Review that keeps up with submissions.
Editors get a line-by-line report in under 30 minutes, so reviewer time goes to judgment, not detection.
Science is outgrowing peer review
Submissions climb every year, and the checking still runs on reviewers’ donated hours. Most of what slips through is honest error, not fraud.
100M+ hours donated by reviewers to peer review in 2020.
33% more journal submissions on ScholarOne in early 2026 than a year earlier.
1 in 2 psychology papers using significance tests misreport a p-value.
3 in 4 flagged paper-mill papers will never be retracted.
SymbyAI checks every submission before a reviewer opens it
SymbyAI flags problems in the figures, statistics, and references. Reviewers start from its findings.
Submission
A manuscript arrives.
Checks
SymbyAI runs them all at once.
- Figures
- Statistics
- Methods
- References
- Arithmetic
- Your guidelines
Report
Line by line, in under 30 minutes.
Triage
Your editors work down the ranked findings.
Decision
Reviewers weigh the evidence. Editors decide, not SymbyAI.
What SymbyAI’s report finds
Every issue carries the evidence it rests on. Decisions stay with the editor.
- Issue · criticalfigure
Panel (ii) of Figure 3b matches the lower-left quadrant of Figure 2d once rotated 90° and contrast-adjusted, down to the same debris at the edge of the field. Figure 3b is the only blot behind the 40% knockdown claimed in §3.2.
- Issue · majorstatistics
A two-sided t-test on 18 degrees of freedom with t = 2.31 gives p = .033. That clears .05, but not the Bonferroni threshold of .017 the methods set for Table 2's three comparisons, so the effect fails the paper's own test.
- Issue · majormethod
§2.3 grows all three biological replicates from one passage of a single culture and splits them before treatment, so the experiment measures one culture three times. The spread in Table 2 is technical variance, not biological.
- Issue · minorreferences
[14] found expression rose after the same treatment in HEK293 cells (its Table 2). The introduction cites it as earlier evidence of a decrease.
Custom rules become enforced checks
Give SymbyAI your journal’s standards and instructions for authors, and it checks every submission against them. A manuscript that breaks a rule gets a finding, with the line where it fails.
Word limits
Keep the abstract under 250 words.
Abstract
MetThe abstract runs to 212 words.
Data availability
Share the data behind every figure in a public repository.
Declarations, Data availability
“The data is available on reasonable request from the corresponding author.”
Issue · majormetadataThe data availability statement offers the data on request and names no repository. Your guidelines ask for the data behind every figure in a public one.
Reporting
Report how the sample size was determined.
§2.4, Sample size
Met§2.4 sizes the sample of 20 adults by a power calculation.
Clinical trials
Reject any clinical trial without a registration number.
Abstract, Design
“Design: a randomized, placebo-controlled trial of the treatment in 20 adults over twelve weeks.”
Issue · criticalmethodThe abstract describes a randomized, placebo-controlled trial. No registry or registration number appears in the abstract, the methods, or the declarations, and your guidelines turn away any trial without one.
The models that read your manuscript
Our proprietary models are trained on open-access science.
SymbyAI’s own servers.
Your manuscript is read on servers SymbyAI operates.
Proprietary models.
Custom-trained on open-access science.
Private data.
SymbyAI never trains on your data.
Talk to us about a pilot
Custom contracts for journals and publishers.
Or write to sales@symbyai.com.