Skip to content
BookFoundry
Detect AI text. Make it sound human.

AI Detector & Humanizer

Paste any text or upload a PDF, DOCX, or EPUB and get a sentence-by-sentence AI verdict with confidence scores and model attribution. Then rewrite flagged passages in a natural human voice that keeps your meaning intact. Free to try, no credit card.

700+
books generated
600+
writers on the platform
50+
book languages
DOCX ยท EPUB ยท PDF
KDP-ready exports

Live platform totals, verified

0 / 80

Paste at least 80 characters (about 15 words) to run a free check.

Your text is analyzed for this check and not stored. Free, no account required.

Why it works

Sentence-level AI detection

Not one vague percentage. Every sentence gets its own verdict, confidence interval, and likely source model, so you know exactly what to fix.

Fair to non-native writers

Most AI checkers flag fluent second-language writing as machine text. Our fairness layer adjusts for non-native patterns before judging.

One-click humanizer

Rewrite flagged text with 10 author personas and a writing fingerprint built from your own document, then export as PDF, EPUB, or DOCX.

How it works

  1. 1

    Paste or upload

    Drop in raw text or a full manuscript as PDF, DOCX, or EPUB.

  2. 2

    Read the verdict

    See which sentences read as AI, how confident the model is, and why.

  3. 3

    Humanize and export

    Rewrite in a human voice and download a clean file, ready to submit or publish.

How the AI detector works

Instead of one black-box percentage, the detector runs several independent layers and combines them. A statistical layer measures burstiness and sentence-length variation, the signals that separate human rhythm from model-smooth prose. A pattern layer looks for the giveaway phrasing that large language models overuse. A language-model judge reads the passage the way a careful editor would and estimates how likely each sentence is to be machine-written. Because the layers are independent, a text has to fool all of them at once, which is far harder than beating a single score. The result is a stable verdict you can trust, not a number that swings every time you paste the same paragraph.

Which AI models it detects

The detector is trained on the writing patterns of the major model families, so it recognizes text from GPT-style, Claude-style, Gemini-style, and Llama-style systems, and it reports the family most likely to have produced your text. Model makers change their systems often, so no detector can promise perfect coverage of every version. What stays reliable is the family-level signature, and that is what the report shows you, alongside a confidence level so you always know how sure the verdict is.

Accuracy and false positives

The honest answer no free tool likes to give: any AI detector can be wrong, and the mistake that hurts most is calling human writing machine-made. Fluent second-language writing is the classic trap, because it can look unusually clean. This detector runs a fairness layer that adjusts for those patterns before it judges, so careful non-native writing is far less likely to be flagged unfairly. Every verdict ships with a confidence level and a sentence-level view, so you are never handed a single accusing number with no way to see why.

Scan PDF, DOCX, and EPUB files

You do not have to copy and paste. Upload a PDF, a Word DOCX, or an EPUB and the detector reads the whole document, returns a verdict for every section, and shows you exactly which passages read as AI. This is built for real work: a full essay, a chapter, a manuscript, or a report, checked in one pass instead of a paragraph at a time. When you want to fix what it finds, the one-click rewrite turns flagged passages into a natural human voice and exports a clean file, ready to submit or publish.

Who uses it

Students check their own drafts before submitting, so a false accusation never catches them by surprise. Writers and bloggers confirm their work reads as human before it goes live. Editors and publishers screen submissions at scale. Authors preparing a book for Amazon KDP run the whole manuscript through before upload. In every case the pattern is the same: check first, on your own terms, then fix anything that reads as AI while it is still private.

Frequently asked questions

Yes. Paste a passage on this page and get an instant verdict with no account and no credit card. A free account adds longer scans, file uploads, and the sentence-by-sentence view; larger documents and bulk rewriting use credits.
The detector recognizes patterns from major model families including GPT-style, Claude-style, Gemini-style, and Llama-style output, and reports the most likely source with a confidence level.
Yes. Sign in and upload a PDF, DOCX, or EPUB and the detector reads the entire file, returns a verdict per section, and flags the exact passages that read as AI, so you can check a full essay, chapter, or manuscript in one pass.
It combines several independent layers for a stable result, but no detector is perfect. The most common error is flagging fluent human writing as AI, so a fairness layer adjusts for non-native patterns before judging, and every verdict comes with a confidence level and a sentence-level breakdown you can review.
Yes. The rewrite changes tone and rhythm, not facts. It builds a fingerprint from your own writing so the result still sounds like you.

More from BookFoundry

Book Formatter & Manuscript Workshop

Already written it? Bring it in.

Learn more

AI KDP Book Generator: Publish & Earn on Amazon

Turn your expertise into a book that sells.

Learn more

AI Novel Writer: Write a Full Novel with AI

Novels in your imagination, in your voice.

Learn more

Ready when you are

No setup, no learning curve. Start in your browser and hold a finished book today.

Scan your text
50+ languages DOCX + EPUB + PDF Free to start