Scite

What is Scite?

Scite is an academic search and citation analysis platform designed for researchers, students, editors, publishing houses, and R&D teams; its key feature isSmart CitationsWhile traditional citation counts merely indicate how many times a paper has been cited, Scite goes further by showing the context in which those citations occur, and uses algorithms to determine whether subsequent papers support the original research, provide contrary evidence, or simply mention it.

As of this verification, Scite has become a product under Research Solutions. Its official website indicates that it covers more than 280 million papers, preprints, books, patents, and datasets; it has analyzed over 1.6 billion citation contexts, and it works in partnership with more than 30 publishers.

The platform includes Assistant, Search, Reference Check, Dashboards, browser extensions, Zotero plugins, API, and MCP, and is primarily used via the Web.

How should Smart Citations be understood?

Scite extracts the quoted sentences from the text along with the surrounding context, indicating which paper the citation comes from and in which section it appears – whether in the introduction, methods, results, or discussion – while also providing automatic classification and a confidence level. The main categories include:

• Supporting：The context of the citation provides results that are consistent with or supportive of the cited research;
• Contrasting：When the results provided in the context of a citation are inconsistent, fail to repeat each other, or are opposite to one another, such an interface can also be described as contradictory.
• Mentioning：The paper merely uses, describes, or briefly mentions the cited works, without providing clear support or comparison;
• Unclassified：The system is not confident enough to classify it into the aforementioned category.

These labels are the model’s outputs.A certain quoted statementThis classification does not constitute a final judgment on the quality or authenticity of an entire paper. The same paper may be cited in different ways by various studies, and one source may refer to the same target paper multiple times within its text; therefore, the number of citation statements is not necessarily equal to the number of papers cited.

In tables of contents and paper writing, a single instance of “Supporting” cannot be interpreted as meaning “this conclusion has been proven,” nor can a single instance of “Contrasting” be equated with “the original paper is fraudulent.”

Core functions

1. Paper reports and citation context

By entering the paper’s title or DOI, or by opening a Scite report, it is possible to view the basic metadata, references, sources that cite this paper, as well as the numbers of supports, comparisons, mentions, and uncategorized items. Each Smart Citation displays the original citation statement, the source paper, the publication date, the chapter number, and the confidence level of the classification, thereby helping users to quickly find information related to reproducing the research, providing criticism, adopting its methods, or carrying out further developments.

Paper reports can serve as a basis for tracing back the introduction and discussion sections: first, read the highly relevant supporting and comparative evidence, and then consult more recent cited papers and reviews. However, the number of such sources cannot replace an assessment of their content.

Old papers tend to be cited more frequently due to their longer publication history, while new papers and those in niche fields receive fewer citations; moreover, it may be harder to publish papers with negative results.

2. Full-text and citation statement search

Scite not only searches for titles and abstracts, but also for citations within the full text that has been indexed. Users can ask questions such as “Which papers cite a particular method in their results?” or “In which studies has a certain hypothesis been compared?”, and they can filter the results based on criteria such as author, journal, year, publication type, chapter, open access status, type of citation, or status of withdrawal or correction.

The Search API also supports Boolean operations, exact phrases, and more than 25 filtering parameters. The results can be sorted by date, total citations, Supporting/Contrasting factors, or journal metrics. Full-text access depends on the publisher’s permissions, on the availability of open resources, and on the quality of the parsing; not all paid articles are available in their entirety.

The presence of metadata does not mean that the entire text can be searched.

3. Scite Assistant

Scite Assistant converts natural language queries into search strategies, retrieves and reorders relevant scientific documents from the Scite index, and then uses a large language model to generate answers accompanied by citations. It supports multiple rounds of follow-up questions, table format, filtering by year and publication type, use of either abstracts or full texts, querying based on specific DOIs, as well as limiting the search to one’s own Dashboard collection.

The Assistant allows sorting by relevance, date, number of traditional citations, supporting citations, comparative citations, or journal indicators; it also enables the search for academic materials and patents. The table format is suitable for comparing individual studies in terms of their methods, results, and other aspects.

The narrative mode is suitable for obtaining a quick overview.

The citations included in the responses should be checked one by one; the model may still misinterpret the context, overlook certain constraints, or compare different research designs at the same level.

4. Reference Check – Verification of references

Reference Check is used to examine the references included in a manuscript or document. The system identifies the citation entries and links them to Scite data, indicating which references have been withdrawn, are the subject of editorial attention, require corrections, or have numerous comparable citations, while also showing the context in which those references are used.

It is suitable for self-checking before submission, preliminary screening by journal editors, updating of reviews, and institutional research integrity processes.

A reference check does not prove that a citation is used correctly. Entries that the system fails to match may simply be due to formatting, DOI, or parsing issues;

Papers that are extensively compared may still yield valid results under specific conditions.

The author needs to return to the main text to check whether the cited paper indeed supports the current sentence, whether it refers to the original research, and whether the version, page numbers, and scope of the conclusions are consistent.

