When peer review gets rushed, does weaker research slip into print, or is overloaded reviewing the deeper problem?

This explores whether rushing peer review, by giving reviewers less time or more papers to get through, lets weaker research into print. To be direct: none of these notes measures review deadlines against the quality of the papers that get published, so the corpus can't confirm a correlation. What it does offer is a sharper way to frame the question. Time pressure is usually a symptom of overload, and overload may cause damage that a correlation alone would miss.

The clearest mechanism is a feedback loop. A two-journal model in Does peer review quality collapse under submission overload? shows that when submissions rise, unpaid reviewers get stretched thin. Journals then either recruit less qualified reviewers or pile more onto existing ones, and review accuracy drops. The surprising part comes next. Noisier review makes it more worthwhile for authors to submit speculative work, because a weak paper now has a better chance of getting through. Submissions rise again and the cycle tightens. So squeezed timelines may do more than let a few bad papers slip by. They can change what authors decide to send in. The authors are candid that the mechanism is plausible but its real-world strength hasn't been measured.

AI speeds up both sides of this loop. A survey of 230 publications in Does AI create a coupled arms race in research production and review? describes production and review scaling together: AI makes papers faster to write, which pushes venues toward automating review, which invites gaming. The gaming is real. Can AI systems safely replace human peer reviewers? found that simply rewriting a paper's text raised AI review scores by almost half a point with no change to the science. And Can AI-generated papers pass peer review undetected? reports a fully AI-generated paper reaching acceptance scores at an ICLR workshop, although its own authors judged it not good enough for the main conference. Fast review that leans on shallow signals is where this kind of mismatch would be most likely to show up.

There's a useful counterpoint, though: speed and depth may not be a trade-off people have to make. An agentic reviewer in Can inference scaling help reviewers catch errors humans miss? spends extra computing time checking proofs line by line, and it caught errors in STOC and ICML papers that human reviewers had passed. A randomized trial at ICLR 2025 (Can LLM feedback help peer reviewers improve their own reviews?) found that AI feedback led 27 percent of reviewers to make their reviews more specific. Policy shortcuts look less effective. Banning LLMs from review barely changed outcomes at ICML 2026, and many reviewers broke the rule anyway (Does banning LLM use in peer review change review outcomes?).

The less obvious takeaway is that the costs of fast or skipped review don't stay inside journals. Can unreviewed preprints shape scientific debate before peer review? shows an unreviewed preprint shaping debate long before anyone questioned it. And because the problem spreads across authors, reviewers, and venues, Can two-stage review and badges fix AI conference peer review? argues for structural fixes, such as letting authors rate reviews before they see the verdict, instead of simply giving reviewers more time. If you want a hard number connecting review time to publication quality, this collection doesn't have one yet.