01
Sentence-level transparency
Alongside the overall result, suspicious and human-classified sections are highlighted separately passage by passage.
Hungarian AI detector that actually works.
Check your Hungarian texts and see which parts were written by ChatGPT, Gemini, Claude, or another AI model.
Recognizes Hungarian text from leading models
ChatGPT
Claude
Gemini
DeepSeek
Llama
Grok
Mistral
Perplexity
Qwen
Z.ai GLM
Kimi
98+%Accuracy
<1%False positive rate
15+Recognized AI models
A real answer in seconds.
Choose from the examples, paste your own Hungarian text, or import a document. The result opens in the Preds dashboard.
Sentence-level annotations
TXT, PDF, DOCX, and Markdown documents
Review past analyses and downloadable PDF reports
1,396 characters
Not just a percentage. A usable answer.
Preds clearly shows which parts are suspicious in a text.
02
PDF, in original view
Upload a PDF document and view detection results directly in the original document.
03
Text from handwriting too
We digitize text from photographed assignments and notes, then detect suspicious parts in the same way.
04
Integratable into your own system
Automate large volumes of content verification with the Preds API. An easily integrated solution for any system.
POST /api/detect
{
"language": "hu",
"prediction": "human",
"confidence": 0.978
}Who is it for?
Speeds up verification across educational, editorial, and corporate environments.
Teachers and students
Fast, consistent screening of assignments, papers, and research materials in Hungarian.
Publishers and media
Verifying content authenticity before publication, with transparent results for the editorial team.
Enterprises and platforms
High-volume verification via web interface or API, tailored to existing workflows.
Preds is a decision-support tool: There is no perfect detector. Despite its reliability, the Preds model can make mistakes. Regardless of the detection result, the uploader should review the entire text and make a responsible, deliberate decision.
From Hungarian researchers - The best possible solution
Preds is built exclusively on Hungarian-language data. We continuously measure and improve our methodology and models.
Verified accuracy
Preds detects AI-generated content in texts with 98% accuracy. Measured on nearly 100,000 documents.
Data protection
Documents are encrypted during both transmission and storage. Uploaded files are not stored, only retained during text extraction.
Trained on diverse texts
The data on which the Preds model was trained was crafted to include as many domains as possible: news, legal and administrative texts, scientific and educational materials, community content, creative writing, as well as technical and financial texts.
Low false positive rate
False positive means flagging human text as AI-generated. This is a critical mistake, which is why our model was designed to keep this error minimal. Based on our tests, it only misclassifies 1 human text out of 1,000 in this direction.
Preds blog.
Fresh articles about research, operations, and advice in the world of AI detection.
All articles
01Minimum hosszDetektálásAug 30, 2026
Miért nehéz rövid szövegből detektálni?
A túl rövid szövegek nem rendelkeznek a szükséges statisztikai és nyelvi jellemzőkkel ahhoz, hogy megbízhatóan el lehessen dönteni a szöveg forrását (AI vagy ember).
Read article
02KézírásOCRJun 17, 2026
Kézírás felismerés fotóból: AI-detektálás kézzel írt szövegeken
A Preds képes kézzel írt szöveget tartalmazó fotókból digitális szöveget készíteni, majd AI-detektálást futtatni rajta. Hasznos tanároknak, diákoknak és intézményeknek.
Read article
03BővítményGyors detektálásMar 14, 2026
Preds Chrome bővítmény: AI-detektálás bármely weboldalon
Preds Chrome bővítmény. A gyors detektáláshoz!
Read articleFrequently asked questions
Key questions about accuracy, data handling, and usage.
Preds is built specifically on high-quality Hungarian-language data, taking into account the unique characteristics of Hungarian texts. Alongside high accuracy, we emphasize a low false-positive rate and transparent, sentence-level results.
The Preds model is itself an AI model that learned to distinguish human writing from machine generation on a massive volume of Hungarian texts. It provides a probabilistic result with confidence scores for pasted text and highlights suspicious passages sentence by sentence. The result is a decision-support signal, not definitive proof of authorship.
Preds is free to use: after registration, every user receives 3 free credits daily, which can cover up to 3 free detections. No credit card is required to sign up. For larger volumes, subscriptions or credit packages are available.
Yes, no AI detector is infallible. In our internal Hungarian-language benchmarks, accuracy exceeds 98% measured across nearly 100,000 independent texts, but the margin of error is higher for short, heavily edited, or translated texts. Results should therefore be evaluated with discretion alongside additional context.
Preds recognizes text written by ChatGPT, Claude, Google Gemini, DeepSeek, Meta AI, Grok, Llama, Mistral, Perplexity, and Deep Research artificial intelligence. In addition, it detects "humanizer" tools such as Grammarly and QuillBot.
Yes, detection from files is supported. We support TXT, PDF, DOCX, Markdown, and image formats. Detection can also be run on handwritten texts, which is available in the Essential and Pro subscriptions.
Start your first detection for free.
No credit card needed, no complex setup. Paste your text and receive fast, transparent results.