Know exactly how much
to trust a study.
PaperGist reads the whole paper — figures, tables, statistics — grades it on a transparent rubric, and re-reads its own write-up to catch anything that overstates the evidence.
Does intermittent fasting actually work?
Evidence grade
A trust score you can actually inspect.
No black-box number. Every score breaks down into named criteria — design, sample size, controls, bias, statistical rigor, conflicts, reproducibility — each rated and explained, and adapted to the field.
- Strong / Adequate / Weak ratings per criterion
- Effect sizes, CIs & p-values for experts
- Honest about small samples and missing controls
Evidence grade
Reported effect
Faithfulness check
Overstated “Coffee prevents cancer.”
Calibrated: linked to lower risk in this sample — not proven to prevent it.
The faithfulness pass
A second, skeptical read catches what the first one overstated — and shows you the calibrated version.
Peer-reviewed research found AI summaries are ~5× more likely than humans to over-generalize. PaperGist is built to do the opposite.
One analysis, two readers
Plain-English, or expert-dense.
People who fasted 16 hours a day lost a little more weight than those who didn’t — but the difference was small, and the study only ran 12 weeks, so we can’t say it lasts.
16:8 TRE vs. control: −1.6 kg (95% CI −2.9 to −0.3), p=0.02, n=212, 12-week RCT.
Δweight −1.6 kg · CI[−2.9,−0.3] · p=0.02 · n=212
Field lenses
Judged by the right standards.
A clinical trial isn’t judged like a machine-learning benchmark. Pick a field — or let it auto-detect — and the rubric adapts.