Wysor vs OpenEvidence: An Honest Comparison for Clinicians (2026)

Two products with overlapping intent

OpenEvidence has earned its place in US clinical practice. Founded in 2023 by Daniel Nadler as part of the Mayo Clinic Platform Accelerate program, it built strong content partnerships — NEJM, the JAMA Network, Cochrane, NCCN, and Elsevier's ClinicalKey AI — and grew to roughly 400,000 verified-clinician users by mid-2025. Doctors searching for a fast, well-sourced answer between patients use it, and for many trivial questions it works.

Wysor is a different product solving an overlapping problem. It is a subscription AI workspace with medical search as one tool among many: chat across multiple frontier models, knowledge bases for your own protocols, email and voice integration. Where OpenEvidence is a single-purpose tool funded by pharmaceutical advertising, Wysor is a workspace funded by paying users.

This page compares the two using only verifiable sources: peer-reviewed accuracy studies, OpenEvidence's own published policies, and independent reviews. Where OpenEvidence is genuinely better — premium journal partnerships, point-of-care speed for US clinicians — the comparison says so.

---

What independent studies have found about OpenEvidence accuracy

Several peer-reviewed papers have evaluated OpenEvidence directly. The numbers are sobering relative to its marketing.

Jolayemi & Hash (medRxiv, 2025) ran the most rigorous benchmark to date. Two evaluators submitted 100 subspecialty board questions from the MedXpertQA dataset (a standardised benchmark designed to test medical reasoning beyond what USMLE-style questions can measure) to both OpenEvidence and Deep Consult. Results:

• Standard OpenEvidence: 34% accuracy
• Deep Consult: 41% accuracy
• Best frontier LLM on the same dataset (GPT-o1, per Zuo et al.): 46%
• Repeatability: 77% for OpenEvidence, 72% for Deep Consult — meaning roughly one question in four returned a different answer when re-run

For comparison, the authors note that pathologists interpreting breast biopsies for invasive cancer reach 89–92% intra- and inter-observer agreement. OpenEvidence's same-question consistency is lower than that.

Deep Consult also takes considerably longer. Median response time was 240 seconds versus 13 seconds for standard OpenEvidence, and Deep Consult averages 33 references per answer (vs 5 for standard), which the authors note increases verification burden materially.

Low et al. compared OpenEvidence with ChatRWD and three general LLMs on 50 open-ended clinical scenarios scored by nine physicians. ChatRWD reached 58% relevant answers; OpenEvidence reached 24%; general LLMs scored 2–10%. The paper's framing is that OpenEvidence performs well when published evidence already exists — a textbook RAG limitation — but struggles when the question requires synthesising real-world data.

Hurt et al. evaluated OpenEvidence on five common primary-care problems (hypertension, hyperlipidaemia, type 2 diabetes, depression, obesity). The clinical impact was minimal: it reinforced existing plans rather than modifying them. Useful for confirmation; less useful for changing minds.

Hajj et al. compared ChatGPT-4o with OpenEvidence on 15 questions about transcatheter tricuspid valve repair and replacement. Subject-matter experts rated ChatGPT-4o the more reliable answer source.

Patel et al. raised separate concerns about OpenEvidence's transparency, including the process by which articles are curated or excluded, the timeliness of indexed content, and clinical relevance.

The Jolayemi pilot also documented a failure mode worth flagging: in both modes, OpenEvidence sometimes returned an answer that was not among the listed multiple-choice options, and neither mode ever responded "I don't know." Confidently wrong, with confident citations, is a recognised pattern.

None of this means OpenEvidence is unusable. It does mean the marketing language — "PhD-level," "medical superintelligence," "100% on USMLE" — does not match the picture peer reviewers see when they test it on harder questions.

---

The pharmaceutical advertising business model

OpenEvidence is free for verified US clinicians. The reason it is free is published openly on their own website: it is funded by pharmaceutical and medical-device advertising. They operate ads.openevidence.com for advertisers, and their published advertising policy describes that sponsored summaries from pharmaceutical manufacturers may appear alongside answers.

OpenEvidence states that advertisements and clinical content are kept separate, and that advertisers cannot influence what the system retrieves or synthesises. This is a self-policed editorial boundary, and how much weight to give it is a judgement each clinician makes for themselves.

What is not in dispute is the underlying economics. CPMs for advertising to verified prescribers can run from roughly $70 to over $1,000 per thousand impressions — orders of magnitude above general consumer media. That economic reality is what makes the product free.

Wysor uses a subscription model. Users pay; pharmaceutical companies do not. There are no sponsored summaries and no advertiser relationships to manage. For a hospital procurement officer evaluating clinical AI tools, this is often a meaningful difference in the diligence file — and for an individual physician, it removes a category of question about whose interest is being served by which answer.

---

Where OpenEvidence's US-centric content matters

For US clinicians, OpenEvidence's content alignment with American guidelines is a feature: AHA, ACC, ADA, and FDA labelling are usually exactly what you want to see. For UK clinicians, it is friction.

Independent reviews note that OpenEvidence tends to cite American Heart Association guidelines and FDA approvals where a UK clinician would expect NICE recommendations and the British National Formulary. Drugs licensed in the US but not in the UK can surface in answers. Dose recommendations sometimes reflect US labelling rather than MHRA-approved labels. Cost-effectiveness logic from US payor systems does not map onto NHS prescribing.

