AI tools in research and publishing: opportunities and limits

News · editorial note

Tools that generate text, images and analyses have entered research and publishing, and their use is being actively debated. They can speed up routine tasks and help authors work in a second language, yet they also raise questions about accuracy, authorship and disclosure. Journals have begun to set out what is and is not acceptable. This news summary outlines the opportunities and the limits, and what you should confirm on the journal's official pages.

Where these tools can help

Language tools can improve the clarity of a manuscript written by a non-native speaker, and routine drafting, summarising and formatting are tasks where assistance saves time. In analysis, software has long played a role, and newer methods extend what can be examined.

Used carefully, such tools can lower barriers and free researchers to concentrate on the science. The key word is carefully, because assistance is only useful when the output is checked.

Where they fall short

Automated tools can produce confident statements that are wrong, and they can invent references or details that do not exist. This makes verification essential, not optional. A fluent output is no guarantee of accuracy.

They also lack the judgement that research requires: understanding context, weighing evidence and recognising when a result is surprising. Those remain the author's responsibility, and no tool can take it over.

Authorship questions

A tool cannot be an author. Authorship implies responsibility for the content and accountability for any problems, which a machine cannot assume. Most guidance therefore holds that only people who meet the authorship criteria should be listed, and that the use of such tools should be disclosed where the journal requires it.

This position keeps responsibility where it belongs. Whatever assistance was used, the named authors remain answerable for the work.

Disclosure and honesty

Journals increasingly expect authors to state how such tools were used, so that readers can weigh the work with that in mind. The exact requirement varies: some ask for a statement, others prohibit certain uses outright. The direction of travel is toward transparency rather than prohibition.

Disclosure protects everyone. It prevents later questions about how a manuscript was produced and keeps the record honest.

Opportunities and limits at a glance

The table below summarises common uses and the caution each one deserves. The rules differ between journals and are changing.

Use	Potential benefit	Main caution
Language editing	Clearer writing	Check that meaning is preserved
Drafting support	Saves time	Verify every factual claim
Summarising	Faster reading	Risk of missing nuance
Analysis assistance	New ways to examine data	Results must be reproducible
Reference generation	Formatting consistency	Invented or wrong citations

What reviewers and editors face

Reviewers and editors are having to adjust as well. They may encounter reports or manuscripts produced with heavy reliance on such tools, and they need to judge the work itself while noticing signs of unverified output. Some journals discourage reviewers from putting confidential manuscripts into external tools, for reasons of confidentiality.

These are practical questions without settled answers. What matters is that the standards of evidence and confidentiality are maintained, whatever tools are involved.

Points for authors to keep in mind

For an author, the safe approach is transparency and verification. Use tools where they genuinely help, check everything they produce, and follow the journal's rules on disclosure. Assume responsibility for the final text, whatever assistance was used.

Check every claim, number and citation the tool produces.
Follow the journal's policy on declaring such use.
Keep confidential material out of external tools.
Remember that the named authors remain responsible.
Do not list a tool as an author.

Reading the trend sensibly

The debate is moving quickly, and policies are being revised as experience accumulates. The principles that hold steady are honesty, verification and responsibility. Tools may change what is efficient, but they do not change what counts as sound research.

Because journal and funder policies on this subject are set by them and change frequently, this summary is deliberately general. For anything that affects your own submission, confirm the current rules on the journal's official pages and follow the guidance that applies to your work.

AI-based tools can help with language, drafting and analysis, but they can also produce plausible errors and cannot bear responsibility. Authorship stays with people, verification is essential, and disclosure is increasingly expected. Policies are set by publishers and change quickly, so confirm the current rules on the journal's official pages.

Research integrity. Sound science rests on honest reporting, transparent methods and respect for ethical standards. Nothing on this site replaces the journal’s official instructions or the policies of your institution. When in doubt, confirm the current requirements with the editorial office and check the journal’s official pages.