AI Text at University: What Is Allowed and What Is Not?

Is AI allowed at university? Yes — but not automatically. Whether you may use ChatGPT, Copilot or any other AI tool in an essay, a seminar paper or your bachelor’s thesis is almost always decided by your institution, and in practice by the person who marks your work. Without clear permission, or without documenting what you did, AI use can quickly be treated as an unauthorised aid.

This guide explains what institutions typically expect, where AI use becomes academic misconduct, and how to use AI without breaking the rules. Mimir Mentor is an AI-powered platform for academic work that lets you check your finished text for AI-generated passages and plagiarism before you hand it in.

Short answer: what is allowed when it comes to AI at university?

AI is usually allowed at university when it is permitted as an aid, when you disclose that you used it, and when the academic substance is clearly your own. What is normally not allowed is having AI produce whole paragraphs, analyses or references and then submitting them as your own work without saying so.

If there is no clear rule in the assignment brief, the module description or the declaration of authorship, ask your marker before you start. General advice you find online never replaces the specific requirements of your own course.

Is AI allowed at university?

AI is often allowed at university, but only on conditions. Almost every current set of institutional guidance comes down to the same three things: what your teaching staff have specified, whether the work is genuinely your own, and whether you have documented your AI use transparently.

That is exactly why the question matters so much right now. In a German survey by Hochschulforum Digitalisierung, more than 96% of the students questioned said they use generative AI tools in their studies (source: Hochschulforum Digitalisierung). At the same time, many institutions still report considerable uncertainty about rules, assessment formats and declarations of authorship. For you that means one thing above all: never rely on general statements you find online — always check what your own module actually requires.

What do institutional rules actually look like?

There is no single rule for AI in assessed work that applies everywhere. Institutions regulate it through their examination and academic misconduct regulations, through guidance notes, through the declaration of authorship and through instructions given in individual modules. The common denominator is remarkably consistent: AI can be allowed if it is permitted, disclosed and combined with genuine work of your own.

The four examples below come from German-speaking universities, which have published unusually detailed public guidance. Read them as illustrations of the pattern rather than as rules that apply to you — what counts in your case is your own institution’s regulations.

The important takeaway is this: the real question is rarely “AI, yes or no”, but on what conditions AI may be used. Even two seminars at the same institution can set different rules.

Don’t stop at your institution’s general AI guidance. Always check the specific module description, the assignment brief and the wording of the declaration of authorship as well.

Does the picture change from country to country?

The details differ, but the underlying logic is strikingly similar wherever you study. The rule is rarely “AI is always allowed” or “AI is always banned”. Instead, teaching staff decide which aids are permitted in a particular assessment, and students carry the responsibility for transparency, accuracy and originality.

The University of Vienna, for example, states explicitly that the relevant teaching staff decide whether and how AI may be used in assessed work. Students there bear full responsibility for the originality, accuracy and transparency of what they submit, and anything not expressly permitted in the course description is treated, in case of doubt, as not permitted (source: University of Vienna).

The direction of travel elsewhere points the same way: regulated use rather than a blanket ban. At ETH Zurich, AI is already being factored into teaching and assessment design, while the institution’s guidance stresses that responsibility and declaration remain central whenever AI is used. Here too, the practical advice is the same — don’t rely on general assumptions, check what your specific module requires (source: ETH Zurich).

When is AI use acceptable?

AI use is typically acceptable when it has been expressly permitted, when you are open about it, and when the academic substance is still clearly your own. The closer AI gets to producing the actual text or the actual argument, the more problematic it becomes.

Usually fine

• Brainstorming topics, research questions or possible structures
• Coming up with search terms for your literature search
• Feedback on your writing, where the content is already yours
• Checking the style and grammar of text you wrote yourself
• Summaries you use for revision or exam preparation

Risky or frequently not permitted

• having ChatGPT write whole paragraphs or chapters and submitting them as your own prose
• letting an AI develop the argument or the central analysis of your work
• copying AI-generated sources or quotations without checking them
• signing a declaration of authorship when substantial parts of the work are not yours
• keeping quiet about AI use when your module asks you to disclose it

Our guide on disclosing AI use walks through how to document what you did, step by step. And if you are unsure whether your text already reads too much like AI, the guide to AI detection will help you judge it.

Check your text for AI-generated passages and plagiarism before you submit

When does AI use become academic misconduct?

