Consensus is an AI-powered academic search engine that fundamentally changes how researchers interact with scientific literature. Instead of returning a list of paper titles like Google Scholar or PubMed, Consensus lets you ask a natural language research question — "Does intermittent fasting improve insulin sensitivity in Type 2 diabetes patients?" — and it returns answers extracted directly from the full text of relevant papers, not just the abstracts. Founded in 2022 by Christian Salem and Eric Olson, Consensus has grown to index over 200 million research papers through Semantic Scholar's database, covering STEM, social sciences, and humanities.
The core innovation is the Study Snapshot: for each relevant paper, Consensus extracts and displays the key finding, population studied, sample size, methodology, and citation count — all in a structured, scannable format. Instead of spending 15 minutes reading a paper's abstract, methods, and conclusion to determine if it answers your question, a Study Snapshot gives you the answer in 15 seconds. For a typical research question with 20-30 relevant papers, Consensus reduces the literature triage phase from 2-3 hours to approximately 20 minutes.
Consensus also offers a Consensus Meter — a visual gauge showing the distribution of scientific evidence on yes/no questions. For example, asking "Does zinc supplementation shorten the duration of the common cold?" might show 68% of studies finding a significant effect, 22% finding no effect, and 10% showing mixed results. This meter is generated by analyzing the conclusions of all relevant studies and classifying them as positive, negative, or inconclusive. The GPT-4o Synthesis (Premium feature) then generates a plain-language summary of the evidence, complete with inline citations to the source papers. This synthesis reads like a mini literature review written by a knowledgeable research assistant.
Key numbers (2026): 200M+ indexed papers | 2M+ monthly active users | 6,500+ university partners including Harvard, Stanford, and Oxford | Covers all major disciplines | Study Snapshots available for ~85% of indexed papers (not all full-texts are accessible for extraction).
Consensus is purpose-built for the literature discovery and evidence synthesis phase of research. Here is what it does that traditional search tools cannot.
Type questions the way you think them, not as keyword strings. "What is the effect of blue light exposure on sleep quality in adolescents?" works better than "blue light sleep adolescents." Consensus uses semantic search to understand the intent behind your question and match it to relevant paper findings, not just keyword occurrences. It handles synonyms, related concepts, and even acronyms. Complex multi-variable questions ("Does A affect B in population C when controlling for D?") are parsed and matched across multiple paper dimensions.
Each paper result shows a structured card: the paper's answer to your question (one sentence), population studied (e.g., "2,847 adults aged 45-72"), sample size, study design (RCT, cohort, meta-analysis), and key findings extracted by AI. You can scan 20 studies in 2 minutes and identify the 3-4 most relevant ones to read in full. Snapshots are AI-generated from the full text and flagged when confidence is low.
For yes/no research questions, the Consensus Meter categorizes each study's conclusion as "Yes" (supports), "No" (refutes), or "Possibly" (mixed/inconclusive) and displays the distribution as a horizontal bar chart. This gives you an immediate visual answer to "What does the evidence actually say?" The meter is most reliable for questions with 15+ studies; for niche topics with only a handful of papers, it serves as a directional guide rather than a definitive answer.
After displaying individual Study Snapshots, Premium users can generate a 300-500 word evidence synthesis that reads like a mini literature review. It summarizes the overall state of evidence, highlights key studies, notes methodological limitations, and identifies research gaps — all with inline citations to the source papers. A professor can use this as a starting point for a lecture or an introduction section. A student can use it to understand the research landscape before diving into individual papers.
Filter results by study design: Meta-Analysis, Systematic Review, Randomized Controlled Trial (RCT), Observational Study, or Case Study. Also filter by sample size (>100, >1000, >10,000), publication date range, journal quality tier, and whether preprints are included. These filters transform a broad literature search into a targeted evidence review, allowing researchers to focus on the highest-quality evidence first.
Every Study Snapshot includes a one-click citation generator in APA, MLA, Chicago, and BibTeX formats. Bulk export selected papers to Zotero, Mendeley, EndNote, or a CSV file. The Bookmark feature (Premium) lets you save papers to named lists ("Dissertation Lit Review," "Grant Proposal Background") for organized research management. Lists can be shared with collaborators via a URL.
A practical guide for integrating Consensus into academic research and teaching workflows.
Go to consensus.app and sign up — free account available with email or Google login. The Free plan gives you unlimited searches, Study Snapshots, and basic Consensus Meter. Try a query from your field. Instead of keywords, phrase it as a question: "What is the relationship between sleep duration and academic performance in college students?" Browse the Study Snapshots to understand the format. The 20 GPT-4o Summaries included in the Free plan let you test the synthesis feature before upgrading.
Formulate your research question precisely. The more specific the question, the better Consensus performs. Compare:
❌ "exercise and depression" (too vague, returns 5,000+ results)
✅ "Does aerobic exercise reduce depressive symptoms in adults diagnosed with major depressive disorder compared to SSRIs?" (specific, returns 20-40 highly relevant results)
After the initial results load, apply methodology filters. If you are writing a systematic review, filter to "Meta-Analysis" and "Systematic Review" only. If you need primary evidence, filter to "RCT." If you want the most recent literature, set the date range to 2020-2026. These filters reduce noise and surface the most methodologically rigorous papers.
For yes/no questions, the Consensus Meter appears at the top of results. Read it carefully: a 75% "Yes" does not mean the question is settled — it means 75% of the studies in this sample found evidence supporting the hypothesis. Consider: How many studies total? (A 75% score from 4 studies is less meaningful than from 40.) What study designs are included? (If all 40 are observational studies and zero are RCTs, the evidence is suggestive but not causal.) Use the methodology filter to see how the meter changes when you restrict to RCTs only — this often reveals that "strong evidence" from observational studies becomes "insufficient evidence" when filtered to gold-standard designs.
