AI writing assistants have quickly become part of many researchers’ toolkits. They can explain a difficult concept, suggest better wording or help organise a messy draft. But they also raise real questions about honesty, accuracy and authorship. Journals and universities are responding with new policies, and authors need to understand where the lines are.
What AI tools can reasonably help with
Used carefully, AI tools can support your work in several ways:
- Language polishing: improving grammar, sentence flow and clarity, especially for authors writing in a second language.
- Understanding concepts: getting a simple explanation of a statistical test or a theory before reading the original literature.
- Brainstorming: listing possible angles, keywords or questions to explore.
- Structuring: suggesting an outline for a paper you have already researched.
- Technical tasks: help with formatting, or understanding how to use data analysis software.
Where the risks are
1. Invented references
AI tools can produce references that look real but do not exist, with plausible authors, titles and journal names. Never include a reference you have not personally found and checked.
2. Factual errors
AI-generated text can state wrong figures, dates or findings with complete confidence. Every fact in your paper is your responsibility.
3. Hidden plagiarism
Generated text may closely resemble existing published work without telling you where it came from.
4. Loss of your own voice and thinking
A paper is supposed to show your analysis and interpretation. If a tool writes the argument, the paper no longer reflects your understanding, and reviewers can often tell.
5. Confidentiality
Uploading unpublished data, interview transcripts or a manuscript you are reviewing to an online tool may breach confidentiality or ethics approvals. Reviewers in particular should not upload manuscripts they have been asked to review.
What most journals now expect
Policies vary, but common principles are emerging across publishers and editorial organisations:
- AI cannot be an author. Authorship requires taking responsibility for the work, which a tool cannot do.
- Disclose significant use. Many journals ask authors to state, in the methods or acknowledgements, if and how AI tools were used beyond basic spelling and grammar correction.
- Authors remain fully responsible for the accuracy, originality and integrity of the entire paper.
- AI-generated images and data are often restricted or not permitted unless they are part of the research method and clearly described.
Always read the specific policy of the journal you are submitting to, and the guidelines of your university.
A simple rule of thumb
Ask yourself: “Could I explain and defend every sentence, figure and reference in this paper without the tool?” If the answer is yes, and you have disclosed the tool’s role honestly, you are likely using it appropriately. If the answer is no, the tool is doing work that should be yours.
Good practice checklist
- Do the reading, data collection and analysis yourself.
- Use AI for language and clarity, not for generating findings or arguments.
- Verify every reference in the original source.
- Check every fact and number.
- Keep unpublished and confidential material out of online tools.
- Disclose AI use according to the journal’s policy.
Technology will keep changing, but the basic expectations of research stay the same: honesty, accuracy and original thinking. Tools that support those values can be useful; tools that replace them put your work at risk.
— Editorial Team, Imaginative Ink Publications