Lit Review Backfill
Already drafted the lit review? Paste each claim that needs a citation. Find Source returns the papers that actually support it, so you can skip the "I know I read this somewhere…" search loop.
Paste a sentence. Get back the peer-reviewed papers that actually support it. A direct quote works. A paraphrase works. That half-remembered stat from a podcast last Tuesday works. Your text lives in memory and is gone the second we send the result back. Nothing logged. Nothing fed to a training set.
Your answer will appear here.
Most citation tools run one direction. You hold a paper, you want the citation. Find Source goes the other way. You wrote a sentence. You want to know who said it first. Paste it. Five academic databases fire in parallel. Real peer-reviewed papers come back.
Here's the guts of it. A reasoning model reads your sentence and pulls out the concept, not the keywords. "Working memory capacity decreases under cognitive load" turns into a query for working memory, cognitive load, dual-task interference. That query hits Crossref, Semantic Scholar, arXiv, OpenAlex, and PubMed at once. The results get deduped and ranked. A second pass reads the candidates and tells you which ones back your claim. Which ones are just adjacent. Which ones to ignore.
Academic databases run first, always. That's the peer-reviewed material you'll actually cite. If they come back thin (low overlap with your wording), a Brave web search fires as fallback for grey literature, government reports, the industry whitepapers academic indexes don't carry. Anything off the open web gets tagged supplementary. Your call whether it clears the bar.
Already drafted the lit review? Paste each claim that needs a citation. Find Source returns the papers that actually support it, so you can skip the "I know I read this somewhere…" search loop.
"Studies show that…" but which studies? Paste the claim and the agent surfaces the original peer-reviewed source, not the news article that summarized it.
Cited paper got retracted? You need a replacement that makes the same argument. Paste the surrounding sentence; Find Source returns clean alternatives.
Use Find Source as a snowball. Each query returns a ranked list of conceptually-similar papers, which is one of the fastest ways to map a literature you don't yet know.
Before you cite a sweeping statement from another paper, paste it into Find Source. See whether the underlying empirical work actually supports the claim or whether the cited author is overreaching.
Working at the boundary of two fields? The five-database search covers medicine, social sciences, physics, computer science, and humanities. Nothing from outside your home field slips through the cracks.
Scholar's a search engine. You give it keywords, it gives you a list. Find Source is an agent. You give it a sentence and it works out the underlying concept, hits five databases at once, dedupes and ranks the haul, then writes a short note on which candidates actually support what you wrote. The output is an answer, not a results page.
Five, every query. Crossref for 130M+ DOI-registered records across every discipline. Semantic Scholar for 200M+ papers with rich metadata. arXiv for pre-prints in physics, math, CS, quantitative biology, and economics. OpenAlex for 250M+ works across all of them. PubMed for 35M+ biomedical citations. If those come back thin, Brave fires as a web fallback for grey literature.
Yes. Quotes are the easy case. Verbatim text gives the agent something concrete to lock onto. Paste it with or without quote marks. The original paper is usually top of the list. Short or generic quote? Paste the surrounding sentence too so the agent has some context to work with.
Broad claims like "exercise is good for mental health" return a flood of candidates, because the literature is huge. You'll get the ranked list plus a nudge to narrow the scope. Add a specific population, intervention, or outcome and the citation set sharpens fast. One thing: paste the sentence as it sits in your draft, not a stripped-down keyword version. The concept extractor needs the context.
Coverage is strongest for English-language work because the underlying databases lean that way. OpenAlex and Crossref do index multilingual records and the agent surfaces them when they fit. For a primarily non-English literature (a French historiographical debate, say), a discipline-specific national database will still beat us on coverage.
The agent will tell you when academic results are thin and the web fallback didn't turn up anything trustworthy. Usually means one of two things. Either the claim is novel enough that nothing directly supports it (rephrase, or split it into smaller pieces). Or the supporting work lives somewhere we don't cover yet. Try once more with domain-specific terminology before you walk away.
Read it first. Seriously. Don't cite something you haven't read. Then hit "Cite this" to send it to the Citation Generator, or save it to your Works Cited library. The agent finds the candidate. Whether it actually backs your specific claim is on you.
10 free credits on signup. 3 more each day you log in. A query runs 2–3 credits, depending on how many fallback searches fire. Top-up packs live on the pricing page. Credits don't expire.
No. ChatGPT will happily invent a plausible-looking citation because it optimises for fluency, not truth. Find Source can't do that. Every result has to come back from a real academic database first. The reasoning model only ranks and summarises what the search returned. It can't make papers up.
Already have a reference and want to confirm it's accurate? Run it through the AI Citation Checker. Every field gets verified against Crossref, Semantic Scholar, and PubMed.
Once Find Source surfaces the right paper, drop the DOI into the Citation Generator to format it in APA, MLA, Chicago, Harvard, ACS, AMA, or any of 10,000+ CSL styles.
Read How to Write a Literature Review for a step-by-step approach that pairs well with Find Source as a discovery engine.
Find Source returns peer-reviewed papers, pre-prints, and grey literature. Our guides on scholarly sources and peer review walk through how to evaluate each.
Once you've found the source, hand the PDF to Verify Source. It reads through the references and lines up each cited claim against what the original paper actually said.
Database results aren't all created equal. How to Find Credible Sources walks through evaluating venues, authors, and methodology before you commit to citing a paper.
Once the sources are gathered, the next risk is paraphrasing too closely. Pair Find Source with the Plagiarism Checker and the Writing Assistant to keep your draft clean.
Click "Save" on any result to add it to your Works Cited library. Build the bibliography incrementally as you research, then export the finished list when you're ready to submit.
AI-powered features require an account. The Citation Generator stays free forever.