Using generative AI in academic studies and research requires transparent labelling and documentation, both to maintain transparency in one's own working methods and to meet academic standards.
If texts are generated using AI tools, they must always be identified as such; otherwise, they may be classified as attempted deception. Even though the rules for AI‑supported literature research are often less strict, it is still advisable to clearly disclose the use of AI, especially when it contributed to key steps such as literature selection or document analysis.
What matters is that academic work remains an independent achievement for which students and researchers take responsibility in terms of content and accuracy. Since requirements vary across disciplines and departments, it is essential to always follow the guidelines provided by supervising instructors and the department.
The following explanations are intended solely as general guidance and can be used either individually or in combination:
The guidelines of common citation systems (APA, MLA, Chicago Style) provide guidance on how to clearly mark AI-generated passages as direct or indirect quotations.
Purpose:
Example in APA Style from the following APA article:
The use of AI in the research process (e.g., literature review, brainstorming, text generation, or text revision) can be explained in the introduction, the methods section, or footnotes.
Purpose:
Example of documenting AI use in formulating the research question:
A brief explanation of how AI influenced the research process can be included in the final part of the paper. This makes it possible to clearly situate the role of AI and to point out potential limitations.
Purpose:
Example for the conclusion/discussion section:
A separate documentation in the appendix provides a structured overview of AI usage, for example in tabular form with details on work step, tool used, and outcome.
Additionally, AI-generated content should be stored locally to enable future reference. This is particularly important for longer processes where the connection to original input might otherwise be lost.
Purpose:
Example:
