Metacheck is an open-source R package that checks research outputs for best practices, using text search, R code and, optionally, queries to large language models. Its 26 modules cover areas such as preregistrations, ethics approval, funding and conflict-of-interest statements, open practices, effect sizes, p-values, power analysis and references, and it cross-references citations against Retraction Watch, PubPeer and the FLoRA replication database. Lisa DeBruine and Daniël Lakens were its first developers, and it is part of the Research Transparency Check project funded by the Dutch Research Council (NWO). The developers renamed it from Papercheck to metacheck, and launched the beta on 5 June 2026 at the AI in Peer Review conference at TU Eindhoven.
Notes & quotes
- The developers say language models are used only if the user chooses, are declared, and are limited to classifying existing text, with non-LLM methods such as regular expressions preferred where possible.
- The developers say the checks assist but do not replace human expertise, and that metacheck will never support automated quality assessments, rankings or scoring.
- Planned work includes a one-click interface that does not require users to install R.
Publisher / company: Scienceverse
Date of case study: 2026
Status: Beta
Built by: Research, non-profit or individual
Sources: