Across 111 AI pre-reviews, methodology issues account for over a quarter of all 2,439 flags

Methodology problems are the single largest source of risk in these 111 manuscripts. Out of 2,439 issues flagged, methodology alone accounts for 656, and it carries far more critical findings than any other category.

Where do the truly high-risk problems cluster?

The severity mix indicates that the most consequential weaknesses are concentrated in how studies are designed and how results are interpreted, not in surface-level writing. Methodology issues have 93 critical flags, far ahead of any other category, which suggests that core design elements such as sampling, controls, measurement, or analytic plans are often fragile enough to threaten the validity of the work.

Conclusions and statistics also carry substantial weight. Together, they combine a large number of major issues with non-trivial critical ones, implying that authors frequently overstate what their data support or apply analyses in ways that are only partially aligned with their research questions or data structure. In contrast, writing and data presentation are dominated by minor and major issues with very few critical flags. That pattern is consistent with manuscripts that are intelligible and structured but whose underlying evidential claims need more careful strengthening.

For authors, the implication is that the highest payoff revisions are rarely cosmetic. Before submission, it is more important to interrogate whether the design actually answers the stated question, whether the analytic strategy is appropriate, and whether the conclusions stay within the bounds of the evidence than to focus first on phrasing or formatting.

What do mid-range rubric scores really signal?

The rubric scores sit in a relatively narrow band: rigor at 3.43, significance at 3.68, and presentation at 3.5 on the 1 to 5 scale. None of these averages are low enough to indicate systematically flawed work, but none are high enough to suggest that manuscripts are consistently ready for acceptance without substantial revision.

A rigor score around the mid-3 range aligns with the concentration of methodology and statistical issues. It points to studies that often have a plausible design and some appropriate methods, yet leave important aspects under-specified or only partially aligned with best practice. Examples might include unclear inclusion criteria, incomplete handling of confounders, or limited robustness checks.

The slightly higher significance score suggests that many projects are asking meaningful questions or addressing relevant problems in their fields, even when the execution is uneven. This is typical of early-stage or resource-constrained work where the idea is strong but the design or analysis needs tightening to fully support the claims.

Presentation in the mid-range indicates that manuscripts are generally readable and organized but could benefit from clearer structure, more precise reporting, and better alignment between text, tables, and figures. The relatively low number of critical writing issues reinforces that most presentation problems are fixable without redesigning the study.

How do citation checks reshape the risk profile?

The citation snapshot shows that verification is non-trivial. Of 463 citations examined, 352 were verified and 39 were flagged as hallucinated. That means a notable share of references either do not exist as cited or are misaligned with the claimed source, with additional citations classified as unverifiable.

This pattern matters because it interacts with the methodology and conclusions categories. When key references are fabricated or cannot be confirmed, it undermines the theoretical framing and the justification for design choices. In cross-disciplinary work, where standards vary by field, unreliable citations can also make it harder for reviewers to assess whether appropriate guidelines or prior findings have been followed.

What should authors change before their next submission?

Taken together, these patterns suggest a clear priority order for pre-submission revision. First, treat methodology and statistical design as the main levers of credibility: revisit inclusion criteria, control strategies, measurement validity, and the match between data and analysis, and ensure conclusions track closely to what the results actually show. Second, use the rubric-style feedback to balance effort across rigor, significance, and presentation, aiming to turn a solid idea into a well-executed study rather than polishing prose alone. Third, audit references systematically, especially any added with AI assistance, confirming that each citation exists, is correctly attributed, and genuinely supports the claim it is attached to.

Authors who approach their drafts with this hierarchy in mind are more likely to resolve the kinds of issues that lead to critical flags in peer review, while still improving clarity and readability in ways that help editors and reviewers engage with the work.