AI Is Reshaping Scientific Publishing in 2026: Biology and Medicine Are Growing Fastest
Global scholarly output is on pace to cross 6 million articles in 2026, up from an estimated 5.5 million in 2025. That is not simply more researchers writing more papers. It is a measurable shift in how papers get written at all.
A December 2025 study in Science, based on 2.1 million preprints from arXiv, bioRxiv, and SSRN, found that researchers who adopted large language models increased their personal output by 23% to 89%, depending on field and language background. Life sciences authors publishing on bioRxiv saw a 52.9% productivity jump, one of the largest gains of any discipline measured, behind only social sciences and humanities.
For anyone preparing a manuscript in biology or medicine in 2026, this is not background noise. It changes how crowded submission queues are, how fast competing labs publish, and how closely journals now scrutinize what lands in their inbox.
The Numbers Behind the Surge
The productivity data is only part of the picture. Submission volume itself is climbing:
- An analysis of ten years of data from three major publishers found submissions up 56% since the release of ChatGPT, a rise the researchers attributed largely to AI-assisted writing;
- Submissions to Organization Science rose 42% between November 2022 and February 2026. By that point, a majority of submissions used AI to some degree, and the share of papers where more than 70% of the text was AI-generated more than doubled compared to early 2024;
- Broader researcher surveys put AI tool adoption for writing and publication tasks at 58% to 62% in 2026, up from 45% in 2024.
For a deeper look at how these shifts are reordering the publishing landscape overall, see our breakdown of academic publishing trends for 2026.
Why Biology and Medicine Are Outpacing Most Fields
Preprint activity tells the clearest story. bioRxiv and medRxiv, now run by the independent nonprofit openRxiv, had their busiest year on record in 2025: 64,229 new preprints combined, up from more than 56,000 in 2024. Roughly 80% of these are eventually published in peer-reviewed journals, typically within 250 days of posting.
The AI-adoption pattern inside that growth is unusually pronounced in biomedicine. An analysis of manuscripts submitted to the American Association for Cancer Research in 2025 found that 36% of 7,177 submissions were flagged as containing AI-generated text, described as a new record. A separate study of more than two million full-text biomedical publications on PubMed Central found that AI-assisted writing adoption grew 400% in non-English-speaking countries after ChatGPT’s release, compared to 183% in English-speaking countries, a gap that is inversely correlated with national English-proficiency scores.
That last point matters for a large share of our readers. Non-native English-speaking researchers are seeing the largest individual productivity gains from AI tools, but they are also more likely to have legitimate AI-assisted writing misidentified as fraudulent by detection software, since several widely used detection tools show measurable bias against non-native phrasing patterns.
The Tools Actually Driving the Work
The acceleration is not limited to drafting text. AI now touches nearly every stage of the biomedical research pipeline:
Literature review and synthesis: platforms such as Elicit, Consensus, SciSpace, and Scopus AI, plus newer end-to-end systems like otto-SR, are compressing systematic review timelines that once took months.
Drug discovery: AlphaFold 3 and open-source successors such as Boltz-2 and Chai-1 have made protein structure prediction routine rather than exceptional. Isomorphic Labs moved its first fully AI-designed drug into Phase 1 oncology and immunology trials in January 2026. Insilico Medicine used AlphaFold-predicted structures to identify a CDK20 inhibitor candidate in 30 days, a process that traditionally took years.
Medical imaging: more than 770 AI-based medical devices focused on radiology have FDA clearance. AI ambient scribes for clinical documentation are now common enough that physicians report up to 83% less time spent writing notes.
Clinical data analysis: adoption here is real but earlier-stage. A review of more than 500 clinical AI studies found that only 5% used real clinical data, while nearly half relied on exam-style benchmark questions instead, a gap worth keeping in mind when AI-driven clinical claims appear in a manuscript you are reviewing or citing.
More Researchers, More Competition for the Same Journal Slots
Faster writing does not translate into more room in high-quality journals. It translates into more submissions chasing the same acceptance rates, and reviewers who are increasingly primed to scrutinize AI-polished prose rather than reward it. The Science study noted a counterintuitive result here: manuscripts with more linguistically complex, AI-smoothed language were less likely to be accepted, not more, suggesting reviewers are already adjusting for it.
We cover exactly how this is playing out for individual authors in our analysis of rising competition for journal slots.
The Integrity Problem AI Created
The same acceleration has produced a documented integrity crisis. A Northwestern University study published in PNAS in August 2025 found that suspected paper-mill output is doubling every 1.5 years, roughly ten times faster than the 15-year doubling rate of legitimate scientific output. Retractions are not keeping pace: they double only every 3.3 years, and of more than 32,000 fraudulent articles identified in the study, just 29% had been retracted.
A related and newer threat is AI-hallucinated citations. Researchers tracking 111 million citations across four repositories, arXiv, bioRxiv, SSRN, and PubMed Central, estimated nearly 147,000 hallucinated citations entered the record in 2025 alone. A Lancet audit published in May 2026, covering 2.47 million biomedical papers, found the rate of fabricated references climbed from about 1 in 2,828 papers in 2023 to 1 in 277 by early 2026. Review articles, which underpin clinical guidelines, showed a 57% higher fabrication rate than other paper types.
This is the same landscape we mapped out in how predatory journals trick researchers: AI has not created a new category of risk so much as it has scaled up an old one, faster than editorial teams can screen for it. Journals that cannot keep pace with integrity screening are also the ones most exposed to the kind of scrutiny we detailed in how Scopus actually decides to remove a journal.
How Journals and Publishers Are Responding
Governance has caught up somewhat in the past year:
- The ICMJE’s January 2026 update added a full new section on AI use: AI cannot be listed as an author, authors must disclose substantive use, and reviewers are barred from uploading manuscripts to AI tools without explicit permission;
- COPE’s September 2025 retraction guidelines now treat undisclosed AI use as a form of misrepresentation, grounds for retraction on its own;
- Wiley, Elsevier, and Springer Nature have all issued or updated disclosure requirements between September 2025 and June 2026. Springer Nature alone runs close to 60 internal AI screening and support tools across its workflow, one of which flagged 25,000 papers for integrity issues in 2025.
The gap between policy and practice is still wide. Roughly 70% to 83% of journals now have a formal AI policy, yet full-text sampling suggests only about 0.1% of post-2023 papers explicitly disclose AI use, a sign that enforcement is lagging well behind the rules on paper.
Conclusion
AI has not just made scientific publishing faster. It has changed who gets published, how quickly, and how hard that work gets checked before and after it appears in print. For early-career researchers in biology and medicine, three things follow directly from the data above: verify every citation an AI tool hands you, disclose AI use according to your target journal’s policy rather than guessing, and treat a journal’s indexing and integrity track record as something to check, not assume.
That last point is where choosing where to submit starts to matter as much as how fast you can write. A journal that is transparent about its review process, its indexing status, and how it screens submissions is worth more than a faster acceptance from one that is not.
Submit to Futurity Medicine
Futurity Medicine (E-ISSN 2956-672X) is an international peer-reviewed open access journal, indexed in Elsevier’s Embase and Embiology, Crossref, Index Copernicus, Google Scholar, EuroPub, and Sherpa Romeo.
Submissions for the September 2026 issue are open now.
Deadline: September 20, 2026.
