Rigor Scan: Reporting Check for Research Papers

1
Read on your device
Your browser reads the text and layout of each page. The PDF itself never leaves your computer.
2
Check with software
Statistics are recalculated, references checked for retractions, cell lines matched against a register and data links opened.
3
Check each item
About 60 reporting items are checked against the text, and each result is tied to the sentence that supports it.
4
Score
Items with strong evidence count as reported and items with weak evidence as missing. The score averages the items that apply to this kind of study.

Your PDF stays with you

The file is read in your browser and never uploaded.
Only the extracted text is sent for checking, and only identifiers and reference citations for the database lookups. K-Dense doesn't save any of it.
Results live in this tab. Close it and they're gone.

Checklist v0.3 · 23 September 2026

How scoring works

Everything that decides a result is published here: every question, every threshold and every software rule. The launch post explains how a scan works and why it's built this way.

Scoring rules

An AI model reads the paper's text and answers each check below with a probability that the paper reports that item. The questions are exactly as written in the table.
A check counts as reported at 70% or above, missing below 35%, and unclear in between.
Warning-sign checks are scored the other way round: finding the problem lowers the score.
The same pass decides which study types the paper contains. Checks for a study type below 60% are skipped and left out of every score.
A domain score is the average evidence across its checks that apply. The completeness index is the average across all checks that apply, so larger domains weigh more.
Software checks are reported on their own and never change the index.
The checklist is versioned. Compare scores only within the same version.

evidence = p          for a reporting check
         = 1 − p      for a warning-sign check
status   = reported if evidence ≥ 0.70, missing if < 0.35, else unclear
index    = 100 × mean(evidence over the checks that apply)

Study types

Asked of every paper. A type counts as present at 60% or above.

New experimental data: Does the paper report its own experiments or measurements on people, animals, cells, tissues, or other biological samples?
Assigned treatments: Does the paper assign subjects or samples to treatments, conditions, or experimental groups chosen by the investigators?
Animal experiments: Does the paper report experiments performed on live animals?
Human participants: Does the paper report data collected from human participants, patients, or patient records?
Clinical trial: Is the paper a clinical trial that prospectively assigns human participants to interventions?
Cell lines: Does the paper report experiments on cultured cell lines?
Antibodies: Does the paper use antibodies, for example for staining, blotting, flow cytometry or immunoprecipitation?
Prediction model: Does the paper build or validate a model whose purpose is to predict outcomes for new cases?
Omics data: Does the paper generate high-throughput data such as sequencing, transcriptomics, or proteomics?

Overall judgments

Graded reads shown next to the score. They don't count towards it.

Methods detail: How much procedural detail does the methods section give for another lab to repeat the experiments? Levels: Minimal → Partial → Detailed → Protocol-level.
Claim strength: How strongly does the abstract's conclusion state its main claim? Levels: Tentative → Measured → Confident → Sweeping.

