The AI Detector that shows its work.
Paste any text and get an AI likelihood score in seconds, with a plain-English verdict and an honest account of what the score can and cannot prove.
Scores 40 to 69 are inconclusive. Never punish on a score alone.
Three steps to a reading
Paste your text
Drop in 120 to 6,000 characters: an essay, article, email or cover letter. No account, no upload, no waiting room.
The detector reads it
The engine measures the statistical fingerprints of machine writing: predictability, uniform rhythm, low variation. It takes about three seconds.
Get an honest verdict
A 0 to 100 reading with a plain-English verdict, including an explicit inconclusive band. We tell you what the score cannot prove.
An honest reading, including the failure zone
Every AI detector, ours included, is a statistical instrument. It measures how predictable and how uniform a text is, because machine writing tends to be smoother and more even than human writing. That measurement produces a probability, and a probability has a gray zone. We put the gray zone on the dial instead of hiding it: scores from 0 to 39 lean human, 70 to 100 lean AI, and everything from 40 to 69 is inconclusive on purpose.
That design choice matters most when the stakes are real. A student accused on the strength of one number, a job applicant filtered by a screening tool, a writer whose careful prose happens to be tidy: these are the people false certainty hurts. So the verdict chip on this site will tell you when it does not know, and the copy around it will tell you what to check next. If you want the mechanics, read how AI detectors work. If you have been wrongly flagged somewhere, start with false positives and what to do about them.
Published research shows detectors flag non-native English writers at elevated rates. Formal, careful writing reads machine-like to a statistical model. If you grade, hire or edit using any AI checker, treat the score as a question to investigate, never an answer.
Built for four desks
Teachers
Pre-screen suspect submissions without buying an institutional license, then follow the honest playbook: a score opens a conversation, drafts and version history close it. Start at the detector for teachers.
Editors
Triage incoming drafts before you spend an hour line-editing one. A high reading means look closer for generic phrasing and invented facts. The content detector page covers the editorial workflow.
SEO managers
Audit content at the draft stage, where fixing voice is cheap. Google rewards helpful content however it is made, so use the score as a quality smell test, not a compliance gate.
Hiring managers
Get a read on whether a cover letter was machine drafted. Then weigh it like any other signal: a polished letter from a strong candidate is not a crime, and the writing detector page explains how to read borderline scores.
What makes this AI detector different
The AI checker market runs on certainty theater: precise-looking percentages, accuracy claims no independent test reproduces, and results gated behind email walls. This site is built on the opposite bet. The detector is genuinely free, asks for nothing, stores nothing, and reports its readings with the inconclusive band drawn in. You get the same engine on every scan, whether it is your first or your fortieth.
The instrument itself stays focused: text in, reading out. No browser extension demanding permissions, no upsell mid-result. When the reading is uncertain, the verdict says so. When the reading is strong, the copy still reminds you what a statistical score can and cannot prove about a human being. That is the work an honest detector owes you, and it is why the tall tick on our dial sits at the threshold where caution begins.
Run your first scan.
Paste a text you are curious about and read it the way an examiner would.
Free. No account. Nothing stored.Frequently asked
Is this AI detector really free?
Yes. Scans up to 6,000 characters are free, with a daily fair-use limit so the tool stays fast for everyone. No account, no credit card, no trial countdown. If you need higher volume, see the pricing page for the current state of paid plans.
How accurate is the AI detector?
No AI detector is fully accurate, including this one. Scores from 0 to 39 lean human, 70 to 100 lean AI, and anything between 40 and 69 should be treated as inconclusive. We explain the mechanics and failure modes in how AI detectors work.
Do you store the text I check?
No. Your text is sent over an encrypted connection, analyzed in memory and discarded. There is no text archive, no training on your data, and no account linking a scan to you. Details are in the privacy policy.
Can an AI detector prove someone cheated?
No. A score is a probability reading, not evidence of misconduct. Treating a high score as proof harms real people, because false positives happen, especially on non-native English and heavily edited text. We cover this in AI detector false positives.
Which AI models does it detect?
The detector is trained on statistical patterns common to modern large language models, including text in the style of ChatGPT, Claude and Gemini. It scores the writing, not the brand: a paraphrased or heavily edited AI draft will read closer to human.
What is the minimum text length?
120 characters. Below that, the statistical signal is too thin to mean anything, and tools that score a single sentence are guessing. For best results paste 200 words or more.