What do universities say about ethical AI use by students?

Universities increasingly treat AI as unavoidable in student work—but they draw a hard line between using AI to support learning and using it to produce “original content” that should come from the student. The practical takeaway is straightforward: ethical AI use means the student remains the intellectual driver, while AI functions as a research assistant, writing coach, or analysis accelerator rather than an invisible replacement.

A major theme across the guidelines Dr Kriukow reviewed is the distinction between AI-generated output and AI-supported work. Relying on AI to write an entire chapter or dissertation is framed as unethical because it substitutes for the student’s intellectual contribution. The same logic applies to data analysis: simply feeding data into ChatGPT and returning results is treated as producing content without the researcher’s involvement. By contrast, using AI to support analysis—when the student directs the process with specific prompts, checks outputs, and stays accountable—can be acceptable. The emphasis is on control, not convenience.

The transcript also maps where AI use tends to be most defensible during the dissertation lifecycle. In the literature review stage, AI can help with brainstorming research gaps, summarizing existing studies, and sorting articles by themes—especially tools like SciSpace that help triage large volumes of papers so students can decide what to read in depth. In writing, AI should not be used as a ghostwriter. Instead, it can generate structure ideas, suggest bullet points, provide feedback on clarity and grammar, and help with rephrasing or transitions—so long as the student reviews and keeps the work aligned with their own argument.

Data analysis is treated as the most sensitive area. The core concern is the lack of an audit trail—evidence of how decisions were made and how the student arrived at results. Full automation is discouraged, but cautious, incremental use is presented as workable: generating initial codes, refining themes, organizing codes into categories, and stress-testing logic in ways similar to intercoder reliability. Even then, the student must be able to demonstrate participation in the process and provide documentation.

Finally, the transcript argues that students should not hide AI use. Some students avoid disclosure out of fear that any mention will trigger suspicion or penalties, including cases where AI detectors falsely flag human-written work. The countermeasure recommended from the university guidance is transparency: reference the tools used, link to them when appropriate, describe the role AI played, and document the workflow. That includes saving prompts and screenshots, showing how AI outputs were incorporated or adapted, and consulting assignment-specific rules from supervisors or lecturers.

The overall message is less about banning AI than about protecting academic integrity and skill development. Students are urged to ask a simple question: does AI use harm learning and professional growth, or does it support the student’s own thinking? Universities highlighted in the review include Newcastle University, Oxford, Stanford University, MacE University, and London School of Hygiene and Tropical Medicine, all of which—despite different emphases—converge on the same principle: AI should assist, not replace, the student’s original work and accountability.