Draft:Consensus (app)

Consensus
Developers	Consensus NLP, Inc.
Release	2022 (2022)
Operating system	Web, iOS
Type	Artificial intelligence, Academic search engine
License	Freemium
Website	consensus.app

Consensus is an artificial intelligence-powered academic search engine developed by Consensus NLP, Inc. It is designed to help researchers, students, clinicians, and the general public search, analyze, and synthesize peer-reviewed scientific literature. The platform searches a database of more than 220 million research papers and uses AI to retrieve relevant evidence, summarize findings, and indicate where the scientific literature agrees or disagrees on a given question.

Unlike general-purpose large language model chatbots, Consensus retrieves and grounds its answers in published research before generating any summary, an approach the company describes as a way to reduce AI hallucinations by tying every response to a real, citable paper.

Overview

Consensus operates as a search engine focused on scholarly and scientific literature. Users pose natural-language questions and receive results drawn from a large corpus of academic papers, along with AI-generated summaries of the underlying evidence. The company positions the product as an alternative to traditional research tools such as Google Scholar and PubMed, and more recently describes it as moving "beyond search" toward a broader suite of research tools.

According to the company, every AI-generated response is produced only after the literature has been searched, so that summaries are grounded in retrieved papers rather than in unsourced model output. The company states that it is "not an AI lab" and does not use customer data to train AI models.

History

Consensus was founded in 2021 by Christian Salem and Eric Olson, former Division I athlete teammates at Northwestern University. The two have said they came from families of researchers and teachers and shared an interest in science while feeling like "outsiders" to academia. Olson serves as chief executive officer and Salem as chief product officer.

The product launched publicly in 2022. Originally a remote-first company, Consensus is now headquartered in San Francisco. The company's stated mission is to "make the world's best knowledge accessible to everyone."

The company's advisory board has included Nicholas A. Christakis, Stephen Wolfram, metascience researcher Jevin West, computational neuroscientist Konrad Kording, and nutrition scientist Layne Norton.

How it works

Consensus describes its search process as proceeding in three stages. First, the system casts a wide net across its database using a hybrid retrieval method that combines semantic search (using AI embeddings to capture the intent of a query) with traditional keyword search (using the BM25 algorithm for exact term matching). Each paper is assigned a relevance score by comparing the query against titles, abstracts, and, where available, full text.

Second, the top 1,500 most relevant papers are re-ranked using research-quality signals, including recency of publication, citation count, and journal impact and reputation.

Third, the top 20 papers are re-ranked a final time using a larger, more powerful AI model, while continuing to weigh recency, citations, and journal reputation. AI is then applied in two ways: to analyze individual papers in depth, and to synthesize findings across multiple papers.

Features

Consensus Meter

The Consensus Meter is a feature that visualizes the degree of agreement or disagreement in the literature in response to a yes-or-no research question. After a user submits such a question, the system analyzes the top 20 returned papers and categorizes their conclusions as "Yes," "No," "Possibly," or "Mixed," with each cited paper color-coded to its stance. The Meter requires a minimum of five relevant papers to display a result. The company states that the underlying results are drawn from word-for-word quotes extracted from papers rather than generated text, in order to reduce hallucinations.

Pro Analysis and Ask Paper

Pro Analysis summarizes the key findings from the top papers returned for a search and provides a higher-level synthesis with citations. Ask Paper allows a user to interact with the full text of an individual paper to obtain answers about its methods, findings, and other details. A related feature, Study Snapshot, extracts structured information about an individual study.

Research agent

In 2025 and 2026 Consensus expanded from a single-step search tool into what it calls a research agent capable of multi-step tool use. The agent can chain together actions such as citation crawling (following references forward and backward from a seminal paper), DOI lookup, author search, and "similar papers" discovery. It can run multiple searches in parallel, read the full text of results, and assemble side-by-side comparison tables across studies. It can also analyze a user's saved collection, build a "research gaps" matrix, and run additional searches to surface papers that fill identified gaps.

Full-text analysis

Consensus analyzes the full text of papers rather than only titles and abstracts. The company states that this approach makes its retrieval approximately 10% more accurate than Google Scholar in internal head-to-head testing. For open-access papers, the full PDF is pulled into the platform, and AI-generated quotes link directly to the corresponding passage in the source document. The company reports that full-text access to paywalled articles is enabled through publisher partnerships (see below).

Library and reference management

The Consensus library functions as a reference manager. Users can import papers from Zotero in one click, or from BibTeX, RIS, or PDF files, and the company reports that more than 1.2 million papers have been imported. Within the library, users can search collections or their entire library, find citations to support a statement, build comparison tables of saved papers, identify gaps in a collection, and generate bibliographies. The company states that uploaded papers remain private to the user's account and are not used to train AI models.