5. Dashboards and reminders

Dashboards are used to store collections of papers, to track the changes in Smart Citations for a set of documents, and to restrict Assistant’s access to specific collections. Users can create different collections based on topics, projects, products, or systematic reviews, in order to continuously monitor the emergence of new Supporting or Contrasting elements.

The API documentation refers to such collections as Collections, and provides interfaces for creating, reading, updating, and deleting them.

It is useful for keeping track of key papers, authors, or research areas, but it should not be the only method for monitoring literature. For formal projects, it is necessary to retain database search queries, dates, exported files, and filtering logs in order to enable reproduction.

6. Browser extensions and Scite Badge

Official browser extensions enable the display of Scite Badges on pages such as PubMed, PMC, Wikipedia, ScienceDirect, and eLife; this allows users to see the numbers of Supporting, Contrasting, and Mentioning references for a paper while reading the page, and to navigate to those references. The badges can also be embedded on journal, laboratory, or knowledge base pages to display citation statistics based on DOI.

The extension reads the currently visited page to identify the DOI and insert the relevant information; before installation, it is necessary to check the store’s permissions and privacy policy. The absence of a badge on the page does not mean that the paper has not been cited – it could be due to an absent DOI, changes in the site’s structure, a mismatch between the extension and the page, or the fact that Scite has not yet indexed it.

7. Zotero plugin

Scite’s official Zotero plugin displays information on support, comparison, and the number of citations within the Zotero library, making it useful for quick filtering within existing reference management workflows. The plugin provides XPI installation packages via GitHub; users should select the official release version that is compatible with their current Zotero version, and after upgrading they should check to ensure that the column displays and data updates properly.

8. MCP integration with AI tools

Scite MCP allows full-text literature searches and Smart Citation data to be integrated into ChatGPT, Claude, Gemini, Copilot Studio, Claude Code, and other compatible clients. Once authorized, users can request the search for papers, verify claims, or compare supporting evidence within existing AI conversations, without the need to manually copy each result.

MCP utilizes OAuth 2.1 and PKCE to support dynamic client registration; in server-to-server scenarios, API credentials issued by Scite can be used to obtain short-term access tokens.

The official website allows users to create an account to try out MCP, but ongoing use and advanced access typically require a Premium subscription or institutional credentials.

When connecting to third-party AI, it is also necessary to consider which services the text in question, the conversation content, and the search results will pass through.

9. API

The Scite API provides paper metadata, Smart Citation statistics, citation relationships, full-text and citation searches, Reference Check, Assistant, a set of dashboards, and an MCP interface. Publicly available papers and the tally endpoint can be accessed at a limited rate.

Searching, citing images, performing Reference Checks, using the Assistant, and for commercial purposes generally require an authorization token and a separate agreement.

The API documentation specifies that individual subscriptions do not include, by default, a license for the Search API intended for use in commercial or research system integrations. Development teams must request credentials from sales in order to determine the scope of data usage, the restrictions on caching and redistribution, as well as the call rates and associated costs.

Patents, clinical trials, funding, and other Resolute evidence datasets also utilize separate access rights.

Data coverage in collaboration with publishers

The differences in Scite stem from the context in which citations appear, and not only from the aggregated metadata. It obtains processable text through open-access content as well as through direct collaborations with more than 30 publishers such as Wiley and SAGE.

According to the information on the official website, the platform covers more than 280 million sources and over 1.6 billion citation contexts, serving more than 2 million users.

The number of entries will continue to increase, and it is necessary to distinguish between “source records,” “fully searchable text,” “parsed citations,” and “automatically categorized items.” Some books, patents, preprints, datasets, or old scanned papers may only contain metadata.

Non-English literature, formulas, tables, and reference formatting also affect extraction.

For systematic reviews that require a comprehensive search, discipline-specific databases should still be used, along with manual removal of duplicate entries.

Price and subscription details

The Scite official website offers a free trial, and subscriptions are available for individuals, teams, universities, companies, and publishing institutions. The Premium version for individuals typically provides access to features such as Assistant, advanced options for Smart Citations, search functions, Reference Check, Dashboards, and MCP.

Institutional plans allow for access, analysis, training, and support to be arranged according to the campus or organizational structure.

During this verification, the public pricing page imposes restrictions on automatic access for users who are not logged in, and the search results on the official website do not display clearly the individual monthly and annual fees that can be used for verification. As a result, the catalog does not use the old figures such as $20 per month or $144 per year as the current prices. Users should check the current currency, trial period, option to pay monthly or annually, student discounts, and taxes on the purchase page of their Scite account.

Institutions, publishers, and API providers must contact sales to inquire about prices.

In particular, the API needs to be distinguished from the individual Premium version: the Search API, which is intended for commercial or research purposes as stated in the documentation, requires a separate license and is not included in the individual plan. MCP allows users to create an account to try it out, but the ongoing permissions for OAuth integration are determined by the current Premium or organizational subscription.

GitHub and the open-source status

Scite’s official GitHub organization is now displayed under the name Research Solutions; it offers browser extensions, Zotero plugins, Scite Badges, VIVO integration, MCP Skills, as well as various research or data processing projects. Some of the repositories related to browser extensions, Badges, and MCP Skills are licensed under the MIT license, while there are also archived repositories and code that has been forked from external projects.

