On 9 November 2026 PubPeer, the site where researchers comment on published papers, launched a redesign that includes a dashboard of running totals. Two days later, on 11 November, Retraction Watch reported on a small survey study of how authors respond to PubPeer comments, and set it beside the figures on PubPeer’s new statistics page. Read together, the two items give research offices and editors a clearer picture of what post-publication peer review does and does not change.
What happened
- The redesign. PubPeer launched its redesigned site on 9 November 2026. According to Retraction Watch, the new statistics page shows more than 500,000 comments over 14 years. PubPeer co-founder Boris Barbour said the redesign shows running statistics more clearly, which may make the platform’s impact more visible.
- The statistics. Retraction Watch reports that PubPeer’s statistics page shows 20,784 papers that had comments before they were retracted, with a median of 10 months between the first comment and the retraction. It also shows more than 2,500 papers with comments a median of 11 months before an expression of concern, and 12,000 papers with discussion before a correction. For context, Retraction Watch notes its database lists 59,843 retractions since PubPeer launched in October 2012.
- The study. Separately, a preprint posted earlier in 2026 on MetaArXiv surveyed authors about PubPeer comments. It was led by Wytske Hepkema, a PhD student at Radboud University, with colleagues in the NanoBubbles group, which studies post-publication peer review; co-author Raphaël Lévy is a professor of physics at Université Paris Sorbonne Nord. The 20,784 figure is PubPeer’s own count, not a result of this survey.
What the survey did and found
As reported by Retraction Watch, the team contacted nearly 100,000 authors: more than 10,000 whose papers had PubPeer comments, and nearly 85,000 who had cited a commented paper. Recipients were chosen from comments that, if true, would affect a paper’s results or content, such as image concerns or “tortured phrases” (odd paraphrasing that can indicate plagiarism or paper-mill output). Retracted papers were excluded.
- The response rate was under 1%: 80 authors of commented papers and 720 authors who had cited one.
- Most respondents in both groups said the comments did not affect their own study’s results.
- About four in five authors of commented papers knew about the comments; those authors were more willing to accept that a comment had identified a real error, and over half said they had asked the journal for an erratum.
- Over half of the citing authors said the criticisms were not relevant to their work.
- Only 10% to 15% of authors who knew of comments replied on the platform.
- Some respondents were strongly critical of PubPeer, including of commenter anonymity.
The authors conclude that PubPeer’s effect operates “through retractions rather than through behavioural change in researchers,” as Retraction Watch quotes them.
Limitations as stated
The response rate is very low. Hepkema said the team was not surprised, described the work as a “first view” of researcher perspectives, and said 80 responses are not enough to generalise about researchers’ opinions. José Luis Ortega of CSIC, who was not part of the study, told Retraction Watch the results may suffer from response bias, since authors in more serious cases may have been less likely to answer, and that citing authors may regard citing a flawed study as acceptable or as needing no action.
The PubPeer figures also need careful reading. They count papers that had comments before a later retraction; they do not show that the comments caused the retractions. Retraction Watch’s reporting does not claim causation. Barbour and co-founder Brandon Stell said the survey findings “largely align” with their experience, and acknowledged that only a modest share of scientists include PubPeer in their workflows. Disclosure from Retraction Watch: its co-founder Ivan Oransky is a volunteer on the PubPeer Foundation’s board.
What it means for research integrity offices, editors and authors
The following is editorial interpretation, not a finding of the study.
- Integrity offices. A PubPeer thread is a signal, not an allegation that meets a formal threshold. A median lead time of about 10 months between first comment and retraction suggests that a comment can be an early prompt for screening under an institution’s own research misconduct procedure, but only an assessment of the underlying data can establish a finding.
- Editors. The survey suggests many authors do not respond on the platform, so editors who receive a concern cannot assume the authors have already seen or answered it. Journal-side processes, such as those described in COPE guidance, remain the route to a correction, expression of concern or retraction.
- Authors. Respondents who knew about comments were likelier to accept a real error and to request an erratum. Monitoring comments on one’s own papers, and replying where appropriate, is the practice the survey data most directly point toward; citing authors may also want to re-check work they depend on when a cited paper is flagged.
Sources
- Retraction Watch, “Do PubPeer comments make a difference? A small study tried to find out”, 11 November 2026 (updated 12 November).
- Hepkema and colleagues, preprint on MetaArXiv (not peer reviewed at the time of reporting).
- PubPeer statistics page, as cited by Retraction Watch.
