VeriCite

A CLI tool for verifying academic citations against the Crossref database.

What it does

Takes a PDF, text file, or a pasted citation string, extracts the references, and checks each one against the Crossref API. Useful for spotting hallucinated or fabricated citations in AI-generated text.

Installation

pip install vericite

To process scanned PDFs (image-based), also install PaddleOCR:

pip install paddlepaddle paddleocr

Usage

vericite

Then enter a PDF path, text file path, or paste a citation directly:

➜ ./paper.pdf
➜ ./refs.txt
➜ Smith, J. et al. (2025). Quantum Entanglement in LLMs. Journal of Fake Science.

Results are shown inline — ✅ verified via Crossref, ❌ not found or low confidence.

How it works

1. For PDFs: locates the reference section (via TOC or heuristic scan), extracts text with PyMuPDF, falls back to PaddleOCR for scanned pages.
2. Each citation is queried against api.crossref.org/works using bibliographic free-text search.
3. A match with a Crossref relevance score above 35 is considered verified.

License

MIT