Formal, formulaic writing (proposals, abstracts, schedules) and text from people writing in a second language can read as machine-made when a person wrote every word. The detector is retrained as new models come out and improves every week.
No. A score is a reason to look closer, never evidence on its own. Don't use it as the only basis for a grade, a hiring decision or an accusation against anyone. Talk to the writer, look at drafts and context, and treat the highlights as a starting point.
If you're signed in, text you check or rewrite on the website is kept in your history for 30 days so you can reopen it, and you can delete one item or everything at any time. The checks that run while you type aren't saved. API requests store only word counts and scores, never the text. Rewriting sends your text to Anthropic's API; detection runs on our own servers. We don't sell your data.
It rewrites in the voice and strength you pick so the draft reads like your own writing: plainer words, varied rhythm and no stock phrases, with every fact, name and number kept. Rewrite flagged only touches the sentences the detector marked. Rewrite all redoes the whole draft, then checks the result and reworks anything that still reads as machine-made. Every pass is kept so you can compare or go back.
Yes. Free accounts can create API keys and call the same detector and rewriter over HTTP, with the same monthly allowance as the website. The API docs cover authentication, limits and every response field.
Read the API docs