What happened
The number of published papers is growing exponentially at 5.6 percent per year, forcing journal editors to struggle finding willing reviewers. Steven Mack, an editor for Human Immunology, now needs to email around thirty researchers to find just one willing reviewer—five years ago it took only five to 10 emails to find three. Some journals are assigning reviews to unqualified or overextended reviewers as a result.
Why it matters
Peer review—the process by which researchers assess whether studies are valid enough to publish—is foundational to scientific credibility, but the system is faltering under strain. Researchers globally devote a collective 15,000 years of work to peer review annually, work that would cost $1.5 billion for the US share alone. When reviewers are rushed or misunderstand papers (as happened to health economist Jason Semprini, whose manuscript was rejected by a single reviewer who misread its central argument), important research gets blocked or delayed.
What to watch
Some AI researchers are abandoning traditional publishing altogether, turning instead to blogs and platforms like the AI Alignment Forum, which use community voting instead of peer review. Helen Qu, an AI researcher at the Flatiron Institute, now publishes only on her own blog; the AI Alignment Forum handles content through upvotes, downvotes, and reader comments rather than expert review. These alternatives are faster and potentially less dependent on opaque company algorithms, though the quality guarantee is weaker.
Summaries like this, in your inbox every morning.
Peer review is a surprisingly recent invention: although scientific societies had mechanisms for assessing submissions as far back as the early 1800s, the practice did not become universal until about 50 years ago. It became the cornerstone of science almost by accident, as a way to solve a credibility crisis. In the 1970s, when the US National Science Foundation faced political pressure to justify its spending, the agency's use of independent external reviewers to allocate grants provided an air of legitimacy that kept funding decisions out of direct political control. The United Kingdom underwent a similar moment in the late 1980s and early 1990s, when the public needed assurance that proper science could be distinguished from quackery. Over time, peer review became the gold standard—but the system was built for a much smaller research enterprise.
Today, the collision is stark. The exponential growth in papers, driven by more researchers, more journals, the shift to online publishing (which removes physical space constraints), interdisciplinary work, AI-assisted writing, and global submissions in English, has overwhelmed a volunteer-based system. Researchers like Sebastian Lourido from the Whitehead Institute receive roughly 10 peer review requests per month when realistically they have time for one or two. Under such pressure, editors are forced to assign reviews to people who lack expertise or time, leading to the kind of mishap that rejected Jason Semprini's research because a single reviewer misunderstood the paper's central argument. The system, in Semprini's own words, is "not standing strong." For AI researchers in particular, the overload is acute—submissions to top AI conferences have increased between two- and tenfold since 2019, and they are experimenting with alternatives that trade the legitimacy of expert review for the speed and transparency of community voting.
For example, today's edition would include:
AI-summarized, only the topics you pick: one digest a day via Email, LINE, or Slack.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
