PDF Reference Checker — Click Any Citation While You Read

EvalCite's PDF reader renders a paper in your browser and makes every in-text citation live: click [12] or (Smith, 2020) and instantly see whether the cited work exists, its correct metadata, and its abstract. No other free tool verifies citations in context, while you read. Free, no signup, and your PDF never leaves your browser unprocessed — it's discarded right after the references are extracted.

Open the PDF reader
Drop in a paper and click any citation. Free. No signup.
Launch the reader

How it works

1. Open a PDF

The file is rendered locally in your browser with PDF.js. On the server, only the references section is extracted and parsed — numbered lists and author-year styles alike.

2. Click a citation

Every citation marker is detected and linked to its reference-list entry: numbered styles ([3], [3,7], [2–5], [4]–[8]) and author-year styles ((Smith, 2020), Smith et al. (2020), [Smith 2020]). Clicking pins a card in the side pane.

3. Read the verdict

The card shows the verification status against 8 databases, the matched record's metadata, its abstract when available, and a Copy-BibTeX button for the corrected entry. Bulk markers like [2–5] resolve to one compact card per reference.

Who uses the reader

Frequently asked questions

Is the PDF uploaded to your servers?

The file is sent for reference extraction only, processed in memory, and deleted immediately after. Page rendering happens entirely in your browser. Nothing is stored or used for training.

Which citation styles does the reader recognize?

Numbered brackets (single, lists, and ranges, including the IEEE [4]–[8] two-bracket form) and author-year (parenthetical, narrative, and bracketed). Hyperlinked citations split across PDF text spans are handled.

Can I verify the whole reference list at once instead?

Yes — upload the same PDF on the main checker and every reference is verified in one batch, with a corrected .bib export.

Does it work on scanned PDFs?

No — the reader needs a text layer. OCR the scan first (any OCR tool) and it will work.