74 checks

Check	Question asked	Applies to	Guideline	Type
Study design · 15
Hypothesis stated	Does the paper state a specific hypothesis or research question that the experiments test?	All papers		reporting
Primary outcome defined	Does the paper name a primary outcome measure for its main experiment?	All papers		reporting
Control group	Does the paper include a control or comparison group for its main experiment?	New experimental data or Human participants		reporting
Group sizes reported	Does the paper report the number of animals, participants, or samples in each experimental group?	New experimental data or Human participants		reporting
Sample size justified	Does the paper explain how the sample size was chosen, for example with a power calculation?	New experimental data or Human participants	ARRIVE 2.0 · item 2	reporting
Randomization	Does the paper state that subjects or samples were randomly allocated to experimental groups?	Assigned treatments	ARRIVE 2.0 · item 4	reporting
Randomization method	Does the paper describe the method used to generate the random allocation sequence?	Assigned treatments		reporting
Blinded allocation	Does the paper state that the people who allocated subjects or gave the treatments did not know which group each subject was in?	Assigned treatments	ARRIVE 2.0 · item 5	reporting
Blinded assessment	Does the paper state that outcomes were assessed, or data analysed, by someone who did not know which group each subject was in?	Assigned treatments	ARRIVE 2.0 · item 5	reporting
Inclusion criteria	Does the paper state criteria for including or excluding subjects, samples, or data points?	New experimental data or Human participants	ARRIVE 2.0 · item 3	reporting
Exclusions reported	Does the paper report whether any subjects, samples, or data points were excluded from analysis?	New experimental data or Human participants	ARRIVE 2.0 · item 3	reporting
Independent replication	Does the paper state that key experiments were repeated independently or replicated in more than one model?	New experimental data		reporting
Replicate type defined	Does the paper say whether replicates are biological or technical replicates?	Cell line work or Animal studies		reporting
Trial registration	Does the paper give a clinical trial registration number, such as an NCT number?	Clinical trials		reporting
Allocation concealment	Does the paper describe how treatment allocation was concealed until assignment?	Clinical trials		reporting
Statistics · 17
Tests named	Does the paper name the statistical tests or models used to analyse the data?	All papers	ARRIVE 2.0 · item 7	reporting
Experimental unit	Does the paper state what the experimental unit of analysis is, such as a single animal, cage, or culture?	New experimental data	ARRIVE 2.0 · item 1	reporting
Assumptions checked	Does the paper describe checking the assumptions of its statistical tests, such as normality?	All papers		reporting
Multiplicity handled	Does the paper describe correcting for multiple comparisons or a pooled approach that addresses them?	All papers		reporting
Uncertainty reported	Does the paper report confidence intervals, credible intervals, or standard errors for its main estimates?	All papers		reporting
Effect sizes	Does the paper report the size of the effect or difference between groups, not only whether it was significant?	All papers		reporting
Error bars defined	Does the paper define what its error bars or dispersion measures represent, such as SD or SEM?	All papers		reporting
Analysis software	Does the paper name the software or packages used for statistical analysis?	All papers		reporting
p-value only reporting	Does the paper report results only as significant or not significant, without estimates of effect size?	All papers		warning sign
Confounders adjusted	Does the paper adjust its analysis for potential confounding variables?	Human participants		reporting
Missing data handled	Does the paper describe how missing data were handled?	Human participants		reporting
Held-out evaluation	Does the paper evaluate its model on data that were not used for training?	Prediction models		reporting
Leakage prevented	Does the paper describe steps taken to prevent information leaking from test data into training?	Prediction models		reporting
Baseline compared	Does the paper compare its model against a simpler baseline method?	Prediction models		reporting
External validation	Does the paper validate its model on an independent external dataset?	Prediction models		reporting
FDR control	Does the paper control the false discovery rate for its high-throughput analyses?	Omics data		reporting
Batch effects	Does the paper describe how batch effects were assessed or corrected?	Omics data		reporting
Materials · 12
Species & strain	Does the paper report the species and strain of the animals used?	Animal studies	ARRIVE 2.0 · item 8	reporting
Sex reported	Does the paper report the sex of the animals used?	Animal studies	ARRIVE 2.0 · item 8	reporting
Age or weight	Does the paper report the age or weight of the animals used?	Animal studies	ARRIVE 2.0 · item 8	reporting
Animal source	Does the paper name the supplier or source of the animals?	Animal studies	ARRIVE 2.0 · item 8	reporting
Housing & husbandry	Does the paper describe how the animals were housed, such as cage conditions, light cycle, or food and water access?	Animal studies	ARRIVE 2.0 · item 15	reporting
Cell line source	Does the paper name the source of each cell line used?	Cell line work		reporting
Cell authentication	Does the paper state that cell lines were authenticated, for example by STR profiling?	Cell line work		reporting
Mycoplasma testing	Does the paper state that cell lines were tested for mycoplasma contamination?	Cell line work		reporting
Antibody identifiers	Does the paper give clone names, catalogue numbers, or RRIDs for the antibodies used?	Antibody use		reporting
Reagent sources	Does the paper name the manufacturer or supplier of key reagents?	Cell line work or Animal studies		reporting
Doses & routes	Does the paper report the doses and routes of administration for treatments given?	Assigned treatments		reporting
Instruments named	Does the paper name the instruments or equipment models used for measurements?	New experimental data		reporting
Data & code · 10
Data statement	Does the paper include a data availability statement?	All papers		reporting
Data in a repository	Does the paper deposit its data in a public repository and give an accession number, DOI, or link?	All papers		reporting
Data only on request	Does the paper say its data are available only on request from the authors?	All papers		warning sign
Code available	Does the paper state that its analysis code is publicly available?	All papers		reporting
Code link	Does the paper give a link or DOI for its analysis code?	All papers		reporting
Protocol shared	Does the paper link to a detailed protocol, such as on protocols.io or in the supplement?	New experimental data		reporting
Preregistered	Does the paper state that the study or its analysis plan was preregistered?	All papers		reporting
Supporting files	Does the paper provide supplementary data, tables, or figures?	All papers		reporting
Raw data	Does the paper state that raw or individual-level data are shared, not only summary statistics?	All papers		reporting
Omics accession	Does the paper give an accession number for its sequencing or omics data, such as a GEO or SRA ID?	Omics data		reporting
Ethics & disclosure · 11
Ethics approval	Does the paper state that an ethics committee or institutional board approved the study?	Human participants or Animal studies	ARRIVE 2.0 · item 13	reporting
Approval number	Does the paper give the protocol or approval number from the ethics committee?	Human participants or Animal studies		reporting
Informed consent	Does the paper state that informed consent was obtained from participants?	Human participants		reporting
Humane endpoints	Does the paper define humane endpoints for the animals?	Animal studies	ARRIVE 2.0 · item 16	reporting
Anaesthesia & analgesia	Does the paper describe the anaesthesia or pain relief given to the animals?	Animal studies		reporting
Euthanasia method	Does the paper describe how the animals were euthanised?	Animal studies		reporting
Funding statement	Does the paper state who funded the work?	All papers		reporting
Funder role	Does the paper state what role, if any, the funders had in the study?	All papers		reporting
Competing interests	Does the paper include a competing interests or conflict of interest statement?	All papers		reporting
Author contributions	Does the paper list what each author contributed?	All papers		reporting
Reporting guideline	Does the paper state that it follows a reporting guideline such as ARRIVE, CONSORT, or STROBE?	New experimental data or Human participants		reporting
Claims · 9
Limitations discussed	Does the paper discuss limitations of its own study?	All papers		reporting
Compared to prior work	Does the paper compare its findings with results from previous studies?	All papers		reporting
Alternative explanations	Does the paper consider alternative explanations for its findings?	All papers		reporting
Null results reported	Does the paper report experiments that found no effect or no difference?	All papers		reporting
Translation caveat	Does the paper caution that results in its model system may not carry over to humans?	Animal studies		reporting
Causal overreach	Does the paper describe a correlation or association as if it proves causation?	All papers		warning sign
Overgeneralization	Does the paper generalize its conclusions to populations or conditions it did not study?	All papers		warning sign
Next steps stated	Does the paper state what further studies are needed to confirm or extend its findings?	All papers		reporting
Adverse events	Does the paper report adverse events or harms in each group?	Clinical trials		reporting

