Free AI detector for GPT-6, Gemini, Claude, Llama, and more.
Check text, PDFs, or web pages for AI writing. See which sentences get flagged, the patterns detected, and how confident the result is.
0 / 1,500 words
Your report will appear here
How the AI detector works
An AI detector is a classifier that estimates the probability that text was written by a language model rather than a person. It works by measuring statistical patterns in the writing, not by tracing where the text came from.
Give it something to read
Paste text, drop in a PDF or Word file, or enter the address of a web page. Documents are opened in your browser, so the file itself never leaves your device.
The text is scored
A detection model trained on human writing and on output from AI systems rates the whole text and each sentence for how closely it matches machine writing.
You see the evidence
The report marks every sentence, lists the writing patterns found, and tells you how much confidence the result deserves.
A score alone tells you little. This one shows its evidence.
In today’s fast-paced world, a good night’s sleep plays a crucial role in overall well-being. I learned that the hard way in March, when three weeks of 5 a.m. starts left me crying over a dropped spoon. Experts recommend keeping a regular schedule, limiting screens, and avoiding caffeine late in the day.
Sentence 1 of 3 · 15 words
Reads as AI-like (94%)
- Uses “In today’s fast-paced world”, a phrase that turns up far more in AI text
- Uses “plays a crucial role”, a phrase that turns up far more in AI text
- Sentence-level marks
- Each sentence is scored separately. Solid blue underline for AI-like, dashed amber for mixed, unmarked for human-like, so the marks read without relying on colour.
- Patterns detected
- Even sentence length, stock phrases, formulaic transitions, missing contractions. We list what we found and say plainly that these are patterns a reader could spot, not the model's reasoning.
- Confidence scores
- Short text, a borderline score or a less-supported language each lower it, with the reason given.
- Share findings
- Create a link that shows the scores only, or the full highlighted text if you choose. Or download a dated PDF.
Who uses an AI checker
Anyone who has to decide how far to trust a piece of writing before they use it, grade it, publish it, pay for it, or put their name on it.
Teachers
Decide which submissions deserve a closer look, and bring specific passages to the conversation with a student.
AI detector for teachersStudents
See how an essay reads to a detector before handing it in, and keep a dated report.
AI essay checkerEditors and publishers
Review freelance and agency copy before paying for it, or audit pages that are already live.
AI content detectorWriters
Find the sentences where an assistant's phrasing survived your edit, and rewrite them in your own voice.
ChatGPT detectorRecruiters
Check cover letters and written answers, bearing in mind that many candidates use AI to polish honest applications.
How accurate is AI detection, honestly?
AI detection is accurate enough to raise a question, but not accurate enough to prove an answer. The best detectors catch most unedited AI text with few false alarms, but accuracy drops sharply when text is paraphrased, lightly edited, or written by a non-native English speaker.
Key takeaways
- Accuracy claims of 99% come from the vendors’ own benchmarks.
- One 2026 test of five detectors put the top two at 72% and 74% overall accuracy.
- False positives fall hardest on non-native English writers.
- A score is a probability estimate, not evidence of misconduct.
Reading the score
| Score | Verdict |
|---|---|
| 0–14 | Human-writtenLittle or nothing resembles AI writing. |
| 15–34 | Likely human-writtenMostly reads as a person, with a few AI-like passages. |
| 35–69 | UncertainMixed signals. Not evidence either way. |
| 70–89 | Likely AI-generatedMostly reads as AI, possibly edited by a person. |
| 90–100 | AI-generatedStrong, consistent AI signals throughout. |
A score in the Uncertain range is reported as Mixed when the sentences back that up, with at least 25% of the words in AI-like sentences and at least 25% in human-like ones.
Supported languages: English, French, Spanish, Portuguese, German, Dutch, Polish, Romanian, Italian and Simplified Chinese.
Questions about AI detection
How does the AI detector work?
Your text is scored by a detection model trained on large amounts of human writing and of output from AI models. It estimates, for the whole text and for each sentence, how closely the writing matches what language models produce.
Separately, we scan the text for surface patterns that tend to accompany AI writing, such as even sentence length, stock phrases and formulaic transitions, and list what we find. Those patterns are shown as supporting evidence. They do not change the score.
How accurate is it?
No AI detector is always right, and anyone who tells you otherwise is selling something. Detection works best on a few hundred words or more of unedited text. It gets less reliable on short passages, on text that has been paraphrased or heavily edited, and on very formal writing.
That is why every result comes with a confidence level and why the middle of the scale is labelled Uncertain instead of forcing a yes or no. Treat the result as one piece of evidence, never as proof.
Can it detect ChatGPT, Claude and Gemini?
Yes. The detection model is trained to recognise text from the major AI systems, including ChatGPT, Claude, Gemini and LLaMA-based models. It does not tell you which model wrote a passage, only how likely it is that a model did.
Why was my own writing flagged as AI?
False positives happen. Writing that is formal, evenly paced and free of personal detail can look statistically similar to AI output. Research has repeatedly found that non-native English writers are flagged more often than native speakers, for the same reason.
If you wrote the text yourself, a high score does not mean you did anything wrong. Open the flagged sentences to see what was noticed, and keep your drafts, notes or version history: they are far better evidence of authorship than any detector.
What does the percentage mean?
It is the model's estimate of how likely the text is to be AI-written, from 0 to 100. It is not the share of the text that is AI. For that, look at the breakdown bar, which shows what proportion of the words sit in sentences that read as AI-like, mixed or human-like.
What are the highlighted sentences and the patterns list?
Sentences with a solid blue underline scored as AI-like; a dashed amber underline means the sentence could go either way. Select any sentence to see its score and the patterns found in it.
The patterns list covers the whole text: sentence rhythm, stock AI phrasing, formulaic transitions, lists of three, contractions, personal voice and paragraph shape. Each is something a careful reader could spot too.
Is it free? Are there limits?
It is free and needs no account. Each check covers up to 1,500 words; longer text is cut to that length and the report says so. There is a daily allowance per visitor, and on very busy days the per-check limit may be lower.
Which files and links can I check?
Upload a PDF, Word document (.docx), OpenDocument text (.odt), RTF, Markdown or plain text file, or paste the address of any public web page and we will pull out the article text.
Scanned PDFs and photos contain no selectable text, so they need to go through OCR first. Our free PDF to Text tool does that.
Is my text stored or used for training?
Files are read in your browser; only the extracted text is sent to our server. We keep the scores for a while so repeat checks are instant, but not the text itself.
The one exception is a share link. If you create one and choose to include the text, we store it so the link works. Links expire after 90 days and you can delete yours at any time.
Which languages are supported?
English, French, Spanish, Portuguese, German, Dutch, Polish, Romanian, Italian and Simplified Chinese. Detection is most reliable in English. The phrase-level patterns are checked for English text only; for other languages you still get the score, the sentence highlighting and the rhythm analysis.
Can AI text be edited to avoid detection?
Often, yes. Paraphrasing, heavy editing and so-called humanizer tools all lower detection rates for every detector on the market. A low score therefore does not prove a person wrote the text, just as a high score does not prove a model did.
Can a teacher or employer use this as proof?
No. A detector result alone is not sufficient evidence of misconduct, and it should never be the only basis for a grade, a disciplinary decision or a hiring decision. Use it to decide where to look more closely, then talk to the writer and review their drafts.