Scite
Last reviewed
Jun 4, 2026
Sources
32 citations
Review status
Source-backed
Revision
v1 ยท 2,890 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Jun 4, 2026
Sources
32 citations
Review status
Source-backed
Revision
v1 ยท 2,890 words
Add missing citations, update stale details, or suggest a clearer explanation.
Scite (stylized lowercase as scite, at scite.ai) is a research-discovery and citation-analysis platform best known for Smart Citations, which use deep learning to classify each citation in the scientific literature as supporting, contrasting, or mentioning the work it references. Rather than counting citations, Scite extracts the sentence in which one paper cites another (the "citation statement") and labels how that paper actually used the reference, so a reader can see whether a study has been corroborated or disputed. The company was founded in 2018 by Josh Nicholson and Yuri Lazebnik and is headquartered in Brooklyn, New York. In late 2023 Scite was acquired by Research Solutions, Inc. (NASDAQ: RSSS), a provider of research-workflow software, in a deal with total consideration of about $20.9 million.[1][2][3]
Conventional bibliometrics treat every citation as equivalent: a paper that cites a study to refute it is counted the same as one that cites it as confirming evidence. Scite was built to address that limitation. Its core product, Smart Citations, analyzes the full text of citing articles, pulls out the exact passage where a reference appears, and uses a machine-learning classifier to assign the citation to one of three categories. The result is presented as a "citation tally" (counts of supporting, contrasting, and mentioning citations) attached to each indexed article, alongside the underlying citation statements so users can read the context themselves.[4][5]
The platform positions itself within the broader response to concerns about reproducibility and citation of unreliable or retracted research. By surfacing whether a finding has been independently supported or contested, and by flagging references that have received editorial notices such as retractions, Scite aims to help researchers, editors, and peer reviewers evaluate evidence rather than simply tally how often a paper has been mentioned.[4][6]
By the time it was acquired, Scite described its index as containing more than one billion Smart Citations; the company later reported that its content agreements with publishers gave it access to over 1.3 billion citation statements drawn from full-text articles, and a corpus of more than 250 million indexed items including journal articles, book chapters, preprints, and datasets.[2][7][8]
Scite grew out of work that co-founder Josh Nicholson had been thinking about since around 2014, when he and Yuri Lazebnik began discussing the problem of citations being used without regard to context, a concern tied to the wider reproducibility crisis in science. Nicholson holds a PhD in cell biology from Virginia Tech and had previously founded The Winnower, an open publishing platform, and served as chief executive of Authorea, a collaborative writing tool. Lazebnik is a cell biologist who had been a professor at Cold Spring Harbor Laboratory. The company was incorporated in 2018, and an early version of the product that attempted to qualify citations by intent was in use by early 2019.[9][10][1]
A third key contributor, Sean Rife, a psychology professor at Murray State University, joined the effort after Nicholson encountered his work online; Rife had built statcheck-style tooling for checking the accuracy of statistics in published papers. Rife co-authored Scite's foundational publications and is commonly named alongside Nicholson and Lazebnik as a co-founder of the company. Scite operated out of Brooklyn, New York.[9][1][11]
Much of Scite's early development was funded through competitive U.S. government research grants rather than venture capital. The company received a Small Business Innovation Research (SBIR) Phase I award from the National Science Foundation (award #1913619) to build "a deep learning platform to identify confirming and refuting citations," a Phase I SBIR grant being worth up to roughly $225,000. In October 2019 Scite was awarded an SBIR Fast-Track grant by the National Institute on Drug Abuse (NIDA), part of the National Institutes of Health, and it subsequently received $1.5 million from the Phase II portion of that Fast-Track award to expand the use of Smart Citations.[12][13][14]
