Research papers are increasingly being written by artificial intelligence.Credit: Yagi Studio/Getty
How much of the scientific literature is generated by AI? The first studies of the size of the AI footprint in scientific journals, preprint repositories and peer-review reports give a spread of answers — and indicate a rapidly evolving situation that it is difficult to get a handle on.
The fear of many in the research community is that poor-quality or entirely fabricated research produced by large language models (LLMs) could overwhelm the ability of current quality-control systems to detect it, thereby polluting the scientific canon.
“The ground is shifting underneath us in ways that we are totally unprepared for,” says Maria Antoniak, a computer scientist at the University of Colorado Boulder.
“We live in an escalating arms race” between people using AI unscrupulously and those who are trying to constrain or detect it, says Richard She, a stem-cell biologist at Nanyang Technological University in Singapore.
Source: ref 1.
AI detectors
Concerns about the extent of AI-generated content in the scientific literature mirror broader online trends. At the end of March, AI-generated articles were estimated to outnumber those written by humans, according to an analysis of 55,000 newly published webpages shared with Nature by the private firm Graphite in San Francisco, California.
LINK AI tool detects LLM-generated text in research papers and peer reviews
AI might have legitimate uses in the production of scientific literature, and can accelerate research progress. But AI-generated content is also potentially problematic because it can be used to create fake or low-quality papers.
To investigate this, researchers are turning to AI-detection tools to measure the scale of the issue. Some of the available tools don’t distinguish between text that was merely edited with AI, and text that was wholly generated by it. The systems also have varying ways of defining what counts as ‘AI generated’, and can falsely flag human-written text as AI generated.
Nevertheless, they can provide some pointers as to what the trends in the use of AI are.
Droplets in a storm
In a study published on 27 April1, researchers used a tool developed by Pangram Labs in New York City to scan nearly 7,000 manuscript abstracts submitted to the journal Organization Science between January 2021 and February 2026, along with some 8,000 peer-review reports. It is the first analysis to estimate the overall amount of AI-generated content in a research journal’s review process, the authors say.
The study reported a 42% increase in submissions since November 2022 — when ChatGPT was released as the first LLM available to the general public — and found that the increase was driven mainly by AI. The authors also estimated that by February this year, submissions with more than 70% AI-generated text had more than doubled compared with the numbers seen in early 2024 (see ‘AI-written text in scientific journals is rising’), and more than 30% of peer-review reports also contained some AI-generated text.
Other researchers, including She and Antoniak, have been attempting to record the total amount of AI-generated research content that currently exists online — a near-impossible task owing to the sheer volume of papers involved. She used Pangram’s AI detection tool to screen some 5,000 biomedical science papers published last year in journals including Science, Nature and Cell.
His analysis — published in a January preprint2 — found that six papers were flagged as fully AI-written, but one in eight articles contained some AI-generated text.
She expects the rate to go up in the coming years. “We’re at the very, very beginning of this new era. What we’re seeing is the first droplets of a storm that’s incoming,” he says.
In another preprint published in January3, Antoniak and her colleague used two AI-detection methods to screen more than 124,000 manuscripts posted on arXiv between 2020 and 2025. They found that for computer science, review preprints containing AI-generated text increased from about 7% in 2023 to 43% in 2025. Non-review manuscripts in this field that contained AI-generated text also grew from around 3% to 23% during the same period.