Detect Suspicious Data Patterns
in Research Papers
No raw data? No problem. Check reported statistics against sample sizes with the GRIM and SPRITE tests, recompute reported p-values with statcheck-lite, or paste any dataset for a 30-second statistical anomaly screening — Benford's Law, terminal digit analysis, variance checks. Research-grade methods, no sign-up required.
📄 Drop a PDF here or click to upload
Auto-extracts table data · never leaves your browser
🔒 Data is processed locally in your browser — never uploaded, never stored.
Why statistical screening?
Statistical forensics uses well-established methods to flag data that may be too clean, too random, or too patterned. These are the same techniques used by research integrity investigators.
GRIM Test
Mean–granularity consistency
Only the sample size n and mean M as printed in the paper: is that mean mathematically possible? (Brown & Heathers, 2017). No raw data needed.
Read the method →SPRITE Test
Mean–SD consistency
GRIM's big sibling: is the reported n + M + SD combination possible given the measurement granularity and scale bounds? Catches SDs that are too small or too large (Peeters et al., 2016).
Read the method →Statcheck (p-value Recomputation)
Exact p-value recalculation
Scans reported t/F/χ²/r/z statistics, recomputes the exact p-value, and flags reporting inconsistencies and significance decision errors (Nuijten et al., 2016).
Read the method →Benford's Law
Leading-digit distribution
Natural data's leading digits follow a specific distribution (1 ≈ 30%, 9 ≈ 4.6%). Fabricated 'random' data often deviates.
Read the method →Terminal Digit Analysis
Last-digit uniformity
Real measurements have uniformly distributed last digits. Hand-made data often favors 0 and 5 (Mosimann, 1995).
Read the method →Variance Anomaly
Too-clean data screen
Real data varies naturally. Fabricated data is often 'too clean' — tiny variance, perfect fits, p-values clustered at 0.04–0.05.
Digit Pair Frequency
Adjacent-repeat rate
Non-leading digits should be roughly uniform. Deliberately avoiding adjacent repeats is itself a fabrication fingerprint.
Citation Verification
Crossref reference check
Fabricated references are the paper mill's most common flaw. Paste a reference list and verify every entry against the public Crossref database.
Read the method →Pricing
🎉 Every feature is free during launch — no registration required. Pro features (batch screening, report library) will be announced here before they ship — start using it now.
Free
- ✓All detection methods
- ✓Multi-group GRIM & SPRITE
- ✓Statcheck p-value recomputation
- ✓PDF report export
- ✓Local processing, zero upload
Pro
- ✓Batch screening
- ✓Report library & history
- ✓Higher data limits
- ✓Priority support
Lifetime
- ✓All Pro benefits
- ✓One-time payment
- ✓Great for lab budgets