Know where the evidence stands in your field, and what just moved.

Priors keeps a public, living evidence review of your field. Each week a four-reviewer AI panel scrutinises every new paper against the primary literature and grades the certainty, so you know what's moving the evidence and what you can skip.

GLP-1 receptor agonistspriors.science/glp-1-receptor-agonists
Example
SignalClaimStandingEvidence certaintyCorpusLast moved
▲ strengthened +3 Reduce alcohol consumption in alcohol use disorderAddiction and reward Uncertain56%20 4 Jul
Stable Reduce MACE in type 2 diabetes with high CV riskCardiovascular protection Established85%20 21 May
▼ weakened −4 Semaglutide increases risk of NAIONSafety signals Uncertain50%20 3 Jul
Stable Semaglutide 2.4 mg achieves about 15% weight lossMetabolic and organ protection Established90%20 2 Apr
▼ weakened −5 Improve or stabilise cognition in established Alzheimer'sNeurodegeneration and cognition Doubtful35%20 3 Jul
39 claims tracked · 633 papers reviewed · re-scored weekly · illustrative weekly movementSee all 39 claims →
How it works

A public, living review of your field, updated weekly.

Every new paper is scrutinised against the live literature by a four-reviewer panel. Continuously updated, fully cited, and built to withstand scrutiny.

01

Every field is a set of claims

Priors breaks your field into falsifiable claims, ordered from most established to most contested.

02

The evidence is assembled

Every relevant trial and paper from the live literature, gathered behind each claim.

03

A four-reviewer panel scores it

Each claim debated and graded for certainty, framed on PICO. The score is the current prior.

04

It stays current

Re-scored weekly, with a feed of exactly which priors moved and why.

What to read first

Priors points you at the papers that are actually moving your field.

For every prior that moved, you get why it moved and the papers driving it. The ones that change nothing, you can skip with confidence.

Illustrative example
▲ strengthened +3 Uncertain · 56%

Reduce alcohol consumption or improve alcohol use disorder outcomes.

Why it movedA 500-patient randomised trial reported significantly fewer heavy-drinking days, the first adequately powered RCT in this indication, and the reason the prior rose three points.

Read these first
▲ movingSemaglutide for alcohol use disorder: a randomised controlled trial · NEJM · 2026
▲ movingGLP-1 agonism and reward signalling: a mechanistic cohort · Nature Medicine · 2026

17 other papers mentioned this claim and changed nothing. The panel read them so you don't have to.

The sceptic's challenge

What the paper didn't show.

Every panel includes a dedicated sceptic whose only job is to attack the finding. For the claims that matter, you get the single strongest objection, and the evidence that cuts against it.

Illustrative example
Established · 90% Semaglutide 2.4 mg weekly · weight loss in obesity

Achieves about 15% body weight loss in obesity.

The sceptic's challengeThe pivotal trial was open-label with a partly subjective co-primary endpoint, which invites expectation bias, and its 68-week window can't speak to weight regain after discontinuation, the question most likely to matter in practice. Blinded maintenance data would resolve it.

Key contested evidence
▼ contestsTwo-thirds of lost weight regained within a year of stopping semaglutide · STEP-1 extension · Diabetes Obes Metab · 2022
▼ contestsOpen-label trials overstate subjective weight outcomes: a meta-analysis · BMJ · 2025

The panel weighs this against the claim before scoring, which is why "Established" here means the effect itself, not its durability after stopping.

Traceable evidence

See how the literature supports or challenges a claim.

Every paper in a claim's corpus is classified supporting, neutral, or challenging, and opens to the trial behind it. Nothing is asserted without a source you can read.

Illustrative example
Uncertain · 49%Semaglutide increases the risk of NAION · ▼ weakened 15 papers in corpus
YearStudyEvidenceRole in corpus
2024 Risk of non-arteritic anterior ischaemic optic neuropathy in patients prescribed semaglutideJAMA Ophthalmology · DOI Supporting Retrospective single-centre cohort: hazard ratio ~4 for NAION in type 2 diabetes. The study that first raised the signal.
2025 Semaglutide and optic neuropathy: a multinational cohort of 1.2 million patientsOphthalmology · DOI Challenging No significant elevation after adjustment for confounders. Largest real-world study to date; the main reason the claim weakened.
2025 GLP-1 receptor agonists and ischaemic optic neuropathy: a disproportionality analysisDrug Safety · DOI Neutral FAERS pharmacovigilance signal; hypothesis-generating only, cannot establish incidence or causation.
2025 NAION events in the pooled SUSTAIN and PIONEER randomised programmeDiabetes Care · DOI Challenging Pooled trial safety data: NAION events rare and balanced against placebo. Randomised evidence weighs against the signal.
Method

Every prior shows its working.

A panel of four reviewers appraises the primary literature behind each claim, using GRADE for certainty and PICO to keep each claim precise. The score you read is their synthesis, with the corpus, the reasoning, and the uncertainty shown behind every judgment.

Priors is a sceptical first reader, not a replacement for peer review, and not medical advice. When the evidence is thin, Priors says so. That is the point.

The panelFour reviewers, synthesised
CertaintyGRADE, 0–100%
Claim structurePICO
Update cadenceWeekly
Behind each claimCorpus + what moved it
FAQ

Questions, answered.

Why does it exist?

Priors was built by a small group of researchers frustrated with the same thing you are: too many papers, no way to filter the noise, and the ones that matter slipping past. AI has mostly made that worse. We wanted to make it help instead, to do the reading that keeps you current, without burying you.

Each week around 35,000 papers are published in medicine alone, and the vast majority don't change the field. Priors is a way for researchers to see how each new paper affects the standing evidence in their field, so we can move faster towards the highest-quality science, and accelerate discovery.

Who is Priors for?

Working researchers, clinicians, and anyone who needs to know where the evidence stands in a field without reading every paper that comes out. You pick your topics; Priors keeps the review current.

How is this different from a search engine or standard AI?

Finding papers was never the hard part. Search already does that, and a chatbot will happily summarise one for you. The problem is knowing which papers actually change what the field knows, and by how much. Priors doesn't just retrieve or summarise; it tracks the key claims in your field, weighs each new paper against the existing evidence, and tells you whether the standing moved. It's a measure of where the evidence stands, not a list of links.

What data does it draw on?

The primary literature: randomised trials, cohorts, and meta-analyses from the indexed biomedical record. Every claim is backed by a traceable corpus of primary papers, each classified as supporting, neutral, or challenging, with the source shown behind every judgment.

Are the scores validated?

Yes. The four-reviewer review protocol was validated and tested against more than 400 Cochrane systematic reviews, the gold standard for evidence synthesis. Against them it delivered roughly twice the GRADE accuracy of a standard LLM, and 85% less overconfidence.

What about AI fabrication?

Fabricated papers and invented DOIs are a well-known failure of standard LLMs. Priors has specific guardrails built to avoid them: every model review is anchored to the real literature, so each score is built from papers that actually exist, traceable back to their source.

What is Priors not?

Priors applies a systematic, standardised protocol to the key claims in a field, but it is not a replacement for the definitive systematic reviews produced by bodies like Cochrane or the IPCC. Our protocol is built for scale, not depth. Think of it as a reliable first pass that keeps you current and gives you a foundation to build on, not the final word, and not a substitute for the comprehensive review you'd write yourself. We don't cover every claim in a field, but subscribers can request coverage of ones we're missing.

Pick your field

Update your priors.