Prompt Generator

AI Detector

Paste text and see which sentences read as AI-written, with a reason for each and an overall range. An estimate you can check, never proof. Free, no sign-up.

0 / 6,000

An estimate per sentence, never proof.

Draft sounds generated? Rewrite it with the AI humanizer

Related writing and prompt tools

View more
AI HumanizerRewrite a stiff AI draft so it reads natural and sounds like you.
Prompt CheckerScore a prompt out of 100 and see what each part is missing.
Prompt OptimizerPaste a prompt and get a clearer, shorter or structured version.
AI Prompt GeneratorTurn a rough idea into a full prompt for ChatGPT, Claude, Gemini, DeepSeek and Grok.
ChatGPT Prompt GeneratorRole, task, context and a clear output format.
Image to TextCopy the text out of a screenshot or photo, as plain text or Markdown.

Before and after

An authored example: a blog intro built from stock phrases with one line a person clearly wrote, the labels a check gives it, and the same text after the humanizer.

A blog intro on remote work

  • Likely AI
  • Mixed
  • Likely human

In today's digital age, remote work has become an integral part of the modern workplace. (Likely AI) It is important to note that this shift offers numerous benefits for both employers and employees. (Likely AI) Furthermore, it allows companies to tap into a myriad of talent across the globe. (Mixed) I tried it for a year and missed the office coffee more than I expected. Moving forward, organizations must foster a culture of trust to unlock the potential of distributed teams. (Likely AI)

Estimated share written by AI

55% to 85%

Reads mostly as AI writing

  • Likely AISentence 1Stock opener that could start any article on the subject.
  • Likely AISentence 2Filler lead-in, then a broad claim with no named benefit.
  • MixedSentence 3Stock transition and phrase around a concrete point.
  • Likely AISentence 5Three stock phrases in one sentence and no specific action.

After the AI humanizer

Remote work is normal now, and it suits plenty of employers and employees. Companies can hire people from anywhere, not only within commuting distance. I tried it for a year and missed the office coffee more than I expected. If teams are going to work apart, managers have to trust them to get on with it.

How the estimate is made

A model's reading and a few plain measurements, shown side by side so you can see where each label comes from.

A model reads each sentence
Every sentence gets a number and one of three labels: likely human, mixed or likely ai. Flagged sentences come with a one-line reason you can check yourself.
Code measures the text
How much sentence length varies, which stock phrases appear and how often sentences start the same way. These are shown separately, so you can see what the model was told.
A range, not a score
The overall estimate is the share of words in flagged sentences, widened to cover the model's own guess and never narrower than 20 points. Short texts get a wider range.

Read the result with care

Every AI detector makes mistakes. These are the ones to keep in mind before you act on a result.

Never the only evidence

This is an estimate, not proof. AI detectors, this one included, get it wrong in both directions: they flag careful human writing, especially from non-native speakers, and miss AI text that someone edited. Never use a result as the only basis for accusing anyone.
01
People get flagged
Formal, formulaic or carefully edited writing, and writing by non-native speakers, often reads as AI to detectors.
02
Edited AI text slips through
A draft someone rewrote by hand, or ran through a paraphraser, can look entirely human.
03
Short texts say little
Below a few paragraphs there is not enough to judge. The tool needs 150 characters and widens the range on short texts.
04
Different tools disagree
Two detectors can give the same text opposite results. No detector, this one included, has been shown to be reliable enough to settle who wrote something.

How to use the AI detector

Three steps from pasted text to sentences you can fix.

  1. 11. Paste the textA paragraph or more, up to 6,000 characters. Longer text gives a steadier estimate.
  2. 22. Read the highlightsSentences that read as AI turn red, mixed ones amber. Each flagged sentence is listed with its reason and any stock phrases found in it.
  3. 33. Edit, or humanizeRewrite the flagged sentences yourself, or send them to the AI humanizer, which concentrates on exactly those sentences.

Frequently asked questions

What people ask before they trust a detector.

1How does this AI detector work?
Two layers. A language model reads the text sentence by sentence and labels each one likely human, mixed or likely AI, with a short reason for every flagged sentence. Code in parallel measures sentence length variation, stock phrases and repeated sentence openers. The overall range combines the sentence labels with the model's own estimate.
2How accurate is it?
Not accurate enough to prove anything, and we do not quote an accuracy figure because we have not measured one we would stand behind. Treat it as a second reader pointing at sentences worth a look, not as a verdict. The reasons are there so you can judge each one yourself.
3Can I use the result to accuse a student or a writer?
No. A detector result, from this tool or any other, must never be the only basis for accusing anyone of using AI. Detectors regularly flag people who wrote every word themselves. If you have a concern, talk to the person, look at their drafts and notes, and follow your school's or organisation's process.
4Why is my own writing flagged as AI?
Because the signs detectors look for, such as even sentence lengths, common transitions and general statements, also appear in careful human writing. Formal essays, reports, templates and writing in a second language are flagged most often. A flag means the sentence reads like typical AI output, not that AI wrote it.
5Why a range instead of one number?
One number suggests a precision no detector has. The range shows how uncertain the estimate is: it is never narrower than 20 points, wider on short texts, and wider again when the sentence labels and the model's overall guess disagree.
6What does Humanize these sentences do?
It opens the AI humanizer with your text already in the box and the flagged sentences marked, so the rewrite concentrates on them. Use it to make a draft read naturally, not to hide AI use where you are expected to disclose it.
7Is it free?
Yes. You get 10 free checks a day on each device without an account, and 20 a day with a free account. The detector shares that allowance with the other prompt tools, and paid plans remove the daily limit.
8Do you keep the text I paste?
The sentences are sent to our AI model provider to be rated and are not shown to anyone else. They are not saved to your account or published anywhere. The length and stock phrase counts are made in your browser.

See which sentences read as AI

Free, 10 a day, no sign-up. Paste a paragraph or more and read the reasons.

Paste text