Once you have explored individual Study Snapshots, click "Synthesize" (requires Premium or uses one of your 20 free GPT-4o Summaries). Consensus generates a structured summary: (1) Overview of the evidence landscape, (2) Key findings from the most cited studies, (3) Methodological considerations and limitations, (4) Research gaps and future directions. Each claim is cited with a clickable reference to the source Study Snapshot. Important: The synthesis is AI-generated and should be treated as a starting point, not a final product. Verify key claims against the original papers. Many professors use the synthesis as a lecture outline or a draft introduction for a paper.
Check the checkbox next to relevant papers and click "Export." Choose your citation format (APA, MLA, Chicago, or BibTeX) and export destination (Zotero, Mendeley, EndNote, or downloadable file). The exported citations include the paper title, authors, journal, year, and DOI — everything needed for a reference list. For Zotero users, the direct integration imports papers with one click, including available PDFs. Create a Bookmark list for each project and share the list URL with research collaborators so everyone works from the same literature base.
How professors and researchers use Consensus in practice.
A public health professor at a U.S. university was conducting a systematic review on "mask mandates and COVID-19 transmission in K-12 schools." Traditional database searches (PubMed, Web of Science) returned 847 papers after keyword filtering. Using Consensus with methodology filters (RCT + Observational, 2020-2025, sample size >500), the same question returned 38 highly relevant papers, of which 31 aligned with the professor's inclusion criteria — a precision rate of 82% vs. ~15% from traditional database searches. The professor reported that Consensus reduced the title/abstract screening phase from 2 weeks to 3 days because Study Snapshots made relevance determination nearly instantaneous. One caveat: Consensus should supplement, not replace, traditional database searches in systematic reviews to ensure comprehensiveness (PRISMA guidelines recommend searching multiple databases).
An undergraduate psychology professor uses Consensus weekly to prepare lectures. Before teaching a unit on "sleep and memory consolidation," she searches Consensus for the latest meta-analyses and RCTs from the past 2 years. The Consensus Meter quickly shows whether the evidence has shifted since her last syllabus update. She uses the GPT-4o Synthesis as a starting outline for her lecture slides and includes a screenshot of the Consensus Meter in her presentation with the prompt: "Notice how 82% of studies support the consolidation hypothesis — but look at the methodological quality. What happens when we filter to RCTs only?" This teaches students critical appraisal of evidence while keeping lectures current with recent research.
A neuroscience researcher used Consensus to draft the background section of an NIH R01 grant proposal on "transcranial magnetic stimulation (TMS) for treatment-resistant depression." She searched 4 related questions ("rTMS efficacy vs sham," "rTMS vs ECT," "rTMS durability at 12 months," "rTMS biomarker predictors") and exported the top 10 Study Snapshots per question. The GPT-4o Syntheses for each sub-question provided structured summaries she adapted into the grant's Background & Significance section. The Consensus Meter for each question helped her articulate the state of evidence: "While 78% of studies support rTMS efficacy, evidence for durability beyond 6 months is limited (only 3 studies, mixed results)." This honest, data-backed framing reportedly strengthened the grant narrative. The researcher emphasized that she verified all claims against primary sources before submission — Consensus accelerated discovery but did not replace verification.
| Plan | Price | What You Get |
|---|---|---|
| Free | $0/month | Unlimited searches, Study Snapshots, basic Consensus Meter, 20 GPT-4o Summaries/month, citation export. Full functionality for casual exploration and light research use. |
| Premium | $11.99/month or $99/year | Unlimited GPT-4o Summaries, Bookmarks & Lists, advanced methodology filters (by study design, sample size, journal quality, population), export to reference managers. Recommended for active researchers and professors. $99/year is the better deal — ~$8.25/month. |
| Teams | Custom pricing | Everything in Premium for an entire department or lab. Centralized billing, admin dashboard, usage analytics, SSO/SAML integration. Typically $8-10/user/month for cohorts of 20+. University-wide licenses available — contact Consensus sales. |
Pricing verified against Consensus official pricing page, June 2026. The Free plan's 20 monthly GPT-4o Summaries reset each calendar month. Premium annual plan effectively costs $8.25/month. Teams pricing is negotiated per-institution.
Honest assessment based on academic user feedback and hands-on testing.
| Feature | Consensus | Elicit | SciSpace | ResearchRabbit |
|---|---|---|---|---|
| Core Approach | AI search + synthesis | AI-powered literature review | Paper reading + chat | Citation graph exploration |
| Paper Database | 200M (Semantic Scholar) | 200M (Semantic Scholar) | 280M+ papers | 200M (Semantic Scholar) |
| Natural Language Queries | ✅ Excellent | ✅ Excellent | ✅ Good | ❌ Keyword-based |
| Evidence Synthesis | ✅ GPT-4o | ✅ GPT-4o | ✅ SciSpace Copilot | ❌ |
| Consensus Meter | ✅ Unique | ❌ | ❌ | ❌ |
| Citation Graph | Basic | Basic | ✅ Full | ✅ Best-in-class |
| Paper Chat/Copilot | ❌ | ❌ | ✅ Explain paper | ❌ |
| Free Tier | ✅ Unlimited searches | ✅ 5,000 credits/mo | ✅ Basic features | ✅ Free entirely |
| Premium Price | $11.99/mo | $12/mo | $20/mo | Free |
Best combo: Consensus for evidence discovery + SciSpace for deep paper reading (Copilot explains individual papers) + ResearchRabbit for citation graph mapping. Elicit is the closest direct competitor to Consensus with very similar feature set.
Explore more AI research tools.