
A plagiarism check for your bachelor’s thesis, master’s thesis or term paper: you want to be sure your text is clean before you hand it in. But how does plagiarism detection actually work? Which types of plagiarism are caught, and what does the similarity score really mean?
Mimir Mentor is an AI-supported platform for academic work that combines plagiarism detection, literature search, text correction and AI detection in a single tool. What makes it different: a purpose-trained AI model that recognizes unsupported claims in your text. More than 45,000 students already use Mimir Mentor. In this guide we explain step by step how the plagiarism check by Mimir Mentor works, which technologies sit behind it and how to use the tool deliberately to protect your work.
What is a plagiarism check and why does it matter?
A plagiarism check is a technical procedure that compares your text against reference databases and uses AI analysis to look for uncredited use of other people’s content. Plagiarism occurs whenever you take over someone else’s intellectual work without a correct source reference, whether you did it deliberately or by accident.
Which types of plagiarism are there?
Plagiarism is not limited to the obvious copying of whole paragraphs. The most common forms are:
- Copy-and-paste plagiarism: You copy sentences or sections straight from a source without marking them as a quotation. The simplest form and the easiest one to detect.
- Paraphrasing plagiarism: You rewrite the original text but keep the line of thought and the structure of the argument, without giving a source. This happens unintentionally particularly often.
- Translation plagiarism: You translate a text from another language and present it as your own work.
- Structural plagiarism: You take over the outline, the structure of the argument or the layout of someone else’s work.
- Self-plagiarism: You reuse passages from your own, already graded work without making that visible.
When you paraphrase, swapping individual words is not enough. To avoid plagiarism you have to render the other person’s idea in your own sentence structures and your own chains of argument. And of course you still have to name the source.
Why should you check your work yourself before you submit?
At most universities today, every final thesis is checked for plagiarism automatically. Common software for this includes Turnitin, PlagAware and iThenticate. Shorter term papers are also being checked more and more, sometimes by random sample, sometimes systematically.
The German Association of University Professors and Lecturers (Deutscher Hochschulverband, DHV) calls on universities to have work submitted in digital form as well, so that it can be checked for matches with plagiarism software. More and more universities also store submitted work digitally in order to compare it internally. If you check your own text for plagiarism before you submit, you avoid nasty surprises.
How does Mimir Mentor’s plagiarism check work behind the scenes?
Mimir Mentor combines several detection methods that go beyond a classic text comparison. The process starts when you upload your document (PDF or DOCX). The text is extracted and split into meaningful sections. Your reference list is recognized automatically and excluded from the scan, so that source references do not trigger false matches. Your text then goes through three central analysis steps:
Step 1: AI pre-analysis for unsupported claims
Before any database comparison happens at all, Mimir’s purpose-trained AI model analyzes every section and identifies passages that look like factual claims without a source. Sentences that are already cited correctly are not checked again. Conventional plagiarism scanners skip this step entirely and only compare against databases. Mimir’s AI pre-analysis therefore also finds problematic passages for which there is no exact match in any database but which still require a source.
Step 2: Database comparison
Relevant sections of text are then compared against an extensive reference database. Mimir Mentor draws on more than 200 million academic publications plus publicly accessible web sources. Academic domains such as university sites and specialist portals are searched with priority.
Step 3: Match detection with semantic analysis
Two methods are used when the results are evaluated:
- Literal similarity: The words and word order of your text are compared with the sources. This is the classic matching for copy-and-paste plagiarism.
- Semantic similarity: Using modern language models (embeddings), the AI compares the meaning of whole sentences. This is how creative rewrites and translations are detected, even when not a single word matches. Mimir supports academic writing in English and German. Detection performance depends on the text, language and sources available.
Finally all results are brought together and the relevant passages are color-coded. Every match is assigned to one of three plagiarism categories: copy-and-paste, paraphrase or translation plagiarism. That way you see at a glance what kind of match you are dealing with.
Evidence from Mimir data: In the Mimir Plagiarism Report for April 2026, 93.6 percent of the recorded flags were paraphrases, 3.2 percent copy-and-paste and 3.1 percent translation proximity. The categories are therefore not just theory: they help you read the report of the plagiarism check by Mimir Mentor in a targeted way.
Which detection methods does Mimir Mentor use?
