Referencing AI

AI outputs should normally be acknowledged rather than referenced in academic work, for two key reasons:

• Authorship: GenAI tools cannot be considered authors in the traditional sense, as they cannot be responsible for the accuracy, integrity, and originality of the work. These are uniquely human responsibilities.
• Traceability: One of the main purposes of a reference list is to give readers the information they need to access the sources a writer has used. With AI-generated content, this is often not possible, as tools generate responses from large-scale language data without clear attribution to specific sources.

This reflects the position taken by major academic publishers, who advise that AI tools should not be treated as authors or as standard citable sources.

You only need to provide a full reference if you directly use AI-generated text in your assessment.

Examples include:

• Quoting AI-generated text word for word.
• Paraphrasing AI-generated content and including it in your work.
• Referring to a specific AI output (e.g. “Copilot explained this theory by saying…”).
• Analysing or critiquing an AI output (e.g. to illustrate a hallucination or bias).

If you only used AI to brainstorm, outline, or clarify your own ideas, a short acknowledgment note is enough. See Acknowledging AI use for guidance.

At Abertay, the standard for Harvard referencing is Cite Them Right Online (CTRO). CTRO is the full, authoritative source and should always be your reference point for less common cases.

For most students, the situation you will encounter is referencing AI text generated in Copilot under the university licence, where a shareable URL can be provided.

Example:

In-text citation:
(Microsoft, 2025)

Reference list entry:
Microsoft (2025) AI-generated text by Copilot with prompt ‘Explain the main ideas of constructivist learning theory in simple terms’, 3 September. Available at: https://copilot.microsoft.com/ (Accessed: 3 September 2025).

If you are using other tools (such as ChatGPT, Gemini, or Claude, or AI image, video, code or music generators), follow the worked examples in CTRO: Harvard – Generative AI.

The authoritative source for APA referencing is the Publication Manual of the American Psychological Association (7th ed.) and the official APA Style Blog. These explain how to treat generative AI, usually as software or as a nonrecoverable source depending on the context.

For most students, the situation you will encounter is referencing AI text generated in Copilot under the university licence, where a shareable URL can be provided.

Example:

In-text citation:
(Microsoft, 2025)

Reference list entry:
Microsoft. (2025). Copilot (March 2025 version) [Large language model]. https://copilot.microsoft.com/

APA guidance also recommends making your use of AI transparent by recording the prompt you used. You can do this in your text, in a footnote, or in an appendix. For example:

• Prompt: “Explain the main ideas of constructivist learning theory in simple terms.”

If you are using other tools (such as ChatGPT, Gemini, or Claude), or need to reference nonrecoverable conversations, see the official APA Style Blog:  Citing generative AI in APA Style

CTRO also provides APA templates, but the APA Manual and Style Blog remain the primary authorities.

Currently, OSCOLA (4th ed.) does not include explicit rules for referencing generative AI. 

The most common recommendation is to treat AI outputs as personal communication / nonrecoverable material, citing it only in footnotes (not in the bibliography). 

For the most typical use case — AI text generated in Copilot under the university licence, where a shareable URL is available — you might use a footnote like:

• Microsoft, Copilot (March 2025 version) https://copilot.microsoft.com/ accessed 3 September 2025.

But note: this is a suggested workaround, not an official OSCOLA rule. Always check you assessment briefs, and consult the OSCOLA handbook  for how to cite unpublished or nonrecoverable sources more generally.

This checklist will help you decide whether you need to provide a full reference, or if a simple acknowledgment is enough. Before you submit your work, ask yourself:

Has my tutor asked me to reference AI? If not, I should only acknowledge its use, not cite it in-text or add it to the reference list.

Have I made it clear which parts are mine and which came from AI? Transparency is essential to maintain academic integrity.

Am I using AI outputs directly in my work?

• Quoting, paraphrasing, or referring to a specific AI-generated output = reference needed.
• Analysing or critiquing an AI output (e.g. to illustrate a hallucination or bias) = reference needed.

Follow these tips to make sure you reference AI responsibly:

• Follow University rules: Always follow the wider University guidance on AI use (see Rules and Responsibilities.
• Use the correct format for your discipline: Harvard, APA, or OSCOLA examples on this page show how to reference AI outputs.
• Record details at the time: Note the tool, version, and date of access or supply when you use the output. If your tutor has given you an AI output, use the date you received it as the access date. This will help you construct an accurate reference if required.
• Be transparent: Make it clear which parts of your work come from your own thinking, which parts include AI outputs you are referencing, which AI examples you are showing for critique (e.g. hallucinations), and which sources are academic references.
• Be critical: Check AI outputs carefully for accuracy, reliability, and potential bias. If you are referencing or critiquing an output, make that purpose clear.
• Stay proportionate: Only reference AI when required. The majority of your references should come from peer-reviewed academic sources.
• Balance with academic sources: AI outputs are not academic sources. Your reference list should be built mainly on peer-reviewed material from LibrarySearch and other databases.