# Reviewer3 > Reviewer3 is grounded, domain-specific AI peer review for researchers and journals. > It catches critical issues in research papers — methodology flaws, reproducibility > gaps, and fabricated/hallucinated citations — before publication. Used by thousands > of researchers; encrypted storage, never used for AI training, SOC 2 compliant. Reviewer3 runs multiple specialized AI reviewer agents over a manuscript and synthesizes their feedback into an editor-style meta-review. It supports two modes: author (researchers reviewing their own work before submission) and journal (editorial screening of incoming submissions). ## Product - [Home](https://reviewer3.com/home): Overview, features, and how the AI review works. - [For researchers](https://reviewer3.com/upload?mode=author): Upload a paper for author-mode review. - [For journals](https://reviewer3.com/upload?mode=journal): Editorial screening for journals. - [API access](https://reviewer3.com/api-access): Request programmatic access to the review pipeline. - [Pricing](https://reviewer3.com/pricing): Plans and pricing. ## Evidence & benchmarks - [Benchmarks overview](https://reviewer3.com/evidence/benchmarks): Benchmarks for peer-review quality and reference verification. - [Reference checker benchmark](https://reviewer3.com/evidence/benchmarks/references): Catches 97.7% of fabricated references at 98.6% precision, tested on 476 real and fake citations. - [Review Arena](https://reviewer3.com/evidence/arena): Blinded "which review is human?" comparison game. - [Case studies](https://reviewer3.com/evidence/case-studies): Real examples of AI catching fraud and methodological flaws. ## Trust & security - [Security](https://reviewer3.com/security): Encryption, SOC 2 compliance, and data handling practices. - [Privacy policy](https://reviewer3.com/privacy): How data is collected, used, and protected. - [Terms of service](https://reviewer3.com/terms): Usage agreement. ## Optional - [Sitemap](https://reviewer3.com/sitemap.xml): Full list of indexed URLs, including individual evidence posts.