AI detector for university texts
AI detectors are also discussed in the university context. Our AI detection tool gives you a sentence-level analysis of statistically conspicuous language patterns in university texts. The value must not be used as proof of origin. What counts is the actual text, the documented writing process and the rules of your institution.
With our system you know before you submit whether your text would be classed as AI-written. Detection works sentence by sentence: each sentence is analysed on its own and checked for patterns that point to machine generation. That way you can revise specific passages, or be sure that your text reads as human-written.
AI check for essays, theses and university work
The detection for bachelor's theses and term papers is geared to the analysis of German texts. Because language models and human revisions keep changing, no detector can reliably attribute every model version. Combined with the plagiarism check, you get two separate evaluations for €0.39 per standard page.
How to read the report
AI detector: what the percentages really mean
A detector gives you a number, but not a verdict. The scale supplies the verdict: it separates the unproblematic range from the range in which you should revise your text again.
The display only turns a clear red above the fifty per cent mark; values below that are mainly for categorisation, not a warning.
You can find a more detailed discussion under Meaning, limits and interpretation.
Which AI models are detected?
- ChatGPT (GPT-6 Astra, GPT-6 Luna)
- Google Gemini
- DeepSeek (R1, V3)
- Microsoft Copilot
- Mistral, Claude, Grok, Perplexity, Meta AI
Start analysis
At a glance
- From €0.29 per standard page
- Result in 30 minutes at most
- All common LLMs detected
- Sentence-level colour marking
- Detailed probability report
- 100% anonymous, no data passed on
- Combo bundle + plagiarism: €0.39 (−33%)
- No registration required
Helpful combo scan
„Helpful combo scan. I ordered the plagiarism + AI scan and was able to correct several citations, and on the repeat scan the result came out at 0%.“
Accurate results from the plagiarism and AI scan
„I have absolutely nothing to complain about. The results are satisfactory and completely comprehensible. Both the plagiarism and the AI scan gave accurate results. Although a technical error occurred during my second scan, I was helped promptly, even on a Saturday. All in all I am very satisfied and can warmly recommend the service!“
Combo scan in 2 minutes
„The combo scan arrived by email in 2 minutes, great!“
Frequently asked questions
What is an AI detector?
An artificial intelligence detector is a tool that analyses texts for linguistic patterns typical of machine-generated content. Our system returns a probability value with sentence-level marking, but not proof of origin.
How much does the check cost?
The check costs €0.29 per standard page. The combo bundle with the plagiarism check costs €0.39 per standard page, 33% cheaper than buying them separately (instead of €0.58). A standard page is 1,800 characters including spaces.
How long does the check take?
You normally receive your report by email within 30 minutes at most. The system runs fully automatically, including nights, weekends and public holidays, with no waiting times.
How reliable is the check?
No AI detector can reliably prove who wrote a text. Our system evaluates statistical features. False positives and false negatives are possible, especially with short, translated, mixed or heavily revised passages. Use the markings as a reason to look at the content again and never as sole proof.
AI detection & AI detector
This page is our AI detector. For a broader report, see the AI scan, which covers all common chatbots.
Where is the service available?
Our service is available online for students in Germany, Austria and Switzerland, with no shipping: you simply upload your file.
What is the AI + plagiarism combo bundle?
The combo bundle combines AI detection with the plagiarism check in a single order, the best price for both checks together. Instead of €0.58 you pay only €0.39 per standard page, a saving of 33%. That way you protect yourself against both AI flags and plagiarism allegations.
Which formats are accepted?
We accept DOCX, DOC, TXT, PDF, Pages, ODT and RTF up to 100 MB. Export LaTeX documents as PDF before uploading. You receive the report by email.
Is my work passed on to my university?
No, absolutely not. Your work is used only for the analysis and deleted afterwards. Nothing is passed on to universities, colleges, databases or third parties. All data is treated confidentially and 100% anonymously.
Is the content of my work changed?
No. The system is a purely analytical tool: it only reads your text and produces a report. Your text stays 100% unchanged.
Plagiarism check & plagiarism checker
Our online plagiarism checker compares your text with millions of sources and delivers a detailed originality report in 30 minutes at most. The plagiarism checker also detects paraphrased passages. Detection runs fully automatically.
Proofreading & editing
Besides detection, we offer professional proofreading (from €1.90) and editing (from €4.90). You can also have machine-assisted texts checked by a professional editor. Bachelor thesis proofreading is carried out by trained specialist editors.
Is the analysis also available in English?
Yes, the detector can also analyse English texts. How meaningful the result is depends on language, text type, length and revision. For a source-based overview, see our comparison of AI detectors.
More questions
How reliable is AI detection compared with GPTZero or Originality.ai?
An independent 2026 study of 160 documents shows clear performance differences between Pangram, GPTZero, Copyleaks and Turnitin. Pangram performed best on this corpus. Newer models, mixed texts and human-revised texts remained difficult for several systems.
Are DeepSeek, Mistral and Claude also detected?
The AI detector analyses statistical language patterns across models. It does not reliably attribute a text to one specific model. Statements such as "written by ChatGPT" or "written by Claude" could not be proven from a detector value alone.
What is the false-positive rate?
A general false-positive rate cannot be stated responsibly without a published, reproducible test design with a fixed corpus, language and decision threshold. False positives are possible, especially with short, formulaic or heavily revised texts. That is why you should always check the marked passages and not just the percentage value.
Does the AI detector also detect translated texts or paraphrases?
In part, but with no fixed detection rate. Translation and paraphrasing change statistical language patterns and can therefore favour both false negatives and false positives. An additional plagiarism check answers a different question, namely whether passages match sources that can be found.
Does my university use AI detectors itself?
More and more universities build AI detection into their exam software, often directly in the plagiarism scanner. Which tools are used differs from university to university. It is worth asking your department or examinations office which software it uses.
What should I do if the detector shows more AI than I expected?
A raised AI value does not automatically mean a problem. If you worked on parts with the help of AI, that is understandable. What matters is that you mark those passages and add your own argument. For passages that are entirely AI-generated, we recommend rewording them and having the text edited afterwards.
Do anti-detector tools such as Humanize.ai help reliably?
A general detection rate for texts from anti-detector or humanizer tools cannot be stated responsibly. Such edits change statistical language patterns and can raise or lower the detector value depending on the text, language and tool. So do not rely on supposedly certain detection or on supposedly certain evasion.
How does AI detection work technically?
AI detectors analyse several statistical signals at the same time: perplexity (how "predictable" the sequence of words is), burstiness (variation in sentence length) and n-gram distributions that are characteristic of language models. Modern detectors add a second neural network trained to classify AI versus human writing.