Good academic practice is not about prohibiting or limiting your use of technology. It is about using it with academic integrity. It requires you are transparent, and that you openly and honestly declare your use of the technology, so others can assess and understand. Likewise, it requires that you remain critical of the output generated by the model and take full responsibility for the academic content you hand in.

To uphold academic integrity, you must choose the appropriate form of acknoledgement based on how AI has contributed to your work. Different requirements may apply regarding both the format and the level of documentation you are expected to provide. The requirements depend on the subject, the assignment, or your supervisor's directions. If you are unsure, you should always look into the matter and discuss it with your supervisor. Doing so may itself be regarded as part of good academic practice.

Declaration (the work process)

A general written account of whether and how AI has been used as a tool during the work process, or as a sparring partner (for example through a declaration template, a cover sheet, or a section within the methodology chapter).

Documentation (audit trail)

The actual, underlying empirical material from your interaction with the model (e.g. prompt history, raw outputs, screendumps, or shared chat links) enclosed as appendices.

Citations (specific text)

Direct rendering of the model's wording (pieces of text or answers) where the text is explicitly marked as a quote; typically used if the specific output from the language model is subjected to text analysis. 

Source reference (bibliographic reference)

A formal reference, often adhering to a specific standard (e.g. APA 7th edition) whenever a model is included as as substantial element or test specimen in the paper/project.

Reflection (what and how) and metareflection (why)

Not part of the academic practice as such, but your supervisor may ask you to consider and document how use of generative AI has affected e.g. your learning in the process. 

What can I do myself?

Get started right away: if you develop a work routine for declaration and documentation from the outset, you avoid the vast task of clearing up your documentation right before submitting your project. For example:

Clarify the formal requirements first: Always check the specific guidelines on AI use for the assignment in question to ensure that you understand what is permitted.

• Agree on AI use: If you are working in a group, establish and document which stages of the project AI may be used for (e.g. outlining, language feedback, or identifying search terms) and which parts must be completed without AI assistance (e.g. analysis, methodological decisions, or conclusions). For further guidance, see the section AI in project groups.
• Maintain a shared prompt log from the outset: Create a central working document in which you continuously record the date, tool used, model version, prompts, and raw outputs. Share the log among all members of the group.
• Colour-code all AI-generated text in drafts: Apply a consistent colour to text suggestions and rewrites produced by AI so that unedited machine-generated text does not inadvertently make its way into the final submission as undeclared content or plagiarism.
• Treat AI-generated sources and empirical material with caution: Adopt a standing rule that no quotation, factual claim, or bibliographic reference produced by a language model may be included in the project/assignment until the original source has been located and verified.
• Protect confidential data and personal information: Do not share personally identifiable information (for example, interview transcripts containing names or contact details), confidential organisational data, or similar material with public language models. Comply with data-classification requirements and use only approved systems provided by the university.

Format formal references correctly: If a formal citation style is required for an AI tool, references should be included in accordance with the relevant style guide. For example, in APA 7th edition:

• In-text citation: (Microsoft, 2026) or Microsoft (2026)
• In bibliography: Microsoft. (2026). Copilot (GPT 5.6) [Large Language Model]. https://m365.cloud.microsoft/chat

Links to several widely used referencing styles can be found here: APA style, MLA style, IEEE style, Harvard style, Chicago style, Vancouver style. Read more at the library's page on sources and references.

Do not confuse an academic source with a tool: Never use a language model as an authoritative source for an academic argument or scientific theory (avoid formulations such as, "According to ChatGPT, motivation is defined as …"). Instead, locate authoritative sources in course literature or through the library's academic resources.

Transparency requires that your use of AI is reproducible, enabling others to assess the basis and validity of your claims. Declaration is particularly important when the use of AI is substantial, that is, when the tool makes decisions, generates data or images, or analyses content in ways that influence the conclusions of a project/assignment. In an academic context, process-related support, such as language assistance, idea generation, and the preparation of outlines, should also be declared openly.

What can I do myself?

Fill in an AI use declaration: Many programmes at AAU use standard templates that serve as practical tools for ensuring clear and consistent reporting. Follow the specifications of your programme, but if no template is provided, the following checklist may be used for inspiration:

Declare the process within the text: If you use AI in research design, literature searching, data processing, or similar activities, both the use of the tool and its role should be described in general terms within the methodology section, a cover sheet, and with reference to supporting appendices:

"During the development of the interview guide for the semi-structured interviews, Copilot (Microsoft, 2026) was used to identify and refine relevant research questions. Documentation of prompts, outputs, and subsequent academic adaptations is provided in Appendix 1. In addition, Copilot was used for language proofreading. The group subsequently reviewed, quality-assured, and revised all content."

Include supporting documentation in an appendix: Regardless of the type of assignment, you should always retain your key prompts and responses throughout the process. A missing audit trail cannot be reconstructed retrospectively. This documentation should be organised in an appendix to which you refer in the project/assignment. Unlike a book or journal article, a reader cannot simply consult a language model and obtain the same response that you received. Your documentation therefore helps ensure that the process is as transparent and reproducible as possible.

The output generated by language models is based on statistical probability and, by its nature, cannot be assigned the status of a reliable source of evidence. Consequently, it can never constitute primary literature or serve as an independent scholarly authority.

For this reason, it is a fundamental academic requirement that you always trace claims back to their original source or identify other authoritative sources that substantiate the information in question.

What can I do myself?

Always verify sources: If a language model suggests an academic source or a theoretical argument, you should never include the reference in your bibliography without first locating the original source; for example in your course literature or through the library's databases. You should then read the source in its original context and evaluate it according to established principles of source criticism. Once this has been done, you should cite the original source in your text.

Be critical of bias and regimentation: Always assess critically whether the model's output overlooks minority perspectives, oversimplifies complex issues, or favours a particular narrative at the expense of alternative academic viewpoints. Read more in the section on Ethics, Disparity, and Bias

Use AI to support search strategies rather than provide answers: Ask the model for suggestions of search terms, synonyms, or key theoretical perspectives, and then carry out the literature search yourself.

When you submit a project/assignment, you are declaring that the work is your own. If you present AI-generated text, code, translations, or analyses as your own work without declaring their use, this constitutes academic misconduct. Depending on the circumstances, it may be regarded as cheating and plagiarism.

What can I do myself?

Be honest, transparent, and accountable: Systematic declaration is the most effective way to avoid this issue. When your use of AI is transparent and properly documented, you protect your own academic integrity and remove any uncertainty about the origin of your work.

Regardless of how advanced an AI tool may be, full and unrestricted responsibility remains with you as the author. Errors, inaccuracies, fabricated sources, or bias within a project/assignment can never be excused on the grounds that they originated from AI.

What can I do myself?

Take ownership of the final product: You bear full personal responsibility for every piece of information, calculation, and reference included in your work. If a language model introduces factual errors, fabricated sources, or plagiarised text, it is your responsibility as the author to identify and correct them through critical review and quality assurance.