Scite

If you've asked ChatGPT for sources on a research question, you know the problem: it'll give you a confident answer and a list of citations, and some of those papers won't exist. Google Scholar has the opposite problem. The papers are real, but it won't tell you whether a finding has held up or been knocked down by later work. Scite sits between the two. It's an AI assistant that answers research questions, but every claim links back to the specific sentence in the specific paper it came from. It reads across what it says are 323M+ full-text scholarly articles, not just abstracts.

The real differentiator is Smart Citations. Instead of a raw citation count, Scite shows how later research has supported, contradicted or simply mentioned a paper, a claim, a method, a journal or an author. That's the feature its testimonials keep pointing to, and it's useful. A citation count of 400 means little if half of those papers are disputing the result. Scite also leans on licensing. It claims direct agreements with Wiley, SAGE and 42+ other publishers plus the open access corpus, so it can read full text that tools stuck behind paywalls can't. Coverage goes past journal articles into preprints, patents, clinical trials, grants and datasets, though some of those datasets are Pro-only.

Scite also doesn't ask you to abandon the tools you already use. There's an MCP server that plugs Scite into Claude, ChatGPT or any MCP-compatible assistant, a Zotero plugin, a browser extension that shows citation badges on pages you visit, and an API for teams building their own research tools. Pricing starts at $12/month billed yearly for Basic (about $20/month if you pay monthly), with Pro at $40/month yearly for API access and bigger limits. For anyone who has to defend a claim to a reviewer, the verification angle is the reason to pay.