What is Preprints.ai?

Preprints.ai is an AI-powered quality control system for academic preprints. It uses multiple AI models (Claude, GPT-4o, Gemini) to independently review preprints from bioRxiv, medRxiv, and arXiv, producing a consensus grade (A–E) based on evidence strength and research significance. All assessments are machine-generated indicators that require human expert review.

How does Preprints.ai grade papers?

Papers receive a two-dimensional grade: a letter (A–E) for evidence strength and a number (1–5) for significance. Grade A means exceptional evidence, B is compelling, C is solid, D is incomplete, and E is inadequate. The significance scale runs from 5 (Landmark) to 1 (Useful). Nine AI agents review each paper independently, then a consensus algorithm aggregates their scores.

What preprint servers does Preprints.ai support?

Preprints.ai supports papers from bioRxiv, medRxiv, and arXiv. You can submit papers using a DOI, arXiv ID, or the full bioRxiv/medRxiv URL. The system automatically fetches metadata and full text when available.

Is Preprints.ai free to use?

Yes, Preprints.ai is free to use for individual assessments. There is also a free API tier for programmatic access. The platform is part of the Infinite Researchers programme, an open science initiative.

Does Preprints.ai replace peer review?

No. Preprints.ai provides machine-generated quality indicators as a first-pass filter. All assessments require human expert review and should not be treated as peer review. The platform assists but does not replace traditional peer review processes.

Multi-agent AI peer review

Quality signal
for preprints.

Nine specialist agents review every preprint independently, then deliberate to produce a consensus grade on evidence strength and significance.

— papers assessed across — indexed preprints from bioRxiv and medRxiv.

Machine-generated indicators. Assists but does not replace expert peer review.

Sample report

10.1101/xxxx.xx.xx

Compelling evidence

Important finding

Reviewers

Agreement

87%

Runtime

4m 12s

Cost

$0.04

Rigorous methodology, replicated across two cohorts. Minor limitations in platform diversity but core claims are well-supported.

Live pipeline

Auto-refreshes every 30 seconds

Live

Indexed

—

Assessed

—

In queue

—

Full text

—

Integrity

letter · A–E

A exceptional · B compelling · C solid · D incomplete · E inadequate

Novelty

number · 1–5

5 landmark · 4 fundamental · 3 important · 2 valuable · 1 useful

Recent assessments

View all →

Grade	Paper	Source	Assessed
Loading…

How it works

Four layers. Nine agents. One consensus grade.

Layer 1

Deterministic checks

Paper-mill detection, statcheck p-value verification, GRIM tests, data availability, retraction cross-referencing — before any LLM runs.

Layer 2

Nine-agent review

Four integrity agents (methodologist, statistician, ethics, validity) plus five domain agents review independently.

Layer 3

Deterministic grading

Evidence_strength and significance labels from each agent map to an A–E grade via a rule-based lookup calibrated to eLife reviews.

Layer 4

Opus arbitration

Borderline or low-agreement cases are arbitrated by Claude Opus with the full agent panel as context.