Artificial intelligence has the potential to make scientific research more efficient, but AI itself can also hinder progress by contributing to an increase in low-quality research papers.
According to a recent study, over 13% of biomedical paper abstracts submitted globally in 2024 appear to have been written with the aid of AI.
Although AI has made writing academic papers easier, the surge in submissions has increased the difficulty of content screening, potentially facilitating the spread of misinformation and hindering the publication of higher-quality research.
In October 2025, arXiv, an open-access online research repository, tightened its posting rules for some computer science papers after a surge in submissions following the emergence of generative AI, which many experts believe has led to an increase in low-quality research.
Researchers present their research findings in the form of scientific papers. Initially, they post their work online for peer review, and later, the papers are published in scientific journals. Researchers gain more recognition as more of their papers appear in leading journals.
The application of AI in scientific research offers numerous benefits. Building upon previous research, AI can generate new hypotheses and analyze experimental data, thereby improving research efficiency and accelerating scientific progress. AI can also reveal insights from perspectives that humans may never have considered.
However, AI can also create problems, as it can generate natural-sounding, human-like text and even write research papers that were previously only done by experts, but the surge in submissions is putting increasing pressure on the academics responsible for reviewing them. A total of approximately 370,000 papers were published on arXiv and the two other major research sharing platforms, bioRxiv and medRxiv, in 2025, a 16% increase from the previous year. Experts believe this increase in submissions is due not only to the growing research community but also to AI itself.
"Leading academic societies in artificial intelligence are facing a shortage of reviewers amid a surge in submissions," said Professor Isao Echizen at the National Institute of Informatics in Japan. While AI can assist in the review process, human researchers remain responsible for the final evaluation, a time-consuming and labor-intensive process.
Studies are underway to identify scientific papers written with the aid of AI. In July 2025, a research team at the University of Tübingen in Germany published a report in an American scientific journal, citing a significant increase in the frequency of use of certain words in biomedical research worldwide . Analyzing 15 million research paper abstracts, the team found a sharp increase in the use of words such as "those" and "important" since OpenAI released its AI chatbot, ChatGPT, in 2022. The research team estimated that AI generation had been used in at least 13.5% of biomedical papers by 2024.
Commenting on this finding, Shota Imai, a visiting professor at the Japan Advanced Institute of Science and Technology, said: "Artificial intelligence has inherited the disproportionate use of certain words from its training materials."
Researchers are still debating the extent to which AI should be used in writing scientific papers. According to a survey of 5,000 researchers worldwide conducted by the prestigious journal Nature in March 2025, over 90% supported the use of AI for proofreading and other editing, but this percentage dropped to 65% when it came to using AI for writing papers.
Currently, Nature magazine does not recognize AI tools as authors or co-authors of scientific papers. The journal requires researchers to take responsibility for the content of their articles and to disclose the use of AI in data analysis and other work.