AI tools for researchers in the UK: what to use them for, and what to check

Most lists of AI tools for researchers in the UK start with products. This guide starts with the work. Researchers, PGRs, and research office staff use AI across six jobs: finding literature, summarising papers, analysing data, writing, grant applications, and admin. For each one, it covers what generative AI in research does well, where it goes wrong, and what to check, along with what UKRI and the major publishers allow.

Can AI help with a literature review?

Yes, for finding and screening. Specialist search tools work over real paper databases rather than generating text from memory. Elicit and Consensus answer a research question with a list of papers and short summaries. Semantic Scholar is a free search engine with AI-generated paper summaries. ResearchRabbit maps citation networks, so you can see what a key paper led to. Scite shows whether later papers supported or disputed a finding.

Where it goes wrong. Coverage. Each tool searches its own index, which may miss paywalled journals, books, grey literature, and much of the humanities. A tool that says “no evidence found” may mean the evidence isn’t in its database.

What to check. Treat AI search as a supplement to your usual databases, not a replacement. If you’re doing a systematic review, record which tools you used and when, because your search needs to be reproducible.

How to use AI to summarise and compare papers

General assistants such as Microsoft Copilot or NotebookLM can summarise a paper you upload, pull out methods and sample sizes into a table, or compare how three papers define the same concept. That’s useful for triage and for organising notes across a large reading list.

Where it goes wrong. Summaries flatten nuance. A hedged finding comes back as a firm one, and limitations sections tend to disappear. Ask a general chatbot for sources without uploading them and it may invent them: an early study found that 55% of GPT-3.5’s citations and 18% of GPT-4’s were fabricated¹. Newer tools do better, but the risk hasn’t gone.

What to check. Open every reference before you cite it. Read the original before you rely on a summary for anything you’ll publish. In our courses we call AI the Super Parrot: it has read a great deal and repeats patterns well, but it doesn’t know whether what it says is true.

Can AI help with data analysis and coding?

This is one of the strongest uses. AI can write and explain R, Python, or SPSS syntax, debug an error message, suggest a suitable statistical test, and draft a first pass at coding qualitative data.

Where it goes wrong. Code that runs isn’t the same as code that’s right. AI can choose the wrong test, drop missing values without saying so, or apply a coding frame inconsistently. The bigger risk is data. Unpublished data, participant data, and anything covered by your ethics approval or data management plan shouldn’t go into a tool your institution hasn’t approved for it.

What to check. Ask AI to write the code, then run it yourself on your data inside your own environment. Test it on a small synthetic dataset first. If you use AI in your method, describe it in your methods section. Our guide to responsible AI in higher education sets out the wider data rules.

What do journals allow when you use AI for writing?

AI is a capable editor. It can tighten an abstract, restructure a paragraph, suggest a clearer title, or turn notes into a draft outline. The publishers’ rules are broadly consistent:

• AI can’t be an author. Elsevier and Springer Nature both say so²³.
• Substantive use must be disclosed. Elsevier asks for a separate AI declaration before the references, naming the tool and its purpose. Springer Nature asks for LLM use to be documented in the methods section²³.
• Basic editing is usually exempt. Grammar, spelling, and punctuation checks don’t need a declaration at Elsevier, and Springer Nature exempts AI-assisted copy editing²³.
• Reviewers mustn’t upload manuscripts. Elsevier tells reviewers not to put any part of a manuscript into an AI tool².

Rules vary by journal and by image type, so read the author guidelines for each journal you submit to.

What to check. Every claim, every number, and every citation in an AI-edited draft is still yours. Keep a note of what you used AI for as you go, so the disclosure statement is easy to write.

Can you use AI to write a grant application?

For UKRI funding, yes, with conditions. UKRI’s policy lets applicants use generative AI, with caution, for tasks such as generating ideas, summarising literature, and drafting abstracts⁴. Applicants mustn’t use it to generate an entire application, or sections of one, without human involvement. Substantive use should be declared, and UKRI says disclosure doesn’t affect the assessment. Minor uses, such as improving English or cutting to a word count, don’t need declaring.

The rules for reviewers are stricter. Assessors mustn’t put any part of an application into a generative AI tool, or use one to summarise or evaluate it⁴. UKRI is trialling AI-assisted assessment from October 2026, but as a test against past decisions, not a change to the rules⁵.

What to check. Read the specific funder’s policy every time: charities and other funders set their own. Never paste a collaborator’s unpublished ideas or partners’ confidential material into an unapproved tool.

How can research office staff use AI?

Research support and research office staff can save time on the routine end of the work: drafting funding call summaries, tidying meeting notes, turning a policy into an FAQ, first drafts of impact case study structures, and checking a draft against a funder’s guidance notes.

Where it goes wrong. Admin is full of personal and commercially confidential material: costings, partner agreements, and peer review comments. The same data rules apply to a research office as to a lab.

What to check. Use your institution’s approved tools, signed in with your work account, and anonymise before you paste. If your institution hasn’t yet set rules for research, our guide to writing a university AI policy covers what it needs.

How to build AI skills across a research community

Researchers rarely need another tool list. They need judgement about which task suits AI, what the funder and journal allow, and how to check what comes back. That’s why research support works best as part of a wider programme, as our guide to implementing AI at your university sets out, built on the foundation in AI literacy training for staff.

Our AI for Researchers course is pitched from foundation to intermediate. Its five learning outcomes are to apply AI to support literature review, summarisation, and synthesis; generate research questions, abstracts, and structured outlines using responsible prompting; organise and manage large volumes of research text and notes with AI support; evaluate AI-generated research content for accuracy, traceability, and methodological validity; and improve research efficiency without compromising scholarly standards. It can run as a workshop or a self-paced course built around your own policy, and we also offer discipline-specific coaching for research teams of up to six.

Find out where your researchers are

The quickest way to see how your research staff use AI now, and where to start, is our free AI Capability Check. It’s a 30-minute call over Teams or Zoom, with no obligation. We’ll place your team on the AI Use Hierarchy, give you two or three prioritised next steps, and send you a written report to keep.

Book your free AI Capability Check

You can also explore the courses, including AI for Researchers.

Frequently asked questions

What is the best AI tool for a literature review? There’s no single best tool. Search tools such as Elicit, Consensus, and Semantic Scholar find papers in real databases; general assistants summarise papers you upload. Use them alongside your usual databases, not instead of them.

Does UKRI allow AI in grant applications? Yes, with caution. Applicants can use generative AI but mustn’t use it to write an entire application without human involvement, and should declare substantive use. Reviewers mustn’t put applications into AI tools.

Do I have to declare AI use in a journal article? Usually, for anything beyond basic editing. Most major publishers ask for a disclosure statement or a note in the methods, and none accept AI as an author. Check each journal’s author guidelines.

Why does AI make up references? General chatbots predict plausible text rather than looking sources up, so they can produce a citation that looks right and doesn’t exist. Open every reference before you cite it.

Can I put research data into ChatGPT? Not unpublished, personal, or ethics-controlled data in a tool your institution hasn’t approved. Use approved tools, anonymise first, and follow your data management plan.