Scite raised a comparatively small amount of private capital. Public funding databases record a seed round of about $1.93 million in October 2022, in addition to the federal SBIR grants described above. The company never raised a large venture round before its acquisition.[15][16]
On November 27, 2023, Research Solutions, Inc. announced an agreement to acquire Scite. Research Solutions, headquartered in Henderson, Nevada, sells the Article Galaxy research-workflow and document-delivery platform to corporate and academic R&D organizations, and it framed the deal as adding AI-driven discovery and evaluation to its product line. Roy W. Olivier, chief executive of Research Solutions, said that "similar to Research Solutions, the heart of scite's success is their unique relationships with publishers." Josh Nicholson, Scite's co-founder and chief executive, joined Research Solutions following the acquisition.[1][2]
The financial terms were disclosed through Research Solutions' securities filings and subsequent reporting. The companies entered into the agreement valued at $14.8 million on November 24, 2023, and the acquisition closed on December 1, 2023. The total purchase consideration, net of cash acquired, was approximately $20.9 million. The structure included an initial payment of about $7.2 million in cash and $6.3 million in stock, a holdback of roughly $0.2 million, and a contingent earnout whose fair value was about $7.2 million as of December 31, 2023; the consideration was split roughly evenly between cash and stock, with the earnout payable in equal quarterly installments over a two-year period.[3][2]
At the time of the deal, Research Solutions cited Scite's roughly 21,000 active business-to-consumer subscribers and a business-to-business customer base spanning corporate entities, academic institutions, and government agencies. As of October 31, 2023, the annualized value of Scite's combined B2C and B2B software subscriptions was about $3.6 million.[1][2]
Under Research Solutions, Scite continued to operate as a branded product (marketed as "Scite by Research Solutions"), and the parent company has used the acquisition to cross-sell Smart Citations alongside its existing document-delivery business and to broaden Scite's publisher network.[2][8]
A Smart Citation is produced by parsing the full text of a citing article, locating each in-text reference, and extracting the surrounding sentence or passage as a citation statement. That statement is then fed to a machine learning classifier that assigns one of three labels: supporting (the citing paper presents evidence or assertions that back up the referenced work), contrasting (it presents contradicting or disputing evidence), or mentioning (a neutral reference that neither supports nor contrasts). Citations that cannot be confidently classified into supporting or contrasting default to mentioning.[4][5]
The classification model has evolved over time. The system has used natural language processing methods and, in the version described in Scite's peer-reviewed work, a classifier built on SciBERT, a transformer language model pretrained on scientific text. Scite has reported tuning the deployed model so that each class is returned with a precision above roughly 80 percent, balancing precision and recall to keep false positives manageable, and it publishes per-category precision, recall, and F-measure figures for the classifier in operation.[17][18][5]
Scite obtains the full text it needs to extract citation statements through indexing agreements with publishers and from open-access sources, supplemented by metadata from services such as Crossref. This full-text access is what distinguishes Scite from citation indexes that rely on reference lists alone, because the surrounding context can only be read when the body of the citing article is available.[19][6]
Scite's approach was documented in peer-reviewed venues. A 2021 paper in Quantitative Science Studies (MIT Press), "scite: A smart citation index that displays the context of citations and classifies their intent using deep learning," was authored by Josh M. Nicholson, Milo Mordaunt, Patrice Lopez, Ashish Uppala, Domenic Rosati, Neves P. Rodrigues, Peter Grabitz, and Sean C. Rife, with Lazebnik acknowledged for help conceptualizing and building the system. A companion piece, "scite: The next generation of citations," appeared in Learned Publishing. These papers describe how machine learning made it feasible to extract citation context and classify citation types at scale.[5][20]
Scite offers a set of tools built on top of its Smart Citations index.
