中文版｜⚙️Demo｜📢Notice

Overview

RefCheck.ai is a lightweight web tool for validating BibTeX references and spotting suspicious or inconsistent citations. Users can upload a .bib file (or paste BibTeX content), watch a live verification pipeline in the browser, and receive a structured report that highlights mismatched fields (e.g., title/author/year) along with a suggested “best match” from trusted scholarly indexes.

Try it on RefCheck-ai.onrender!

How it works

RefCheck.ai treats each BibTeX entry as a claim and verifies it via a multi-source, evidence-driven matching pipeline:

1. Parse & normalize
   BibTeX is parsed and normalized (basic LaTeX cleanup + canonicalized title/author/year) to reduce formatting noise.

2. Multi-source retrieval
   The system queries scholarly metadata providers (e.g., Semantic Scholar, Crossref, OpenAlex) to retrieve candidate works.

3. Evidence-based matching
   Candidates are scored primarily by title similarity, with additional signals from author and year (±1) consistency. The best-matching candidate becomes the reference “ground truth” suggestion.

4. Three-level decision

   • verified: title + author + year (±1) all agree with the best match
   • uncertain: the work likely exists but has one or more mismatched fields
   • suspicious: multiple strong inconsistencies or no plausible candidates despite reachable sources

User API Key model

RefCheck.ai supports bring-your-own Semantic Scholar API key:

• Without login or an API key, the tool runs in a degraded mode.
• After login, users can store exactly one API key in their profile (encrypted at rest) to enable full Semantic Scholar retrieval.