Find out instantly if it's human orAI-written.

Pontbot is a software text checker. For pasted or uploaded text it returns a statistical probability of how closely the writing resembles text generated by artificial intelligence — for example ChatGPT, Claude or Gemini. It is a support tool for teachers, editors and students: not an expert opinion, not evidence, and not a substitute for an institution’s, examination board’s or employer’s own process.

The engine looks at linguistic patterns: how predictable the wording is, how even the sentence rhythm is, and how typical the formulaic, machine-like turns of phrase are. The result is a probability that the text looks AI-like — not a finding of which specific model wrote it. The method can be wrong, especially on short, heavily edited, translated or formal academic text.

No. Detection is a statistical estimate, not proof and not an expert opinion. Human text can be flagged as AI, and AI text can look human — especially after a paraphraser or humaniser. Results must always be interpreted with human judgement. Pontbot does not take automated decisions about people, and a score on its own is not a basis for disciplinary, examination or employment proceedings.

You need to sign in. Paste the text, or upload a .txt, .docx, .md or .pdf file, then start the analysis. On registration you receive 3000 character credits; a free scan analyses at most 1000 characters. You can download the report as a text (.txt) or JSON file. Larger allowances and extra detectors are on paid plans. Re-running the same unchanged text within 24 hours does not use extra credits.

Pontbot looks for AI-like linguistic patterns common to many large language model outputs — including styles associated with OpenAI, Anthropic, Google and other publicly used models. It does not reliably name which model wrote the text, and it does not warrant that every current or future model version will be recognised. Models change continuously, so the result is always indicative.

The in-house engine is calibrated for Hungarian and English. The interface is also available in German, Spanish, French and Polish, but detection in those languages is indicative, not calibrated. Short text, machine translation and non-native writing increase the chance of a false flag.

Submitted text is stored in your account history so you can open it later. We do not use it to train our own or any third-party model, we do not sell it, we do not put it in a plagiarism database, and we do not hand it to an institution. Deleting your account deletes your texts. Details are in the Privacy Notice.

As a support tool, yes; as a decision, no. Pontbot is not an educational, assessment or legal service: it does not prove authorship, it does not certify originality, and it does not excuse you from your institution’s rules. Using the service to breach the rules of an institution, exam body or employer — including trying to evade a detector — is prohibited. A “human” score does not prove that you wrote the text yourself.

Formal, formulaic or very regular academic style, short text, machine translation and non-native writing can resemble AI patterns. That is not an accusation and not evidence. Read the sentence-level explanation and decide — or let your teacher decide — based on the full text and the circumstances.