Wysor's medical search is structurally broader. It pulls from:

• PubMed / MEDLINE — 40M+ primary records, MeSH filters, citation-based reranking, retraction detection
• FDA drug labels + FAERS — 256K labels and 20M adverse-event reports as a structured source, not a snippet
• RxNorm — drug brand, generic, ingredient, and dose-form resolution
• International regulatory data (EMA, ATC) — for clinicians who cross borders or work with international pharma
• ICD-10 — both ICD-10-CM and international variants

OpenEvidence has deeper full-text access to a small set of premium journals (NEJM, JAMA, Cochrane, NCCN, ClinicalKey) via partnership. Wysor has broader structured regulatory and coding coverage. Different shapes; for UK clinicians, regulatory breadth usually matters more than premium-journal full text.

---

Privacy and data handling

Patient-context queries reveal protected information even when no name is attached. "75-year-old male on warfarin with new atrial fibrillation" can re-identify a patient in a small clinic. Where that query is sent, what is logged, and who else has access matters.

OpenEvidence is a US-infrastructure platform with a HIPAA-aligned posture for its primary market. Its business model is pharmaceutical advertising, which means clinical queries flow into a system that simultaneously serves sponsored content to the same user. OpenEvidence states ads and clinical content are kept separate; this is a self-policed boundary.

Wysor takes a different approach. Every model provider — Anthropic, OpenAI, Google, Mistral — has a signed DPA covering no-training guarantees and processor obligations. The DPA template and subprocessor list are public. There is no advertising layer for queries to flow into. For multi-physician procurement, having the contract on hand is usually what unblocks the purchase.

---

The workspace around the search

OpenEvidence is a single-purpose tool: one search box, one synthesis pane. It does that well. But a clinician's day is not just literature lookups.

Wysor wraps medical search inside a workspace:

• Multi-model chat — the same conversation can call medical search, then switch to GPT-5 for letter drafting, then to Claude for analysis of an uploaded patient summary, then to Gemini for a multimodal question
• Private knowledge bases — upload your hospital's antibiotic protocol, your departmental guidelines, your internal SOPs. The agent cites from your private corpus alongside published literature
• Email — Gmail and Outlook sync. Draft a referral letter referencing literature you just searched, without copy-paste
• On-device voice transcription — dictate consult notes, audio never leaves your device
• Browser extension — Chrome integration for working with hospital web tools
• Voice agent — patient booking and SMS confirmation flows for private practice

OpenEvidence will refer you back to your other tools for everything except the literature question.

---

Pricing

	Wysor Plus	OpenEvidence
Price	€17.99/month	Free for verified US clinicians (NPI-based verification)
Funding model	Subscription	Pharmaceutical and medical-device advertising
Eligibility	Open to anyone	Clinician verification required
Structured medical data	PubMed + FDA + FAERS + RxNorm + EMA + ATC + ICD-10	PubMed + NEJM + JAMA + Cochrane + NCCN + ClinicalKey partnerships
Other AI models	GPT-5, Claude, Gemini, DeepSeek	Single proprietary stack
Knowledge bases	5 included	Not available
Email management	Full Gmail + Outlook	Not available
Voice transcription	On-device	Not available
Verification required	None	NPI-based clinician verification

For verified US clinicians who only need literature lookup, OpenEvidence's free tier is hard to beat on price. For UK and other non-US clinicians who hit the NPI verification wall, anyone uncomfortable with pharmaceutical-funded clinical AI, or anyone who wants more than literature search in one place — Wysor Plus at €17.99 is the realistic comparison.

---

Who should choose which

Choose OpenEvidence if you:

• Are a US clinician with an NPI and want a free literature search tool
• Need fast, one-shot answers between patients
• Primarily want NEJM- and JAMA-backed evidence synthesis
• Are comfortable with a pharmaceutical-advertising-funded clinical platform
• Do not need workflow tools beyond literature lookup

Choose Wysor if you:

• Want a clinical AI that does not show pharmaceutical sponsored content
• Are a UK clinician (or any non-US clinician) who hits the NPI verification wall
• Want the option to use GPT-5, Claude, Gemini, and other frontier models directly — and to compare their answers
• Need structured FDA, FAERS, RxNorm, EMA, ATC, or ICD-10 data alongside PubMed
• Want the search inside a workspace that also handles email, dictation, and your private guideline corpus
• Want contractual privacy guarantees with a published DPA, not just a privacy policy
• Are uncomfortable with the accuracy and repeatability numbers in the published peer-reviewed evaluations

---

Start using Wysor today

Try Wysor free — no clinician verification required. The free tier includes medical search across all integrated databases, five AI models, and one knowledge base for your own protocols.

When you need the full toolset — multiple knowledge bases, every model, email and voice integration — Wysor Plus at €17.99/month is less than a single journal subscription, paid for entirely by users rather than advertisers. Contractual DPAs and no training on your queries are included on every plan.

---

References

The accuracy claims about OpenEvidence on this page are drawn from the following published sources:

• Jolayemi & Hash. "The accuracy and repeatability of OpenEvidence on complex medical subspecialty scenarios: a pilot study." medRxiv preprint, 2025.
• Zuo et al. "MedXpertQA: a benchmark for evaluating medical reasoning in LLMs." 2025.
• Low et al. "Answering real-world clinical questions using large language model, retrieval-augmented generation, and agentic systems." 2025.
• Hurt et al. Study on OpenEvidence and primary care decision-making.
• Hajj et al. ChatGPT-4o vs OpenEvidence in structural heart disease clinical decision support.
• Patel et al. Commentary on OpenEvidence transparency and curation.
• OpenEvidence published advertising policy: openevidence.com/policies/advertising
• Independent UK clinician review series, iatroX, 2025.