Model Context Protocol server

Consensus offers a Model Context Protocol (MCP) server that connects AI assistants such as ChatGPT and Claude, and other MCP-compatible clients, directly to its index of more than 220 million papers. Through the MCP server, an external AI assistant can search peer-reviewed papers, build multi-step search strategies (including boolean and PICO frameworks), generate structured outputs such as reading lists and literature reviews, and run pre-built "Skills" for tasks such as grant research, literature reviews, and curriculum development. The company has reported more than five million uses of the MCP server.

API

Consensus provides an API that allows developers to integrate its search results into external applications, such as private chatbots and internal research tools. The company has advertised the service at a per-request price plus an annual platform fee, returning data including the top abstracts, metadata, study-snapshot data, summaries, author, study type, citation count, journal, and year.

Data and coverage

Consensus searches a database the company describes as containing more than 220 million research papers, including peer-reviewed journal articles and preprints, spanning fields from astronomy to sociology to biochemistry. Paper data is sourced largely through a partnership with Semantic Scholar, which itself draws on more than 50 direct partnerships with publishers, data providers, and aggregators covering over 500 journals, university presses, and scholarly societies, including preprint servers such as arXiv, bioRxiv, medRxiv, and SSRN. The corpus includes the entirety of PubMed (about 37 million biomedical studies) and a majority of journals ranked in the top quartile (Q1) by SCImago Journal Rank. Company documentation has described the index as being updated weekly in some materials and monthly in others.

Publisher partnerships

Consensus has formed licensing and indexing partnerships with several academic publishers to enable full-text analysis of paywalled articles within the platform. Named partners include Wiley, Taylor & Francis, SAGE Publishing, the American Chemical Society (ACS Publications), the American Association for the Advancement of Science (AAAS), and the American Psychological Association (APA). Through these arrangements, the platform can analyze the methods, results, and discussion sections of participating articles, although access to read or download the full article may still depend on the user's own institutional or personal subscription.

Funding

In 2021, Consensus raised an initial round of funding. In 2024, the company raised an $11.5 million Series A led by Union Square Ventures, with participation from investors including Nat Friedman and Daniel Gross. The round was reported by Bloomberg News and other outlets.

In May 2026, Consensus announced a $30 million round led by GreatPoint Ventures, whose partner DJ Patil, a former Chief Data Scientist of the United States, joined the company's board. Existing investors Union Square Ventures, NFDG, and Draper Associates also participated. The company said the funding would support expansion beyond search toward what it calls an "AI operating system for research."

Usage and adoption

Consensus has reported a range of usage figures across its materials. At the time of the May 2026 funding announcement, the company stated that its literature-review tool served more than 2.5 million monthly active users, and its marketing materials describe the product as used by researchers and students at more than 10,000 universities worldwide. The company's about page has cited figures including more than 7 million total researchers, students, and clinicians; users representing more than 12,500 universities; and more than 75 million research questions handled.

Reception

A peer-reviewed rapid review published in Cureus in 2025 examined how Consensus was used and reported across academic publications. Following PRISMA guidelines and analyzing ten papers, the authors concluded that the application remained "underutilized and significantly underreported" in the academic literature, and called for clear quantitative benchmarks, such as retrieval accuracy, search speed, transparency, and reproducibility, to enable rigorous head-to-head assessments of such tools. The review noted that Consensus ties its outputs to real studies and does not fabricate citations, but that it inherits the limitations of the underlying literature, including mixed-quality studies, publication bias, and unresolved debates, and that AI-generated summaries can still be flawed even when the cited sources are genuine.

Several university libraries, including those at the University of Virginia, Oklahoma State University, and Bentley University, have published research guides describing Consensus and, in some cases, providing institutional access to it. Library and academic-skills reviewers have generally described the tool as useful for quickly orienting to a topic and building an initial reading list, while advising that AI-generated summaries be verified against the underlying papers. The platform's grounding of answers in citable papers has been noted as a distinguishing feature relative to general-purpose chatbots.

Consensus has also been examined in comparative evaluations of AI research assistants alongside tools such as Elicit, Scite, and Scopus AI. A 2026 scoping review of AI tools for automating evidence synthesis, published in the Journal of Medical Internet Research, identified 65 such tools, including Consensus, and noted a lack of studies pragmatically comparing different AI approaches or evaluating their effectiveness in real-world review settings. Library evaluations have characterized Consensus as oriented toward fast, evidence-backed answers and directional summaries of the literature, in contrast to tools designed primarily for structured screening and data extraction.

Business and technology outlets, including Bloomberg News, have covered the company primarily in connection with its funding rounds.

See also

• Google Scholar
• Semantic Scholar
• PubMed
• Elicit
• Scite
• Academic search engine
• List of academic databases and search engines

External links

• Official website

Category:Artificial intelligence applications Category:Academic search engines Category:Online databases Category:Science websites Category:Companies based in San Francisco Category:American companies established in 2021 Category:2021 establishments in the United States