Semantic Scholar

Semantic Scholar, from the Allen Institute for AI, is what academic search should have been a decade ago. Its index covers over 220 million papers across every discipline, and the difference from Google Scholar is structural: every paper is enriched with AI-derived TLDR summaries, citation context analysis that tells you whether a citing paper endorsed or criticized its source, and citation graphs that filter out the review-echo so common in scholarly networks. Searching a topic surfaces not just the best-known papers but the influential ones ranked by a model trained on actual citation behavior.

The features researchers actually use daily: the TLDR one-liners that triage a 40-paper reading list into a manageable afternoon; 'Citation Intent' showing which references are background vs. method vs. result; the Semantic Reader that adds inline definitions and figures in-context; and Research Explorer for tracing how an idea evolved across fields. Alerts and paper feeds keep a lab group current without anyone manually checking journals.

It is free, ad-free, and run by a nonprofit, which is rarer than it should be. Limits exist: coverage of very recent preprints lags arXiv by days, humanities indexing is thinner than STEM, and the UI occasionally feels utilitarian next to Connected Papers' visualizations. But as the primary discovery layer over the scientific literature — with an API for those building on it — Semantic Scholar is essential infrastructure for students, academics, and anyone doing evidence-based work.