In this third installment of my summary of the 10th International Congress on Peer Review and Scientific Publication I focus on research integrity. Delegates at the Congress heard evidence of paper mills pumping fraudulent science into journals, fake reviewers infiltrating editorial systems, and image manipulation contaminating the literature. But alongside these threats, there were also signs of progress: new tools and cross-publisher collaborations.
This post summarizes some of the evidence and innovations presented at the Congress.
Paper Mills: Industrial-Scale Misconduct
One of the most striking studies detailed the scope of the Tanu.pro paper mill. Investigators linked 1,517 fraudulent papers across 380 journals, involving more than 4,500 scholars from 46 countries. Springer reported receiving 8,432 submissions tied to Tanu.pro, most of which were detected and rejected, but nearly 80 still made it into print.
Another study mapped 26 fake author identities, half of which managed to publish papers and later acted as reviewers. These personas were not isolated actors: they cited each other’s fabricated work, building closed networks of validation that fooled editorial systems.
The message was clear: paper mills are no longer small-scale fraud factories. They operate globally, infiltrating authorship, peer review, and citation ecosystems.
Image Manipulation: A Persistent Plague
Paper mills are the industrial threat, and image manipulation is part of the raw material.
- An analysis of 8,002 retracted papers found that gel blots were the most frequently manipulated images, with copy-paste duplications, spliced lanes, and fabricated bands accounting for the majority of cases. Nearly half of these retractions were tied to paper mills, underscoring the overlap between industrial and individual misconduct.
- A forensic scan of 11,314 materials-science papers containing scanning electron microscopy (SEM) images found that 2% had mismatched instrument metadata. These papers were significantly more likely to contain analytic errors, linking poor reporting with potential fraud.
Such findings show why image checking is becoming a frontline defense. Several publishers reported piloting or adopting automated tools that flag duplications or anomalies before peer review even begins.
Authorship Manipulation: Who Wrote This Paper?
Authorship problems also surfaced repeatedly.
- Taylor & Francis audited 452 manuscripts where three or more authors were added post-submission. 81% of these requests were denied, and denied requests strongly correlated with other suspicious markers such as “tortured phrases” and undeclared conflicts of interest.
- In some cases, AI tools were incorrectly listed as co-authors, raising questions of accountability and transparency.
The Congress echoed a call for stricter contributor accountability through CRediT roles, ORCID verification, and better affiliation checks.
Signals of Fraud and Integrity Problems
Beyond the cases of paper mills, manipulated images, and author accountability, several presentations highlighted the early warning signals that can indicate deeper problems with research integrity.
- One study found that countries with higher scores for rule of law, press freedom, and governance tended to have more retractions, even when misconduct was suspected. In contrast, nations with stronger democratic institutions had lower retraction rates, not because they had less fraud, but because they were better at detecting and correcting it.
- Citation data can also reveal red flags. One presentation showed how suspicious citation concentration, whether around a single author, a small cluster of journals, or coordinated self-citation networks, can indicate manipulation. These patterns may reflect coercive citation practices or citation cartels.
- Fraudulent organizations are often highly resilient and adaptive. A study showed how fraud networks adjust when detection tools improve, shifting to new tactics, repackaging services, or even using more sophisticated AI tools to evade detection. A key signal is repeated, slightly altered submissions or recycled fraudulent methods appearing across journals.
Taken together, these studies point to the importance of looking not just at the blatant cases of misconduct, but also at the patterns and signals that suggest deeper systemic risks: a suspicious absence of retractions, abnormal citation clustering, or repeat submissions that look “too similar.”
Retractions: Too Few, Too Late
Even when misconduct is caught, the system often fails to correct the record. One study showed that fewer than 25% of known paper-mill articles are ever formally retracted. Where retractions did occur, notices varied widely in clarity: some publishers provided detailed reasons, while others issued vague statements that did little to protect readers.
Several studies show that without consistent retractions, fraudulent work continues to be cited and incorporated into systematic reviews and meta-analyses, compounding the damage.
- Radiation Oncology: 34 retracted papers were cited 576 times after retraction; 92% of citing studies treated the work as legitimate (Hamilton et al. 2019).
- Exercise Physiology: 9 retracted papers were cited 469 times after retraction; in a 20% sample of citing manuscripts, none acknowledged the retraction (Hagberg et al. 2020).
- Dentistry: 180 retracted papers were cited 530 times after retraction, with 89.6% of citations judged inappropriate (Rapini et al. 2020).
- Anesthesiology: 478 retracted papers were cited 1,402 times after retraction; only 10.8% of surveyed authors were aware the cited papers had been retracted (Cassai et al. 2022).
- COVID-19 literature: 212 retracted papers were cited ~650 times after retraction; 80% of total citations were uncritical and treated the retracted work as valid (Meyerowitz-Katz et al. 2022).
Across disciplines, retracted papers continue to circulate in the literature, with most citations failing to flag the retraction.
Integrity Audits at Scale: The PLOS Example
Amid the grim data, there were also encouraging signs of progress. PLOS described how it has scaled integrity screening across its entire portfolio.
Between 2021 and 2025, PLOS combined:
- STM Integrity Hub checks for duplicate submissions,
- plagiarism and image analysis,
- audits of contributor behavior, and
- targeted pre-review screening for study types prone to misuse, such as Mendelian randomization and systematic reviews.
The result: desk rejection rates rose from 13% to 40%, filtering out poor-quality or fraudulent papers before they reached peer reviewers. This not only protected the literature but also conserved reviewer capacity, a precious resource in today’s over-stretched ecosystem.
Technology, Collaboration, and Policy
What distinguished the 2025 discussions from earlier congresses was the sense that solutions are finally starting to scale.
- Cross-publisher collaboration through the STM Integrity Hub enables journals to detect duplicate or simultaneous submissions across portfolios.
- Automated tools are increasingly applied at submission: image forensics, duplication detectors, and AI-based reporting guideline checks.
- Policy innovations like mandatory CRediT roles, stricter checks on author-addition requests, and enhanced retraction notices aim to close loopholes that paper mills exploit.
Still, participants stressed that no tool or policy is foolproof. Fraudsters adapt quickly, and automated detectors must be combined with vigilant human oversight.
Human Judgment Remains Central
One refrain across sessions was the need to keep humans in the loop. AI can spot anomalies, such as outcome switching in trial registries or duplicated gel images, in minutes. But human experts are still needed to interpret context: is a deviation a fatal flaw, a reporting oversight, or a legitimate adjustment?
As one presenter put it: “AI can flag the smoke, but editors still have to decide if there’s fire.”
Conclusion: Scaling Integrity
Peer review has always depended on trust. At the 10th Peer Review Congress, I saw the scale of fraud facing scholarly publishing, but I also learned how publishers and researchers are meeting the challenge. Paper mills, fake reviewers, image manipulation, and authorship abuse threaten the credibility of the research record. Integrity must be defended at scale. That means scaling tools, scaling collaboration, and scaling the cultural shift toward researcher and institutional accountability.
– By Tony Alves
