Pathway 3 — Using daily

AI in your research

Where AI can genuinely help in research workflows — and where it cannot. What you need to disclose, to whom, and why this matters more than it might seem.

This is a fast-moving area. Publisher policies, funder requirements, and institutional research ethics guidance on AI are all evolving. Check your specific publisher's author guidelines, your funder's terms, and your institution's research integrity policy before submitting any work.

Where AI genuinely helps in research

The most valuable uses of AI in research workflows are the ones where you are directing the work and evaluating the output — not the ones where you are asking AI to do the intellectual work for you. That distinction matters both for research integrity and for the quality of what you produce.

Literature scoping. AI can help you quickly scope an unfamiliar field, identify key themes, and generate a starting list of search terms. It cannot replace systematic literature searching — it will miss papers, hallucinate references, and cannot access most subscription journals. Use it for orientation, not for the review itself.
Writing assistance. AI is useful for improving clarity, restructuring arguments, simplifying dense passages, and drafting sections you then substantially revise. It is not useful as a ghostwriter — the intellectual content, the argument, and the interpretation must be yours.
Grant application drafting. AI can help you draft lay summaries, public engagement descriptions, and structured sections of applications. It cannot know your research context, your track record, or the specific priorities of your funder. Drafts need heavy editing and factual verification.
Data analysis support. AI tools can help with coding qualitative data, suggesting analytical frameworks, and explaining statistical concepts. They cannot replace your interpretive judgement, and any AI-assisted analysis must be documented and disclosed.
Peer review preparation. AI can help you anticipate reviewer questions, identify gaps in your argument, and stress-test your methodology section. The intellectual defence of your work remains entirely yours.

What AI cannot do in research — honestly

Understanding the limits is as important as understanding the uses. These are not temporary limitations that will be resolved in the next version — they reflect fundamental characteristics of how these tools work.

→AI cannot access most academic literature. Free AI tools do not have access to subscription journal databases. They may have been trained on some open-access papers, but their knowledge is incomplete, potentially outdated, and they cannot retrieve current publications. Always verify any paper or finding an AI cites.
→AI hallucinates citations. AI tools will confidently produce plausible-looking references that do not exist. Every citation an AI generates must be verified before use. This is not occasional — it is a known and consistent failure mode.
→AI cannot be an author. No current journal or funder accepts AI as a named author. Authorship requires accountability — the ability to stand behind the work, respond to correspondence, and take responsibility for errors. AI cannot do any of these things.
→AI cannot interpret your data. AI can describe patterns in data you show it, but interpretation — what this means for the field, how it relates to prior work, what it implies for theory or practice — is your intellectual contribution. Outsourcing that is outsourcing the research itself.
→AI does not know your field's standards. Different disciplines have different conventions for what counts as evidence, rigour, and appropriate argumentation. AI produces generically academic-sounding text that may not meet your discipline's specific standards.

Disclosure — what to declare and where

The landscape here is moving quickly. What was optional disclosure a year ago is becoming mandatory in many journals and funding contexts. The safest position is to disclose AI use unless you have a specific reason not to — and if you are unsure, to ask your editor or programme officer before submission.

Journal submissions

Most major publishers now require disclosure of AI tool use in the preparation of a manuscript. Check the specific journal's author guidelines — requirements vary between publishers and between journals within the same publisher. As a minimum, disclose in the methods or acknowledgements section: which tool was used, for what purpose, and what you then did to verify or substantially revise the output.

Grant applications

Funder requirements vary widely. Some have no specific policy. Others require disclosure if AI was used in preparing the application. A small number prohibit AI use in applications entirely. Check your funder's terms before submitting. If no policy exists, consider disclosing anyway — it is increasingly expected and reflects good research practice.

Research involving participants

If AI tools are used to process, analyse, or generate content involving research participant data, this should be considered in your ethics application. Free AI tools almost certainly do not meet the data security requirements for sensitive participant data. Check with your research ethics committee before using AI in any stage of participant-facing research.

Ready-to-use research prompts

These prompts are designed for research contexts. Each is structured to keep you in the intellectual driving seat while using AI for what it does well.

LITERATURE SCOPING

I am a researcher in [DISCIPLINE] working on [BROAD TOPIC]. I want to scope this area before conducting a systematic literature search. Please give me: 1. The five to eight key themes or debates currently active in this field 2. Ten search terms I should use in database searches — including discipline-specific terminology 3. Five adjacent fields or disciplines where relevant work may be published that I might not think to search 4. Three methodological approaches commonly used to study this topic Important: do not provide specific paper references or citations — I will conduct the literature search myself. I need orientation, not a reading list I cannot verify.

ARGUMENT STRESS-TEST

I am preparing a [journal article / conference paper / book chapter] arguing that [STATE YOUR CENTRAL ARGUMENT IN 2-3 SENTENCES]. Please act as a critical peer reviewer and identify: 1. The three strongest objections a reviewer in this field is likely to raise 2. Any assumptions in my argument that I may not have stated explicitly 3. Any evidence gaps that would weaken my case 4. One alternative interpretation of my position that I should address Do not rewrite my argument. I want to strengthen it myself based on your critique.

LAY SUMMARY

I need to write a lay summary of my research for a [grant application / public engagement section / institutional website]. My research [DESCRIBE YOUR RESEARCH IN 2-3 SENTENCES OF TECHNICAL LANGUAGE]. Please draft a lay summary of approximately 150 words that: — Explains what the research is and why it matters to a non-specialist reader — Avoids jargon, or explains it briefly if unavoidable — Makes the real-world significance clear without overstating it — Uses active voice and plain language throughout I will substantially revise and personalise this draft.