Screen reference lists before they reach review

Submissions increasingly arrive with machine-drafted reference lists — and some of what they cite was never written. The big publishers run integrity pipelines built for the big publishers. If you run a preprint server, an independent journal, or a small press, you are mostly expected to catch invented citations by eye. CiteStamp’s checking engine is free to use today, and we are building the bulk version with pilot partners.

What your editors can do today

• Paste a submission’s bibliography into the health check. Per-reference report in seconds: does it resolve (Crossref, DataCite, OpenAlex — datasets and preprints included), does the title match the DOI it claims, is the work marked retracted, does it carry recorded refutations from named researchers. Free, unlimited, no account.
• Confidentiality is structural, not contractual. The reference list is never uploaded to CiteStamp as a document. Each reference is looked up from the editor’s own browser against the public registries it already lives in — Crossref, DataCite, OpenAlex — the same requests that clicking a DOI would make, under those registries’ own policies. There is no CiteStamp server-side copy of the submission to trust, retain, or leak.
• Embed the engine in your own tooling. The checking engine is one script tag, no build step, no key: https://citestamp.com/check/v1.js. It exposes the functions; a small form and one call to checkBibliography() is the glue your developer writes.

What we are building with pilot partners

A bulk bibliography-screening API for submission pipelines: reference lists in, per-reference reports out, priced for journals that will never buy an enterprise integrity platform. It is not live yet — we describe nothing before it works — and it is being shaped by the editors who pilot it. Pilot access is free while that happens, and pilot partners set the roadmap.

What the check is — and is not

It reports checkable properties of the public scholarly record: resolution, DOI–title agreement, retraction notices, and recorded refutations, each with a source. It is not a plagiarism screen, not image forensics, and never a verdict on the quality of the work — it is one screen among several, for the failure mode that got cheap to mass-produce.