OpenEvidence — AI clinical answers for doctors, ad-supported, viral among physicians

Investment Memo: OpenEvidence Stage: Hypothetical Series A Sector: Digital Health / Vertical AI

Thesis

OpenEvidence has the potential to be a fund-returner by capturing the most valuable real estate in healthcare: the physician’s search bar at the point of care. For two decades, UpToDate has monopolized clinical decision support through expensive, top-down enterprise contracts. OpenEvidence flips this model. By offering a superior, AI-native product for free and monetizing through pharma advertising, they have unlocked viral, bottom-up adoption among doctors. If OpenEvidence becomes the default daily search engine for clinicians, it will command a massive share of the $30B+ pharmaceutical digital marketing spend, effectively becoming the "Google for Medicine" while bypassing the notoriously slow hospital procurement cycle.

Product & Wedge

The wedge is speed, accuracy, and zero friction. Physicians are severely burned out, and traditional clinical reference tools require sifting through dense, textbook-style articles. OpenEvidence acts as a medical-grade Perplexity. Doctors ask complex, patient-specific questions in natural language, and the AI synthesizes answers instantly, citing peer-reviewed literature, clinical guidelines, and FDA labels.

Crucially, the product is free for clinicians. By removing the paywall, OpenEvidence bypasses hospital IT and procurement committees, driving organic, viral growth through physician word-of-mouth. The monetization engine—pharma advertising—is seamlessly integrated, allowing drug manufacturers to reach high-intent prescribers exactly when they are researching treatments.

Market & Competition

The market for clinical decision support and physician networking is highly lucrative, but ripe for AI disruption.

• The Incumbents: UpToDate (Wolters Kluwer) is the gold standard, generating hundreds of millions in recurring revenue. However, its UI is legacy, and its search is keyword-based. Epocrates owns drug reference but lacks deep clinical reasoning capabilities.
• The Ad-Model Pioneer: Doximity proved that an ad-supported, physician-only network can be a multi-billion dollar public company. However, Doximity is a newsfeed and directory; OpenEvidence captures higher-intent, point-of-care queries.
• AI Challengers: Glass Health is building AI for clinical notes and diagnosis, but leans more toward workflow SaaS. General AI tools like ChatGPT and Perplexity are used by some doctors, but lack the strict medical grounding, trust, and compliance required for clinical use.

Traction & Business Signal

• User Growth: Viral adoption among physicians across the US. It is publicly known that tens of thousands of clinicians use the platform, driven by strong word-of-mouth.
• Validation: Backed by early partnerships and validation from top-tier institutions, including a notable collaboration with the Mayo Clinic, which provides immense credibility.
• Revenue & Retention: Unknown. The exact daily active user (DAU) to monthly active user (MAU) ratio, long-term retention cohorts, and current revenue run-rate from pharma sponsors remain unknown.
• Ad Performance: Unknown. The click-through rates and return on ad spend (ROAS) for early pharma partners are currently unknown.

Risks

To underwrite this deal, we must accept three existential risks:

1. The "One Bad Hallucination" Liability: Trust is the only currency in clinical decision support. If OpenEvidence’s LLM hallucinates a drug dosage or contraindication that leads to a high-profile adverse patient event, the resulting PR disaster and potential legal liability could destroy the company overnight. Medical AI has zero margin for error.
2. The UpToDate AI Counter-Offensive: Wolters Kluwer is not standing still. UpToDate has a massive proprietary dataset of physician-authored content and existing enterprise contracts with nearly every major health system. If UpToDate successfully launches a fast, reliable LLM interface, doctors may default back to the trusted incumbent, neutralizing OpenEvidence’s product advantage.
3. Point-of-Care Ad Blindness & Compliance: While pharma has deep pockets, advertising at the point of care is tricky. Doctors are in a rush; they may develop "banner blindness" to ads while reading clinical answers. Furthermore, pharma legal/regulatory teams are notoriously conservative. They may balk at placing brand ads adjacent to generative AI outputs that they cannot pre-approve, limiting the scalability of the revenue model.

Verdict: Invest

Decision: Invest at the Series A.

The consumerization of healthcare IT is long overdue. OpenEvidence has found a brilliant wedge by combining the UX of modern generative AI with the proven, highly lucrative business model of Doximity. Bypassing hospital enterprise sales in favor of an ad-supported, product-led growth motion is the exact right playbook to build a massive business in this space. The upside of owning the primary interface between doctors and medical knowledge justifies the clinical and competitive risks.