| Product / feature | What it does |
|---|---|
| Smart Citations | Classifies each citation as supporting, contrasting, or mentioning and shows the citation statements behind the tally for any indexed article. |
| Search / Citation Statement Search | Lets users search across the body of extracted citation statements and full-text content, filtering by citation type, section, and other facets. |
| Scite Assistant | An AI chat assistant, launched in early 2023, that answers research questions in natural language and grounds its responses in citations drawn from the index. |
| Reference Check | Analyzes a manuscript's reference list to flag retractions and other editorial notices and to show how each cited reference has itself been supported or contrasted. |
| Custom Dashboards | Aggregate Smart Citation data across a chosen set of papers (for example a lab, author, or topic) to see which are most supported or contrasted. |
| Badges | Free, embeddable widgets (a Citation Type badge and a Section badge) that display Smart Citation tallies on publisher and repository pages. |
| Browser extension and plug-ins | A browser extension and reference-manager integrations (including a Zotero plug-in) that surface Smart Citation tallies on sites such as Google Scholar and PubMed and inside users' libraries. |
Scite Assistant is an LLM-based chat tool, released in early 2023, that lets users ask research questions or give free-form prompts and returns answers grounded in the scholarly literature indexed by Scite. Early descriptions characterized it as combining a large language model (it was built around OpenAI's ChatGPT technology) with Scite's search, so that generated answers are accompanied by citations to specific papers and to the citation statements supporting each claim; hovering over a citation reveals the exact text and source. At launch the tool searched over roughly 179 million articles, book chapters, and preprints and more than 1.2 billion citation statements. Because the assistant cites both abstracts and citation sentences, Scite and outside reviewers have noted it can occasionally produce indirect citations that should be checked against the original source. The tool is an example of retrieval-augmented generation applied to scientific literature, and it competes with research assistants such as those from Semantic Scholar, Elicit, and Consensus.[21][7][22]
Reference Check is aimed at the publishing and peer-review workflow. An author, editor, or reviewer uploads a manuscript or pastes its reference list, and Scite scans each cited work to report whether it has received editorial notices such as retractions or corrections, how many times it has been supported or contrasted, and whether anything about the references warrants concern. The tool can surface citations to retracted papers in seconds, which is intended to catch reliability problems before submission or publication.[6][23]
Custom Dashboards let users group papers by DOI and view aggregate Smart Citation statistics for the group, including which articles in the set are most supported or most contrasted. Free-plan users are limited to a single dashboard of up to 100 DOIs, while paid users can create unlimited dashboards of up to 1,000 DOIs each. Separately, Scite offers free embeddable badges, a Citation Type badge and a Section badge, that publishers and repositories can place on article pages so readers can see Smart Citation tallies in context. Some platforms have implemented the badges across large article collections; for example, ScienceOpen added Scite badges to tens of millions of article records.[24][25]
Scite has historically used a freemium model. A free tier provides limited access, while a paid Personal plan unlocks full access to Smart Citations, full-text search, and the AI assistant; the Personal plan has been listed in the range of roughly $12 to $20 per user per month (with discounted annual billing). Organization and institutional plans, which add single sign-on, analytics, API access, and dedicated support, are sold via custom quotes rather than published prices.[26][27]
Because Smart Citations require full text, indexing agreements with publishers are central to Scite's business, a point Research Solutions emphasized when describing the acquisition rationale. Scite has built content agreements with more than 30 major publishers, among them Wiley, BMJ, Karger, Thieme, Sage, Cambridge University Press, Future Science Group, and Rockefeller University Press. The Rockefeller University Press agreement (announced in 2020) brought Smart Citations to journals including the Journal of Cell Biology and Journal of Experimental Medicine.[19][28]
After the acquisition, Research Solutions continued to expand this network. In 2024 it announced a collaboration with Karger Publishers to enrich Karger's journal content (around 100 medical journals) with Scite Smart Citations, and on June 18, 2025, it announced an indexing agreement with the American Society for Microbiology to add ASM's journal portfolio to the platform.[29][30]
On February 26, 2026, Research Solutions launched Scite MCP, a Model Context Protocol server that connects external AI tools, including ChatGPT, Claude, Microsoft Copilot, Cursor, and Claude Code, directly to Scite's index of more than 250 million scholarly items and its Smart Citations. The integration lets those AI agents ground their answers in peer-reviewed literature and distinguish well-supported findings from contested ones. Josh Nicholson, by then in a strategy role at Research Solutions, described the problem as AI tools being unable to tell "which findings are well-supported and which have been contradicted." Scite was also made available as a verified connector in Anthropic's Claude directory. At launch Scite MCP was offered to paid subscribers with access to open-access content, with discussions underway to extend coverage to paywalled material.[8][31]
Scite has been widely reviewed in the library and scholarly-communications community as a notable advance over raw citation counts, and many university libraries publish guides on using it. Reviewers have praised the value of seeing supporting versus contrasting citations and the Reference Check retraction screening, while cautioning that the automated classifier is imperfect, particularly in distinguishing supporting from contrasting citations, so the labels are best treated as a starting point for human judgment rather than a definitive verdict. Coverage in venues such as Quantitative Science Studies, Learned Publishing, and the Journal of the Medical Library Association helped establish the tool's credibility within bibliometrics and research-evaluation circles.[5][20][32]