The table below summarizes which methods are used in the analysis:
| Method | What it detects | How it works |
|---|---|---|
| Semantic analysis | Reworded passages | Modern language models (embeddings) compare the meaning of whole sentences with sources, independently of the words chosen. Creative rewrites are detected this way too. |
| Cross-lingual detection | Translation plagiarism | Comparison across languages: if you translate a German or Spanish journal article, for example, and present the result as your own English text, it is flagged as translation plagiarism. Most conventional scanners do not detect translation plagiarism, because they only compare within one language. |
| AI detection | AI-generated text | An integrated AI detector analyzes whether passages show the typical language patterns of generative AI tools such as ChatGPT. |
| Unsupported-claim detection | Missing source references | Mimir’s purpose-trained AI model identifies claims that require a source but do not have one. This happens before the database comparison. |
Check your text for possible plagiarism right away with Mimir Mentor
Understanding and using the plagiarism report
After the check you receive a detailed report. Here we explain how to read it, what the similarity score really means and how to work with the results in practice.
What does the similarity score mean?
The similarity score states as a percentage how much of your text matches sources in the reference database.
Important: A high similarity score is not automatically plagiarism. Correctly cited direct quotations, standard phrases and technical terms can push the score up without anything being wrong. Conversely, a few uncited passages alone can be judged to be plagiarism.
There is no officially “permitted” percentage. What always matters is whether the flagged passages are cited correctly. The final judgment is made by your professor. The report is a working basis: the flagged passages show you where you should look more closely.
What does the Mimir Mentor plagiarism report show?
Mimir’s report goes beyond a bare percentage. The premium check analyzes your entire document sentence by sentence and shows you, for every potential match:
- Plagiarism category: A clear assignment to copy-and-paste, paraphrase or translation plagiarism, so you know immediately what kind of match this is.
- Similarity and risk assessment: How high the match is and how critical it should be considered.
- Original source: A direct link to the source found, so you can check the context.
- Additional reports: Alongside the plagiarism report you receive the AI detector report and an academic correction report covering spelling, grammar and style.
Mimir Mentor does not only flag problems, it helps you fix them directly. The AI rephrasing assistant supports you in developing your own, academically phrased alternative. With the integrated literature search you find suitable sources from more than 200 million publications without leaving the platform, and you can format them directly in the citation style you need (APA, Harvard and so on). In parallel, the academic text correction checks spelling, grammar, informal or imprecise wording and gender-inclusive language. And through concrete style suggestions you learn to express matters clearly and in your own words. That is the best remedy against unintentional paraphrasing plagiarism.
How should you work through the report?
A plagiarism report is only useful if you work through it systematically. Here is the best way to go about it:
- Get an overview: Look at the overall score and the number of matches. Are there many flagged passages or only a few?
- Check the matches one by one: Go through the flagged passages in order. Is the passage cited correctly? Is a source reference missing, or is this simply a technical term?
- Add missing sources: Use the integrated literature search to insert suitable sources directly in the right citation style.
- Rework your phrasing: For paraphrasing matches, the AI rephrasing assistant helps you restate the thought in your own words.
- Check the correction report: Finally, work through the correction report to improve grammar, style and academic tone.
Use the flags as a checklist: go through every flagged passage and make sure that each borrowing is cited correctly. Not every match is a problem, but every one deserves a quick look.
Are the reference list and quotations excluded?
Yes. Mimir recognizes your reference list automatically and excludes it from the plagiarism scan. That matters, because entries in the list inevitably match databases and would otherwise trigger false alarms.
Direct quotations in the body text, by contrast, are checked and flagged. This lets you verify whether your citations are complete and correct. Technical terms and common definitions are shown as context but are usually unproblematic.
What sets Mimir Mentor apart from other plagiarism checks?
Conventional plagiarism scanners compare your text with a database and display matches. That works well for copy-and-paste plagiarism but reaches its limits in more complex cases. Mimir Mentor goes further on several points:
| Feature | Conventional scanners | Mimir Mentor |
|---|---|---|
| Detecting unsupported claims | No. Only matches from databases are shown. | Yes. Mimir’s own AI model recognizes passages that require a source, before the database comparison. |
| Translation plagiarism | Usually not detected, because comparison happens only within one language. | Cross-lingual detection finds plagiarism across languages, for example texts translated from English. |
| Academic language | Mostly tuned for general English, not for academic register. | Built for academic writing in English and German; results still require review. |
| AI detection | Sometimes available as a separate tool. | Integrated AI detector in the same workflow. |
| Self-plagiarism risk | Some providers store submitted work permanently in their database. | Documents are not added to a public comparison database. This does not remove the obligation to disclose reuse of your own earlier work. |
| Correction and source search | Not included. The plagiarism report has to be processed separately. | Integrated literature search (200 million+ publications), rephrasing assistant and academic text correction in one tool. |
| Data protection | Often US-based servers, data protection varies. | GDPR-compliant, server location Mainz. |
What happens if your work contains plagiarism?
