See the flagged passages
Preds highlights sentences within the full text. You can see what the result refers to instead of guessing from a percentage.
We trained Preds on student essays too. Finished a Hungarian essay or assignment? Paste it in to see which sentences receive an AI signal.
Made for your kind of writing
The model saw Hungarian student essays during training. See which sentences it flags in your own work before you submit.
Preds highlights sentences within the full text. You can see what the result refers to instead of guessing from a percentage.
You do not need to create extra proof: notes and revision history can already show how you reached the final draft.
Some courses allow AI for brainstorming; others require you to disclose any use. Check your teacher’s or institution’s policy before submitting.

In practice
Preds shows an overall estimate and highlights passages in the text. Read them with the surrounding paragraphs, and check that your sources are accurate.
A signal is an estimate, not proof. A low result does not prove authorship, and a high result does not prove AI use.
More and more Hungarian universities and faculties set rules for AI use in a thesis: some require a declaration, others allow AI only for brainstorming. Check your faculty’s thesis regulations and your supervisor’s expectations before you submit.
No AI detector is infallible. If sentences in your own text are flagged, reread them and keep your drafts, notes and document revision history: they show how your thesis came together.
A plagiarism checker looks for matches with other sources; an AI detector looks for wording patterns typical of machine writing. One does not replace the other, and Preds is not a plagiarism checker.
Uploads travel over an encrypted connection. Original files are kept only while the text is extracted and then deleted; the extracted text is stored encrypted until it is submitted for checking.
Privacy policyOne credit covers up to 1,000 words, so a 15,000-word thesis takes 15 credits. The 2 free daily credits suit shorter excerpts; a full thesis needs the 100-credit pack (2,500 HUF) or a subscription.
PricingOn the public test dataset of 1,000 texts, Preds 3 reached 99.6% accuracy and did not label any human text as AI-written. The error rate is higher for short or heavily rewritten text.
Fresh articles about research, operations, and advice in the world of AI detection.
All articles
01Csábító azt gondolni, hogy mivel az LLM-eket az emberi szöveg modellezésére tanítják, a kimenetük megkülönböztethetetlenné válik az emberi írástól. A valóságban azonban az LLM-ek nem csupán megközelítik az emberi szöveget: utótanítással (post-training) olyan segítőkész asszisztensekké alakítják őket, amelyek követik az utasításokat, ez pedig egy szűk tartományra korlátozza az írásukat, amely felismerhető marad.
Read article
02A Preds 3 jelentős előrelépést képvisel a korábbi 2.0-s verzióhoz képest. A mérési eredmények kimagaslóak lettek: tovább csökkent a téves felismerések (fals pozitívok) aránya, miközben a detektálás pontossága érezhetően nőtt.
Read article
03Sok AI-detektor állítja magáról, hogy ők a legjobbak a szintetikus szövegek felismerésében. Ebben a témában szeretnék egy szemfelnyitó mérést és cikket bemutatni, amely megmutatja a valóságot.
Read article