How to search 250 million papers with the Literature Finder

Most academic search tools make you choose: search one database at a time, match the exact keywords you typed, and hope the paper you need used the same words. The Literature Finder does the opposite. One relevance search reaches CiteDash's own indexed corpus plus a live sweep of OpenAlex, PubMed, Semantic Scholar, and arXiv (over 250 million works between them), matches on meaning rather than exact wording, and returns a single reranked list. This guide covers how to drive that search well.

One search, both scopes

Type a topic, a question, or a set of keywords and search. Two things happen at once. The internal corpus is searched by meaning (your query is embedded and compared against paper abstracts) and by keyword, while a live query fans out to the four external sources at the same time. Every candidate, internal and external, is cross-encoder reranked together into one ordered list (up to 250 results), so a brand-new arXiv preprint and a paper already in your corpus compete on the same relevance scale.

External hits that are not in your corpus yet are marked "wider literature" and stay read-only until you act on one. Saving or reading such a paper mints it into the corpus (a step CiteDash calls "fetch to cite"), and it is pulled in for the next search too.

Point at an exact paper with a DOI

When you already know the paper, do not describe it, paste its DOI. The search box detects a DOI (a bare 10.xxxx string, a doi.org URL, or a doi: prefix) and routes to an exact lookup instead of semantic search. This matters: embedding a DOI string would return the nearest strangers, not the paper you meant. The exact match resolves straight to the one record you asked for.

Read the badges before you save

Every result wears honest access and trust signals, so you can judge each one at a glance:

• Full text: the parsed PDF is already in the corpus. It opens instantly in the Library and is ready to cite.
• Open access: a fetchable open-access PDF exists, so "Read From Library" pulls it in and parses it.
• OA, no direct PDF: flagged open access but only a landing page or DOI resolver. View it at the publisher, or upload a copy.
• No PDF: not open access. Reach it via the publisher link, or upload your own legal copy.
• Retracted: shown in red. Retracted papers can never be added to your library or cited, though the view links stay so you can examine the retraction.

Filter and sort the result set

A control strip sits above the results and applies live. The Filter row narrows what the search returns: a year range, which of the four sources to include, an author name, and a "With PDF only" toggle that keeps papers with an accessible open-access PDF. The View row shapes how the set is shown: how many to display (20, 50, 100, or all 250), and the sort order (most relevant, least relevant, most cited, or newest). Retracted papers are hidden by default; a "Show retracted" toggle brings them back, still badged and still uncitable.

A per-search stats rail totals what you found: papers shown, the year span, how many have an accessible PDF, how many carry an abstract, and a breakdown by source.

Save a search and let it keep working

You rarely find everything in one sitting. Save a query as an alert and new papers matching it surface automatically over time, feeding the screening queue of a living review in the Synthesis Lab. Separately, every search you run is written to a server-backed history, so you can restore any past search with one click, even on a new device or after a cleared cache.

What to save, and the one honest limit

Saving a paper is how your project library grows, but saving is not the same as being able to cite it. CiteDash only lets generated, verified text cite a paper whose full text it holds, because that is the only way a claim can be checked against a source. So the practical rule when searching: prefer results badged Full text or Open access, since those become citable automatically. For a paywalled paper you have legal access to, save or star it now and upload the PDF in the Library to make it citable later.

For a deeper look at each paper once it is in your library, read "Reading papers with sentence-level provenance." To discover more by following citations rather than keywords, read "Find the right sources with citation maps and AI recommendations."