AI use becomes misconduct when you use an aid you are not allowed to use, or when you pass off someone — or something — else’s work as your own. The problem is never the tool itself; it is the breach of the assessment rules and of the requirement that the work be your own.

It becomes particularly problematic in these situations:

• You use AI even though the assignment brief explicitly forbids it.
• You present AI-generated text as your own writing.
• You don’t disclose your AI use even though transparency is required.
• You use invented sources, quotations or facts taken from an AI answer.
• You cannot explain, when asked, how your text came about.

Many institutions now handle such cases much as they would any other form of academic misconduct. At the same time, several universities stress that purely technical proof of AI use is not currently reliable enough on its own. In practice, suspicion therefore tends to arise from breaks in style, missing sources, follow-up questions in a viva, or the documentation of how you worked.

What are the possible consequences?

The consequences depend on your institution’s examination and academic misconduct regulations and on the individual case. Typically, though, they are the same sanctions that apply to any other form of misconduct: a fail, the loss of an attempt, or more far-reaching measures in serious cases.

One case shows how seriously this is now taken. On 25 February 2026 the Administrative Court of Kassel in Germany upheld two decisions by the University of Kassel in which assessed work was failed for serious misconduct through prohibited AI use, and the students were additionally barred from resitting. Leave to appeal was granted, so the decisions are significant but not yet final (source: Administrative Court of Kassel). The ruling has no direct effect outside Germany, but the reasoning behind it is familiar everywhere: undisclosed AI use undermines the declaration you sign when you submit.

What does this mean for essays, seminar papers and theses?

The more open-ended and the longer the assessment, the more likely it is that some permitted AI use is on the table — but even then, only within rules. For essays, seminar papers and final theses in particular, institutions generally expect a recognisable contribution of your own, traceable work with sources and clear documentation.

Essays and term papers

Here AI is often tolerated, or explicitly permitted, for generating ideas, structuring the argument and giving feedback on your writing. Generating the text itself without disclosing it, on the other hand, is high-risk.

Seminar papers

A lot depends on the discipline and on what the seminar is trying to achieve. Where independent thinking, argument and critical engagement are the point of the exercise, generated prose is especially sensitive.

Bachelor’s or master’s thesis

Institutions are usually strictest with final theses, because independent academic work is the whole point of them. Check especially carefully which aids are permitted. For a final quality check, our guide to the plagiarism check for a bachelor’s thesis is worth reading too.

How to use AI within the rules

Using AI within the rules comes down to three habits: use it deliberately, check what it gives you, and document what you did. The steps below will cut your risk substantially.

1. Check the rules for your module: Read the module description, the assignment brief and the declaration of authorship carefully. If there is no clear rule, ask.
2. Use AI as support only: Use it for ideas, feedback or alternative phrasings — not as a substitute for your own argument.
3. Document what you used: Record which tool you used and what for. Many institutions expect exactly this.
4. Verify sources and facts: Check every reference and every factual claim yourself.
5. Check the text before you submit: Look for AI-sounding passages, missing sources and plagiarism risks.

A good workflow therefore looks like this: gather ideas with AI, write the text yourself, verify your sources properly, and finish with a pass through the text correction and the combined AI and plagiarism check.

Rewrite AI-assisted passages in proper academic prose

How to check your text before you submit

The safest approach is a short self-check before submission. That way you spot early on whether your text still sits too close to AI phrasing, or whether sources are missing.

1. Finish the final version: Write the whole thing first, then come back to it after a break and read it again yourself.
2. Upload the document: Upload your PDF or DOCX to mimir-mentor.com.
3. Read the report: Look at the flagged passages for AI-generated content, plagiarism and unsupported claims.
4. Assess what has been flagged: Check accuracy and attribution, revise where necessary, and follow the AI-use rules. Do not treat a detector flag as proof or rewriting as a substitute for disclosure.
5. Complete your documentation: Record which tools you actually used, and how.

Don’t leave this self-check until the evening before the deadline. A buffer of two or three days often makes all the difference with essays and final theses.

Mimir Mentor is GDPR-compliant, with servers located in Mainz, Germany. Your document is processed solely for the purpose of the analysis and is deleted in full after 30 days at the latest. Your text is never added to a public database.

If you now want to move from the rules to the practicalities, read on in disclosing AI use, forgot to cite a source and AI detection at university.

FAQ: common questions about AI at university