Research integrity, automated

Find the numbers in a paper that can't be real.

Statlint reads the statistics in any research paper and mathematically proves which results are impossible — unifying seven peer-reviewed forensic methods into a single, editor-ready report.

No signup. Nothing stored. Works on PDF, text, or data files — across every field.

statlint report · trial.pdf
Statlint reportrisk score 100 / 100
4 impossibilities found
impossibleGRIMMERTable 2 · Anxiety
SD 0.82 is impossible for mean 2.50, N = 4.
impossibleGRIMTable 2 · Mood
Mean 3.43 can't occur with N = 20 integers.
impossibleSPRITETable 2 · Stress
SD 3.50 exceeds the max a 1–7 scale allows.
reviewCarlislebaseline
Trial arms are implausibly similar (p = 6e−10).
Methods peer-reviewed in Behavior Research Methods· Social Psychological & Personality Science· PeerJ· Anaesthesia
Why it matters

Science has a numbers problem.

Papers report means no dataset could produce, standard deviations that break algebra, and p-values that don't match their own tests. Most are never checked — reviewers and editors don't have the time or the tools. Statlint gives them both.

~10,000
research papers retracted in 2023 — a record, and the curve is still climbing.
>50%
of researchers can't reproduce another scientist's published results.
Seconds
for Statlint to run a full forensic check that takes an expert hours by hand.
How it works

From paper to verdict in three steps.

01

Drop in a paper

Paste a results paragraph, or upload a PDF, JSON, or CSV. Nothing is stored.

02

Statlint runs the battery

Seven forensic checks recompute and cross-examine every number you reported.

03

Read the report

Proven impossibilities first, heuristic flags clearly separated, each in plain English.

The battery

Seven checks, one report.

Five return mathematical proofs. Two return heuristic flags. Statlint never confuses the two.

statcheck recompute proof
Recomputes every p-value from its test statistic and degrees of freedom.
GRIM proof
Tests whether a reported mean is achievable for the stated sample size.
GRIMMER proof
Tests whether a reported standard deviation is possible for that mean and N.
SPRITE bounds proof
Checks the SD against the hard limits the response scale imposes.
Descriptive recompute proof
Verifies reported group means and SDs imply the reported test statistic.
Digit & baseline analysis heuristic
Benford, terminal-digit, and Carlisle baseline checks that flag what merits a look.
No discipline left out

Works across every field.

If a paper reports means, SDs, p-values, or test statistics, Statlint can check it. The methods are field-agnostic by construction.

PsychologyMedicine & clinical trialsNeuroscience EconomicsEcologyPublic health EducationSports scienceNutrition Political scienceSociologyPharmacology
The discipline

Proof, never accusation.

Statlint draws a hard line between a mathematical impossibility — proven, the numbers cannot all be right — and a statistical suspicion, a flag worth a second look that proves nothing on its own. The engine is validated against thousands of real datasets to guarantee it never flags genuine data. Every finding is a question to put to the authors, not a verdict on them.

Impossibility — mathematically proven
Suspicion — heuristic, unproven
Try it now

Check a paper in the next 30 seconds.

Paste a results paragraph or upload a file. Your report opens in a new tab — nothing is saved.

Statlint analyzes your input in memory and returns a report. It does not store documents.