Plagiarism counts as an attempt to deceive. The German Association of University Professors and Lecturers classifies it as “intellectual theft and a violation of copyright”. The standard consequence is failing the assessment. In serious or repeated cases, expulsion or the subsequent withdrawal of an academic degree are possible. How to protect yourself and what rights you have if the worst happens is covered in our detailed article on the consequences of plagiarism.
How do you protect yourself against unintentional plagiarism?
The best protection is a clean way of working from the start plus a careful check before you submit:
- Separate your notes: While you read your sources, mark immediately which thoughts are yours and what comes from the original text.
- Cite consistently: Better once too often than once too rarely. Use consistent citation styles such as APA or Harvard.
- Check before you submit: Upload your document to Mimir Mentor and work through the report systematically. That way you find the gaps before your university does.
Check your text for possible plagiarism right away with Mimir Mentor
Data protection and self-plagiarism: is Mimir Mentor safe?
Yes. Mimir Mentor is GDPR-compliant, with its server location in Mainz. Your document is processed exclusively for the analysis and is deleted completely after 30 days at the latest. Mimir does not add your text to a public comparison database. However, reusing earlier assessed or published work can still raise self-plagiarism concerns, regardless of where you run the check.
That is an important difference from many other providers: some conventional plagiarism checks store submitted work permanently in their own database. If you check there before you submit, your university’s check can then flag your own work as a match. With Mimir Mentor that does not happen.
Mimir Mentor offers two levels of plagiarism check:
- Free document scan: You upload your PDF or DOCX and receive an overview of how many plagiarism passages, AI matches and corrections were found in your text. This gives you a first assessment of your risk.
- Full report with editor (premium): The complete sentence-level analysis with a detailed display of every match, the additional reports (AI detector, correction report) and integrated tools for source research and text improvement.
Check and improve your work with the plagiarism check by Mimir Mentor. Which providers achieved the best results in an identical premium test is shown in our plagiarism checker comparison 2026.
Scientific Formulations in Minutes Seconds
FAQ: common questions about plagiarism checks
-
What exactly does a plagiarism check do?
A plagiarism check analyzes your work on two levels: first, an AI pre-analysis recognizes unsupported claims in the text. Relevant passages are then compared against a reference database of more than 200 million academic publications and web sources. The report categorizes every match as copy-and-paste, paraphrase or translation plagiarism.
-
How does Mimir Mentor detect paraphrasing plagiarism?
Mimir uses semantic analysis with embeddings to compare the meaning of whole sentences, not just individual words. That is how creative rewrites are detected too. In addition, Mimir’s own AI model identifies passages that look like unsupported claims, before any database comparison takes place.
-
Does a high similarity score automatically mean plagiarism?
No. The score shows what percentage of your text matches the database. Correctly cited quotations, technical terms and standard phrases can push the score up. There is also no officially permitted percentage. What always matters is whether the flagged passages are cited correctly. The final decision is made by your examiner.
-
Can AI-generated texts be detected as plagiarism?
AI use and plagiarism are separate issues. AI output can contain uncredited material, while unauthorised or undisclosed AI use may breach assessment rules even without plagiarism. Mimir Mentor contains an integrated AI detector that identifies the typical language patterns of tools such as ChatGPT. Beyond that, the plagiarism check itself also flags AI text, because it typically contains no source references.
-
What are the consequences of confirmed plagiarism?
Plagiarism counts as an attempt to deceive. The standard consequence is failing the assessment. In serious or repeated cases expulsion is a risk, and for academic degrees the subsequent withdrawal of the title.
-
Does Mimir Mentor also detect translation plagiarism?
Yes. Mimir’s cross-lingual detection finds plagiarism across languages. If you translate a German or Spanish journal article, for example, and present the result as your own English text, it is flagged as translation plagiarism.
-
Is the reference list excluded from the plagiarism scan?
Yes. Mimir recognizes your reference list automatically and excludes it from the scan. That way you avoid false matches from source references, which inevitably match databases.
-
Is Mimir Mentor GDPR-compliant?
Mimir Mentor is GDPR-compliant, with its server location in Mainz. Your document is processed exclusively for the analysis and is deleted completely after 30 days at the latest.
-
Do I risk self-plagiarism?
Mimir does not add your text to a public comparison database. However, reusing earlier assessed or published work can still raise self-plagiarism concerns, regardless of where you run the check.
-
The free document scan shows you how many plagiarism passages, AI matches and corrections were found. The full report with editor view, the detailed sentence-by-sentence analysis, all additional reports (AI detector, correction report) and the integrated tools for source research and text improvement come with premium.



