Semantic Scholar

Semantic Scholar is a free, AI-powered search and discovery tool for scientific literature, built by a research and product team at Ai2, the Allen Institute for AI, the nonprofit founded by Microsoft co-founder Paul Allen. It launched in 2015 and indexes papers from all fields of science, sourced through publisher partnerships, data providers and web crawls. Ai2's own pages size the corpus differently: over 200 million papers on the About page, 214 million on the API page, and 238 million in the site's search bar on 30 September 2026, which the FAQ names as the current count. Coverage focuses on English-language publications, book coverage is very limited and patents are not included. Search results and paper pages carry signals from Ai2's own models. TLDRs are one-sentence AI summaries of a paper's objective and results, currently limited to computer science and the biomedical domains. Highly Influential Citations flag citations where a machine-learning model judges that the cited paper had a significant impact on the citing one, and citation intent labels each citation as background, method or result extension. Results can be sorted by relevance, citation count, influence or recency. Ask This Paper answers questions about the paper being viewed, with supporting statements quoted from it, using OpenAI's GPT-3.5 model; the FAQ says it has been tested only on English-language papers and works on a subset of papers. Semantic Reader, still in beta, shows citation details in context while reading and is available for arXiv papers only. Topic pages, generated with large language models, cover computer science only. No account is needed to search. A free account adds a library with folders, shared folders and bulk citation export, Research Feeds that recommend new papers based on the papers in a folder, email alerts for new papers and citations, and claimable author pages. Papers can be saved to Zotero from paper pages and from Semantic Reader through the Zotero browser extension, citations can be copied in BibTeX, MLA, APA and Chicago formats, institutional sign-in runs through OpenAthens, and Chrome and Firefox extensions search the site from any page. There is no mobile app. For developers, the Semantic Scholar Academic Graph API is free. Most endpoints work without authentication, rate-limited to 1,000 requests per second shared among all unauthenticated users. A free API key is required for some endpoints and starts at an introductory limit of one request per second. Monthly bulk datasets and the S2ORC full-text corpus are also free, and use is governed by an API License Agreement last updated on 17 May 2023. Several research tools in this index name Semantic Scholar as a data source, including Elicit, Consensus, Undermind, Litmaps and ResearchRabbit, and Ai2's API page quotes the founders of Litmaps and Connected Papers on building on it. Ai2 also runs a Model Context Protocol server, the Asta Scientific Corpus Tool, documented on its Asta resources pages with the endpoint asta-tools.allen.ai/mcp/v1. The documentation lists seven tools covering paper lookup, citations, author search, an author's papers, relevance and title search, and snippet search over full-text passages, and an optional key, requested through a form, raises rate limits. That endpoint and a second, undocumented one at api.semanticscholar.org/mcp both accepted an MCP connection without a key on 30 September 2026, and the second also lists a batch paper-lookup tool. Community-built wrappers around the public API are separate projects and not this server. Semantic Scholar has no paid tier: search, the AI features, accounts, the API and the datasets are all free. Ai2 is a nonprofit, and the About page presents equal access to science as one of the project's values. Ai2's privacy policy, last updated 19 February 2025, lists saved papers, email alerts, search queries and prompts among the data it collects and says it uses them to improve its services and for research. It does not state whether that data is used to train models. Questions asked through Ask This Paper are sent to OpenAI together with the paper's content. Semantic Scholar finds and annotates individual papers; it does not write a synthesis across many papers or extract data into tables. Teams that want answers synthesized across the literature can look at Elicit (from $49/mo billed annually) or Consensus (from $12/mo billed annually). Teams that want to see whether later work supports or contrasts a finding can look at Scite.ai, and teams that want to explore a field through a citation map can look at ResearchRabbit or Litmaps. Current state Q4 2026: the search bar counts 238 million papers, and Ai2 documents the Asta Scientific Corpus Tool MCP server. The s2-folks GitHub repository, the API community space, has been archived since January 2025, and the Semantic Scholar Open Data Platform paper (Kinney et al., arXiv 2301.10140) was last revised in April 2025. G2 lists an unclaimed Semantic Scholar profile with no reviews, and Product Hunt shows two reviews.