AI × Research · 2026
AI is now part of every stage of research, including peer review
Authors write with AI, reviewers review with AI, and journals screen with AI. When polished prose is no longer rare, what gets a manuscript past the editor and out to review? And where is AI-driven research heading?
See the data Get in touch01 · Where we are
AI use in research is already widespread
The figures below come from recently published surveys and studies, compiled by SciAutoLab. Follow each source to the original report.
of surveyed researchers use AI in their work, up from 57% a year earlier.
of surveyed peer reviewers use AI when reviewing, mostly to draft reports and summarise findings.
of ICLR 2026 reviews were flagged as fully AI-generated, and more than half showed signs of AI use.
of 2024 PubMed abstracts are estimated to have been processed with a large language model.
These studies differ in samples, fields and detection methods, so the numbers are not directly comparable. The direction is consistent, though: AI is now present on both the writing side and the review side.
02 · Desk review
When AI makes writing look alike, how do you get sent out to review?
Many journals reject a large share of submissions at the editor’s first screening; some report desk-rejection rates of 60–80% [5]. As AI lifts the baseline of language quality, editors look harder at everything beyond the prose.
- Make the question and contribution clear on page oneEditors often decide from the title, abstract and introduction. The research question, what is new and the main result should be obvious up front.
- Target the right journalFit the journal’s scope and engage with what it has published recently, so the paper is not simply at the wrong door.
- Be transparent and reproducibleSpell out identification, data sources and robustness checks, and make code and data reproducible where possible. That credibility is hard for AI to fake.
- Disclose AI use honestlyMost publishers ask authors to describe how AI was used, and AI cannot be listed as an author. Proper disclosure is safer than silence.
- Run a mock review before submittingUse AI for a first-pass check of logic and wording, then have someone who knows the field judge which points really matter.
The rules on the review side are tightening too
Elsevier, for example, asks reviewers and editors not to upload manuscripts into generative AI tools and not to use generative AI to evaluate them; responsibility for the review stays with people [6]. What decides a paper’s fate is still human judgement about the value of the research.
03 · What next
Does AI-driven research have a future?
It depends on what we mean by “AI research”.
Execution is being automated fast
Literature work, data processing, modelling, writing and typesetting can all be sped up considerably. In 2025 a paper generated end to end by an AI system passed peer review at an ICLR workshop [7].
Judgement and trust remain scarce
That paper was a workshop submission, withdrawn before publication, and it contained citation errors. Researchers’ concern about inaccuracy and hallucination is also rising, from 51% to 64% [1].
Value is moving upstream
As execution gets cheaper, good questions, distinctive data, rigorous identification and accountability for results become a researcher’s real edge.
Our view: AI research has a future, and it looks like people asking the questions and making the calls, with AI speeding up the work, not AI replacing the researcher.
Take part in the SciAutoLab researcher survey
We are preparing a survey of authors and reviewers on how AI is actually used in submission and peer review. If you would like to take part or share your experience, please get in touch.
Sources
- Wiley. AI Adoption Jumps to 84% Among Researchers… (October 2025)
- Frontiers. Most peer reviewers now use AI… (December 2025)
- Plagiarism Today. AI Conference Overrun with AI Peer Review (December 2025, reporting Pangram Labs’ analysis)
- Kobak D. et al. Delving into LLM-assisted writing in biomedical publications through excess vocabulary. Science Advances, 2025. arXiv version
- Jawaid S.A., Jawaid M. Common reasons for not accepting manuscripts… Pakistan Journal of Medical Sciences, 2019.
- Elsevier. The use of generative AI and AI-assisted technologies in the review process
- TechCrunch. Sakana claims its AI paper passed peer review… (March 2025)