Your manuscript, and what happens to it

Placing citations sends your manuscript to the model that does the placing, and the result comes back to your browser. It is not written to a database, not kept in a file, and not retained after the request finishes. There is no table in this system that holds a manuscript.

The same is true of the document you get back. If you close the tab without exporting, it is gone, and we cannot recover it for you.

The Review mode reads your document on your own machine. Finding the citations, matching them to your reference list and reporting what does not line up all happen in the browser — no model, no upload, nothing sent anywhere.

The one exception is that we ask CrossRef and PubMed whether each reference resolves to a real paper. That sends the reference — the authors, title and DOI of a published work — and nothing from your own text.

Anthropic runs the model. Your manuscript and your reference list are sent there to be processed. Anthropic does not use data submitted through its API to train its models. Their policy is at anthropic.com/legal/privacy.

CrossRef and PubMed resolve each reference to its real authors, year, journal and DOI. Only the reference is sent — never your manuscript. This is deliberate and it is the core of how the tool works: bibliographic details come from those databases, never from the model, so nothing is invented.

Vercel and Supabase host the application and its database.

• Your email address, if you gave one — to count free runs, to answer you, and to tell you when access opens.
• Your reference library, if you save references to it. These are bibliographic details of published papers.
• Resolved reference metadata, cached so the same paper is not looked up twice. This is public bibliographic data about published work, and it is shared across everyone who uses the tool.
• Notes about each run — how long a manuscript was, how many references, how long it took, whether it worked. Sizes and timings, so we can find out why something failed. No text from your document.
• Anything you write in the feedback box.

• We do not sell or share your data with anyone beyond the services above.
• We do not train any model on your work.
• We do not read your manuscript. We could not — it is not stored.
• We do not run advertising or third-party trackers.

Write to korenma94@gmail.com and we will delete your email address, your library and your feedback. Since your manuscript was never stored, there is nothing of it to delete.

Cached bibliographic data about published papers is not personal information and stays, as it would in any reference database.

If your institution or your funder has rules about sending unpublished work to third-party services, those rules apply here. Some do. It is worth checking before you paste in a grant application under embargo.