The One Question We Need Answered First: What is the D30 and D90 retention rate of your daily active users, and how many queries are they running per week? (If doctors are treating this as a novelty and churning after a month, the ad model collapses. If it has become a daily habit replacing UpToDate, this is a generational company.)

Investment Memo: OpenEvidence Hypothetical Series A | AI clinical search for physicians

Thesis OpenEvidence can become a fund-returner if it captures the default “first query” slot for practicing physicians. Clinical questions are high-frequency, high-stakes, and currently answered via slow, expensive, or untrustworthy tools (UpToDate subscriptions, PubMed spelunking, or general LLMs). An ad-supported, citation-grounded AI that physicians actually prefer creates a Google-like flywheel: usage → better retrieval/ranking → more usage → premium ad inventory from pharma, device, and diagnostics companies. Physician networks are tight and status-driven; viral adoption among attendings and residents can lock in a generation of users before incumbents or general-purpose models fully adapt. If it reaches even a mid-teens share of U.S. physician clinical searches, the resulting data, brand, and ad platform would be worth multiple billions.

Product & Wedge OpenEvidence is a conversational interface that returns evidence-based answers to clinical questions, pulling from peer-reviewed literature and guidelines with inline citations. The wedge is speed-plus-trust versus the status quo: faster than paging through UpToDate or PubMed, more current and less hallucinatory than vanilla ChatGPT or Claude, and purpose-built for the 30-second window between patients. It is not an ambient scribe or EHR copilot; it is the search layer. Ad support (rather than per-seat SaaS) removes the procurement friction that kills most point solutions in hospitals and lets individual doctors adopt immediately. The product therefore starts at the individual physician and can later expand into enterprise, CME, or decision-support APIs.

Market & Competition The addressable market is the time and money physicians spend on clinical information: UpToDate (Wolters Kluwer) remains the incumbent gold standard with hundreds of millions in revenue; DynaMed, BMJ Best Practice, Epocrates, VisualDx, and Isabel occupy adjacent niches. General search (Google, PubMed) and frontier models (ChatGPT, Gemini, Claude, Perplexity) are already used off-label. Newer medical-specific efforts include Glass Health, Hippocratic AI (more agentic), and various Med-PaLM-style research systems. OpenEvidence’s differentiation is the combination of medical grounding, physician-first UX, and a consumer-like (ad-funded) distribution model rather than another enterprise sales motion. The TAM is large because every practicing doctor is a potential daily user; the SAM is the subset willing to trust an AI layer over legacy tools.

Traction & Business Signal Publicly reported: rapid organic spread among U.S. physicians via word-of-mouth and social media; claims of high query volume and usage by a meaningful fraction of practicing doctors. Exact MAU, DAU, query counts, retention curves, and revenue are unknown. Funding history, valuation, and unit economics (ad CPM, fill rates, pharma advertiser concentration) are unknown. No public FDA clearance or peer-reviewed accuracy studies versus gold-standard sources have been widely cited.

Risks (the three that actually kill the deal)

1. Accuracy + liability. A single high-profile error that leads to patient harm, or a class of systematic hallucinations on rare conditions, can trigger malpractice exposure, FDA medical-device classification, or a sudden collapse in physician trust. Incumbents already have editorial processes and legal shields; an AI startup does not.
2. Incumbent + big-tech response. Wolters Kluwer can embed similar AI into UpToDate and leverage existing hospital contracts. Google, OpenAI, or Anthropic can ship medically fine-tuned models with better base capability and distribution. If OpenEvidence’s moat is only “first mover among doctors,” it evaporates.
3. Ad-model fragility and perceived bias. Pharma advertising on a clinical-answer engine invites accusations of influence, even if technically separated. If CPMs are low, fill rates poor, or doctors revolt against ads, the company is forced into a painful pivot to enterprise SaaS—the exact motion it currently avoids. Unknown advertiser concentration makes this risk unquantifiable today.

Verdict: Pass at hypothetical Series A until one question is answered Pass. The thesis is attractive and the wedge is real, but the three killer risks are existential and currently un-de-risked by public data. The single question that would change the decision: “What are the independent, prospectively measured accuracy, citation fidelity, and harm-rate metrics versus UpToDate and a frontier LLM on a representative set of real physician queries, including rare and high-stakes cases?” Without that, the product is an unproven clinical tool riding a viral wave. With strong numbers, the conversation becomes about defensibility and monetization rather than whether the core engine is safe enough to scale.