AI text detector

Estimate with suspicious sentences highlighted

No database: AIMRAN doesn't store your files, text or results

What is AI text detector?

This detector analyzes a text and estimates how likely it is to have been written by an artificial intelligence such as ChatGPT, Gemini or Claude. Besides the number, it highlights the sentences that most resemble AI writing and explains why.

It's a helpful aid for reviewing assignments, articles or client copy, not definitive proof: no detector is always right. A very formal human text can score high, and well-edited AI text can score low.

How to use it

  1. Paste the text you want to analyze (ideally more than 100 words).
  2. Check the estimated probability and the level: low, medium or high.
  3. Review the signals in the side panel, sorted by weight.
  4. Tap or hover over the highlighted sentences to see the reasons.

Advantages

Technical details

It's a heuristic detector, not a language model. It combines ten stylometric signals on a logit scale, with weights calibrated on sample texts in Spanish and English: variation in sentence length (burstiness), filler phrases and expressions typical of AI, repeated sentence openings, lists of three items, use of em dashes, uniform punctuation, lack of personal markers, formal register, lexical diversity and chatbot formatting (headings, bullet points, bold). Each sentence gets its own score; sentences are highlighted in orange from 0.25 and in red from 0.5. With fewer than 60 words the result is marked as unreliable, and texts under 20 characters are not analyzed.

Frequently asked questions

How can I tell if a text was written by ChatGPT?

Paste it into the detector and review the probability and highlighted sentences. Use it as a clue alongside your own judgment, not as proof.

Is this AI detector reliable?

It gives a rough estimate. No detector is 100% reliable: very formal human writing can produce false positives, and edited AI text can produce false negatives.

What signals give away AI-generated text?

Sentences of very even length, filler phrases like “in summary” or “it's worth noting”, lists of three, lots of em dashes, an impersonal tone and chatbot formatting.

How many words does the detector need?

It works from just a few sentences, but the result is more reliable from 100 words. Below 60 words you'll see a warning that it's unreliable.

Which languages does it support?

English and Spanish. It's calibrated for both and detects the language automatically.

More Text & AI tools