Covidence offers automatic extraction in Extraction 1 (DE1) where we can source reliable suggestions, provided a DOI is available. Extraction suggestions are enabled by default and draw from the following data sources:

  • Study metadata from external repositories (for both open and closed access studies)
  • Large Language Model (LLM) extraction from full-text articles (for both open and closed access studies)

Covidence ensures that PDFs and data derived from them remain confidential, with no access granted to individuals outside the respective review. The platform utilises Large Language Models (LLMs) that do not train on, store, or distribute any user-provided full-text PDF data. The LLMs use this data solely to generate outputs that assist the user’s review, without retaining or repurposing the input for any future use.