Software checks

Reported statistics

t, F, χ², r and z results that come with a p-value are recomputed. The rounding of the reported statistic is allowed for. A mismatch is flagged, and marked serious when the recomputed p falls on the other side of 0.05.

Method follows statcheck (Nuijten et al., 2016, Behavior Research Methods).

Ratios and intervals

For hazard, odds and risk ratios with a 95% interval, the estimate must lie inside the interval and near its centre on the log scale, and any p-value must agree with the one the interval implies.

Altman & Bland (2011), BMJ 343:d2090.

Percentages

Every "x of n (y%)" is recomputed and must match to the precision reported.

Plain arithmetic.

GRIM

A mean of whole-number data (counts, scores, ratings) must be reachable with the reported n. Only applied when the sentence says the data are whole numbers and n is small enough for the test to say anything.

Brown & Heathers (2017), Social Psychological and Personality Science.

Retractions and notices

Each reference DOI is looked up in Crossref, which carries the Retraction Watch database, and each PubMed ID in PubMed. A reference without either is matched in Crossref by its citation, and the match only counts when the title agrees. Retractions and expressions of concern are flagged; corrections are listed. The paper's own DOI is checked too.

Crossref REST API, Retraction Watch, NCBI E-utilities.

Cell lines

Cell line names in the text are compared with the lines Cellosaurus lists as contaminated, misidentified or misclassified. Named lines are also looked up for their Cellosaurus accession and STR profile.

Cellosaurus (SIB), CC BY 4.0.

Data, code and registries

Repository DOIs must resolve and, where the repository says, hold files. GitHub repositories must exist. GEO accessions must be public. Trial registration numbers must exist in their registry. RRIDs must resolve.

doi.org, Zenodo, GitHub, NCBI, ClinicalTrials.gov, SciCrunch.

What leaves your device

Only identifiers (DOIs, PubMed IDs, accession and registration numbers, cell line names) and the citations of references that have no identifier. Never the body of the paper.

Limits

Text only. Figures, and tables drawn as images, aren't read.
Supplementary files aren't included unless they're part of the PDF.
Scanned PDFs without a text layer can't be read.
English works best.
A check reads what the paper says, not whether it's true. A paper can report every item and still be wrong.
The AI can misread. Every result links to its evidence so you can check it.

Rigor Scan

Study design

Statistics

Materials

Data & code

Ethics & disclosure

Claims

Checked by software

Close the gaps with K-Dense Web

Overall judgments

Read on your device

Check with software

Check each item

Score

Your PDF stays with you

How scoring works

Scoring rules

Study types

Overall judgments

Software checks

Reported statistics

Ratios and intervals

Percentages

GRIM

Retractions and notices

Cell lines

Data, code and registries

What leaves your device

Limits