Scite’s core website, the database with 1.6 billion citations, the classification models, Assistant, Reference Check, and the business API are not fully open-source products.There is no officially available local version of the Scite community edition. The fact that a certain plugin is open source means that its code can be used in accordance with the relevant license, but it does not mean that the publisher’s full texts, Smart Citation data, or cloud services can be copied freely.

Scite Usage Tutorial

Complete a basic task.

1. Clarify the issue, time frame, location, source priority, and output format;
2. In Scite, upload materials for which you have permission to use, or enter search queries;
3. First, establish a framework using paper reports and the context of citations, and then seek additional evidence through searches of the full text and citation statements.
4. It is necessary to distinguish between factual information from the source, the author’s opinions, and AI-generated conclusions.
5. Check each item for dates, numbers, the original location, and any conflicting evidence;
6. The conclusions are manually revised, the verification time is recorded, and then they are published;

Create reusable professional workflows

1. Break down complex topics into four categories of questions: background, data, comparison, and conclusions;
2. Combining paper reports with citation contexts, searching through the full text and citation statements, along with Scite Assistant, constitutes a set of standard research steps.
3. Give priority to using the official website, research papers, regulatory documents, and raw data;
4. A second person is assigned to review conclusions that are considered high-risk;
5. Save queries, evidence, versions, and unresolved issues;
6. Re-run after the data changes and update the conclusions;

Which users is it suitable for?

• Researchers who need to determine whether a paper has later been supported or challenged;
• Students and teachers who write papers, proposal reports, reviews, and grant applications;
• A team responsible for literature review, drug development, medical affairs, and evidence assessment;
• Editors and publishing houses that check the references in manuscripts, as well as those involved in withdrawal requests and disputes;
• Users who view the context of citations within Zotero or in the browsing process;
• Developers who wish to integrate scientific literature evidence into AI agents, knowledge bases, or internal systems.

Product advantages

• It not only counts the number of citations but also shows the specific citation sentences, chapters, and automatic categorization.
• It supports the identification of supportive, comparative, and generally mentioned follow-up studies;
• It covers both free and partially paid content, and works directly with various publishers;
• The Assistant’s responses can be linked to actual papers, and it supports tables as well as various filtering options.
• Reference Check is suitable for batch verification before submission and during the editing process;
• It offers various ways of access, including browsers, Zotero, APIs, and MCP;
• The official team has made some plugins and integration codes available, to facilitate the expansion of reading workflows.

Restrictions and Precautions

• The SmartCitation classification is carried out by a model, and it may misclassify background descriptions, reuse of methods, polite citations, or complex discussions.
• The classification confidence does not indicate the credibility of a paper, nor does it reflect sample quality, statistical power, bias, conflicts of interest, or the level of evidence.
• A proper evaluation of evidence requires reading the paper and taking into account the research design;
• Having many citations does not mean high quality, and having numerous supporting elements does not imply that the conclusions hold true for all populations and under all conditions.
• Differences may arise from various experimental conditions, endpoints, doses, samples, or analysis methods.
• Scite is suitable for helping users find the context in which a text should be read, but it cannot replace peer review, tools for assessing risk bias, or evaluation frameworks such as GRADE.
• Assistant is still a generative AI, and it may miss determiners, combine sentences incorrectly, or produce inaccurate statements.
• The content related to medicine, law, policies, and investment must be checked against the original text;
• Before uploading unpublished manuscripts or institutional documents to ReferenceCheck, Dashboard, or external MCP clients, it is necessary to verify the rules regarding confidentiality, copyright, data retention, and organizational security.

Frequently Asked Questions

Is Scite free?

The official website offers a free trial, as well as access to some papers and statistical data; however, full functionality of Assistant, Search, Reference Check, Dashboards, and MCP requires a personal Premium subscription or an institutional subscription. The current pricing is indicated on the account purchase page.

Can Smart Citations determine whether a paper is authentic or not?

No. It classifies the context of citations, helping to locate supporting, comparative, and referenced materials.

To determine whether a paper is reliable, it is necessary to evaluate the research design, the data, the statistical methods used, the possibility of reprocing the results, potential biases, and any status changes regarding withdrawal of the paper.

Can Scite search the full text?

It is possible to search for the full texts and quoted sentences that have been obtained and parsed, but it is not guaranteed that the full text of all 280 million records will be available. The extent of coverage depends on whether the content is in the open access domain, on the terms agreed with publishers, and on the quality of parsing.

Does Scite support Zotero?

Supported. The official Zotero plugin allows for the display of statistics related to Supporting, Contrasting, and Mentioning elements in the bibliography; the installation package is available on the official GitHub Releases page.

Does Scite offer MCP and APIs?

Available. MCP can be connected to various AI clients;

The API covers search, papers, citation statistics, citation graphs, Reference Check, Assistant, and collections.

Using the Search API in commercial or research systems requires a separate license.

Is Scite open source?

The core platform is not open source. The official team has made available certain components such as browser extensions, Zotero plugins, Badge, and MCP Skills, but the Smart Citation database, the models, and the cloud-based services remain proprietary.