Reference Check resolves each entry in your bibliography against Crossref and reports what disagrees: a retraction, an entry that no longer resolves, an author or year or page range that does not match the published record. Give it the manuscript and it reads the body text too — so it can name a citation with no reference entry, and a reference entry the manuscript never cites. 1 credit for a list, 2 for a manuscript.
Reference errors are rarely invented citations. They are a reference number that shifted when a paragraph moved, an entry left in the list after the sentence citing it was deleted, a year copied from a preprint that was later published, an author initial dropped by a reference manager three papers ago. None of them are visible while reading your own list, because you already know what it was supposed to say.
And a retraction is invisible by definition. A paper cited in good faith in 2023 can be retracted in 2025 without anything in your document changing. Nothing in a manuscript goes stale as quietly, or looks as bad in review.
1 credit for a pasted reference list and 2 for a full manuscript audit, charged once when the check runs. Every style, every re-export and the .docx are free after that. A check that finds nothing at all is free — that outcome is usually a vendor outage, which is the worst possible moment to take a credit — and re-running the same list on the same day returns the original receipt rather than charging again.
Every account gets 5 free credits per calendar month, which reset on the 1st. Most things cost one credit; a readiness check against a named journal and a conversion to your own uploaded template cost 2. Free credits cover all of it. Additional credits cost ₹500 each and never expire. There is no subscription and no card on file.
An entry that cannot be resolved is flagged and left exactly as you wrote it. Nothing is invented, nothing is corrected on a guess, nothing is silently dropped. A reference missing from Crossref is often perfectly real — a book, a thesis, a preprint, a conference paper nobody deposited — and the report says that rather than pretending to know better.
Crossref itself is not perfect. Publishers deposit their own metadata and some deposit it badly: a missing page range, an initial where a full name belongs, a journal title abbreviated a different way. Disagreements are therefore reported field by field with both values shown, so you are judging a difference rather than accepting a correction.