Documenting and disclosing the use of AI
A template for term papers and coursework
Why disclose?
If you use an AI system (such as ChatGPT, Claude, Gemini, or Copilot) at any stage of your academic work, you are expected to document what you did and disclose it transparently. Doing so is not an admission that you cut corners — it is part of good scientific practice. It lets readers judge how your results came about, protects you against accusations of academic misconduct, and forces you to reflect critically on where the AI actually helped and where it did not.
The core principle is simple:
While AI tools may assist your work, you remain fully responsible for every claim, analysis, and conclusion you submit.
This page gives you (1) a fillable template to include in your work, (2) a worked example, and (3) a short note on how to cite AI in your reference list. Always check the concrete requirements of your specific course or examination regulations — this template is a sensible default, not a rule that overrides them. If in doubt, always clarify this topic with your instructor as early as possible.
How to use this template
Copy the table below into an appendix (or, for short pieces, summarise the core rows in your methodology section). Fill in the right-hand column for your own work. The elements are grouped into two tiers:
- Core elements — always report these, even for a one-paragraph disclosure.
- Additional elements — report these when they are relevant to your work (e.g. when AI touched your data, or when its limitations affected your results).
The disclosure template (fill this in)
Core elements — always report
| Element | What to include | Your entry |
|---|---|---|
| AI tool & version | Name and provider of the system, the specific model/version, and the date(s) you accessed it. | |
| Access method & plan | How you used it (web interface, API, IDE plugin …) and which subscription tier, noting capabilities that shaped the output (e.g. document upload, extended reasoning). | |
| Contribution scope | Exactly which parts of the work involved AI (e.g. brainstorming, restructuring, code, language editing) — and, just as importantly, which parts did not. | |
| Human oversight & verification | How you checked, corrected, and revised AI output, and against which sources. | |
| Documentation location | Where the full record lives (e.g. chat logs in Appendix B, key prompts in the methodology section); only if this is required to report. |
Additional elements — report when relevant
| Element | What to include | Your entry |
|---|---|---|
| Prompt & iteration strategy | How you structured your requests and how many rounds of refinement were needed to get usable output. | |
| Ethical & data-protection considerations | How you handled confidentiality, personal data (GDPR — e.g. interview data), and possible bias in the model. | |
| Limitations noted | Known constraints of the tool that affected your work (knowledge cut-off, hallucinated references, topical bias). |
Integrity statement
Close your disclosure with an explicit statement of responsibility. You can adapt the following:
While AI tools assisted with aspects of this work, as documented above, I maintain full intellectual responsibility for all content, analysis, and conclusions presented. All AI-generated content has been critically evaluated, substantially revised, and personally endorsed as academically sound. No confidential or personal data was shared with the AI system except where noted and permitted.
Worked example: using an AI assistant for a term paper
The following shows the template filled in for a fictional Management-Studies term paper. Your actual tool, model name, and cut-off date will differ — record whatever your tool actually reports.
| Element | Example entry |
|---|---|
| AI tool & version | Claude (Sonnet 5) by Anthropic; accessed 21–25 September 2026. |
| Access method & plan | Web chat interface at claude.ai, Pro subscription — used for document upload of research papers, extended reasoning, and higher message limits. |
| Contribution scope | (a) Organising the initial literature review; (b) generating alternative interpretations of my interview data for me to evaluate; (c) checking the clarity of my argument structure. The research design, data collection, coding, and all conclusions were my own. |
| Human oversight & verification | Every AI suggestion was cross-referenced with course literature, substantially edited to reflect my own analysis, and verified against additional peer-reviewed sources. |
| Documentation location | Full conversation logs in Appendix B; key prompts summarised in the methodology section. |
| Prompt & iteration strategy | Provided specific management theories and asked how they might apply to my case data; used iterative follow-up prompts; shared sample paragraphs for feedback. Roughly three rounds were needed to clarify the theoretical framework. |
| Ethical & data-protection considerations | No confidential company data and no personally identifying interview details were shared. AI-suggested frameworks were always verified against published sources. I stayed alert to the model’s tendency toward mainstream management perspectives. |
| Limitations noted | The model’s knowledge cut-off preceded some recent developments in the field; I checked recent literature manually. The tool occasionally produced plausible-looking but incorrect citations, which I removed. |
While AI tools assisted in aspects of this work, as documented above, I maintain full intellectual responsibility for all content, analysis, and conclusions presented in this paper.
Citing AI in your reference list
Disclosure (the table above) and citation (an entry in your bibliography) are complementary, not interchangeable. Most style guides now treat generative AI as a retrievable source. In APA 7, for example, the author is the company that made the model, and the format is roughly:
Anthropic. (2026). Claude (Sonnet model) [Large language model]. https://claude.ai
Check the current guidance for the citation style your course requires (APA, MLA, Chicago, or the journal style used in this module), as recommendations for AI sources are still evolving. Personally, I find acknowledging and documenting the use of AI far more useful and important than citing an AI model.