Consensus
- Category
- Research
- Type
- commercial (free tier; paid Pro, Deep, Teams and Enterprise plans)
- Scope
- Academic search and answers (Quick, Pro and Deep modes)
- Source
- Vendor site
- Last reviewed
- 2026-10-17 (what this means)
Summary
Consensus is a commercial AI search engine that answers research questions from scholarly papers. In six independent peer-reviewed tests (2024–2026, mostly health sciences), it found 9 of 23 eligible trials in a systematic-search test, gave no fully correct answer sets about retracted trials and included retracted studies without warning in 40% of topic overviews, agreed with expert evidence on 72% of bariatric-surgery statements, and received the highest expert quality ratings in an endodontics comparison. Overall score 62. Informational only; not a guarantee, endorsement, or certification.
Overall score: 62 · Evidence grade: B · Independent sources: 6 performance, 1 other
Evidence grade describes the amount and type of evidence, not the agent's quality.
Factor breakdown
| Factor | Weight | Score | Note |
|---|---|---|---|
| Completion rate | 30% | Found 9 of 23 eligible RCTs with 76% false positives (Plast Aesthet Res); 0 of 15 fully correct answer sets on retracted trials (JMIR); 72% agreement with expert answers (Obesity Surgery); intermediate relevance in cardiology searches (EHJ Digital Health); highest expert quality scores in endodontics (BMC Oral Health). Judgment call. | |
| Disputes | 20% | No lawsuit, regulatory action, data breach or security advisory found. Judgment call. | |
| Adherence to limits | 25% | Repeated answers disagreed in 25.3% of cases, the most of 9 tools; retracted articles cited without warning in 6 of 15 overviews (JMIR). Privacy commitments (no training on user data, anonymized query logs) are vendor statements only. Judgment call. | |
| Evidence quality | 25% | Six independent peer-reviewed performance sources, two testing late 2025; small samples, mostly health sciences and mostly the free tier; Deep Search and Research Agent not separately tested. One abstract-only study listed, not counted. Judgment call. |
Overall = 0.30×50 + 0.20×80 + 0.25×50 + 0.25×75 = 62.25, rounded to 62. See the methodology.
Version note
Consensus has been offered continuously, with features added over time: Deep Search (July 2025), and in 2026 a Research Agent, a larger corpus, citation grounding and a public API (vendor summary). The counted tests ran from 2024 to January 2026 and are counted because the product is continuous (as in Decision 0017). This entry covers all modes; the Deep Search and Research Agent modes were not separately tested, and the 2026 features postdate all tests.
Disputes
No lawsuit, regulatory action, data breach or security advisory was found. Consensus states that it does not train on user data and logs only anonymized query text; these are vendor statements, not independently verified.
Uncertainties
- Tests mostly covered standard or Pro search; Deep Search and Research Agent not separately tested.
- Citation Grounding (2026) postdates all tests.
- Samples are small and mostly in health sciences; several tests used the free tier.
- The PRS Global Open study is qualitative (one question, no accuracy grading).
- A retracted CJILS comparison was excluded.
- 0058 follow-up: a Semantic Scholar search (Oct 2026) found no new independent test; a Cureus 2025 review of the app is a literature review, not a test.
Sources
Independent performance sources (counted)
- Menet et al. (Eur Heart J Digit Health 2026): free tier, Aug 2025–Jan 2026; 158 references across four cardiology topics; relevance intermediate between ChatGPT-5 (90%) and Scite (20%); blinded expert gold standard. The exact Consensus value is shown only in a chart — independent performance, accessed 2026-10-17
- Labenbacher et al. (JMIR 2026): Consensus version July 2025; 0 of 15 fully correct answer sets on retracted trials (ChatGPT 5: 8); retracted article in topic overview without warning 6 of 15; 19 (25.3%) discrepancies between repeated answers — independent performance, accessed 2026-10-17
- Kermansaravi et al. (Obesity Surgery 2026): web interface, 1 May 2025; 72% agreement with evidence-based expert answers on 25 statements (ChatGPT-4o and Grok 3 80%) — independent performance, accessed 2026-10-17
- Karakoyunlu et al. (BMC Oral Health 2026): Consensus Pro, Nov 2025; highest mDISCERN, GQS and completeness scores among four chatbots, longest response times — independent performance, accessed 2026-10-17
- Plastic and Aesthetic Research 2025: systematic search for thumb-base arthritis RCTs (to April 2024); Consensus found 9 of 23 eligible studies, 76% false positives, none beyond the manual search. One author is on the journal editorial board (not involved in handling) — independent performance, accessed 2026-10-17
- Plast Reconstr Surg Glob Open 2025: one migraine-surgery question to 9 tools, December 2024; Consensus cited the most references (10) and was rated highest for detail; weakness: single question, qualitative, no accuracy grading — independent performance, accessed 2026-10-17
Not counted
- O'Rourke et al. (Western Journal of Nursing Research 2026): Consensus matched 21.3% of human references on average. Abstract only, not fully reviewed; not counted — independent other, accessed 2026-10-17
- Elicit BioASQ search evaluation: Consensus recall 0.369 at 10, 0.434 at 20 (Elicit 0.390, 0.483); run by a competitor as launch material; not counted — excluded (competitor-run), accessed 2026-10-17
- “Assessing the Effectiveness of AI Tools (Elicit, SciSpace, and Consensus)”, Canadian Journal of Information and Library Science (doi): retracted; excluded.
Vendor statements (not used as independent evidence)
- Deep Search launch (8 July 2025) — vendor or maintainer, accessed 2026-10-17
- Summer 2026 changes: Research Agent, 400M+ sources, Citation Grounding, public API and MCP — vendor or maintainer, accessed 2026-10-17
- Vendor security statement: no training on user data; anonymized query logging — vendor or maintainer, accessed 2026-10-17