OpenEvidence — AI clinical answers for doctors, ad-supported, viral among physicians
September 14 at 04:16 · $0.132 total
Investment Memo: OpenEvidence
Stage considered: Hypothetical Series A Author: [Partner], drafted for IC discussion
---
Thesis
OpenEvidence is attempting to become the default clinical decision-support layer for physicians — the "Google + UpToDate replacement" at the point of care. If it wins, it owns the highest-value professional attention pool in the economy: doctors making prescribing and treatment decisions. Pharma spends ~$20–30B annually on US physician marketing, most of it inefficiently. A product that captures daily physician workflow and monetizes via contextual, compliant advertising could support a multi-billion-dollar revenue business on a free-to-user model — the classic consumer-internet playbook applied to the single most lucrative user demographic. That's the fund-returner case: not "AI health tool," but the Bloomberg terminal for medicine, paid for by pharma the way Google is paid for by advertisers.
Product & wedge
The product is an AI question-answering engine that gives physicians fast, citation-backed answers grounded in peer-reviewed literature (NEJM, JAMA, etc. — OpenEvidence has announced content partnerships with major publishers including the NEJM Group and JAMA Network). The wedge is speed and trust: UpToDate answers take minutes of reading; OpenEvidence answers a specific clinical question in seconds, with sources. It's free to verified clinicians (NPI-gated), which removes the adoption friction that kills enterprise health IT sales cycles. Distribution has been bottoms-up and viral — residents and attendings recommending it to each other — which is rare and precious in healthcare, where products are usually forced on doctors by administrators.
The gating of access to verified clinicians is strategically important: it creates a clean, auditable, high-value audience for advertisers and a defensible data asset (real clinical questions at the point of care).
Market & competition
- UpToDate (Wolters Kluwer): the incumbent, ~$1B+ revenue segment, deeply embedded in hospital subscriptions, now adding AI features. Slow but trusted and institutionally entrenched.
- Doximity: the proven comp for the business model — physician network monetized by pharma marketing, ~$500M revenue, highly profitable. Doximity has launched its own AI tools (Doximity GPT) and could bundle aggressively.
- Epocrates (athenahealth), Medscape (WebMD): legacy ad-supported physician tools; Medscape in particular proves pharma will pay for this audience, but both are dated products.
- General LLMs (OpenAI, Google/Med-Gemini, Anthropic): doctors already use ChatGPT informally. The frontier labs could commoditize the answer engine.
- Abridge, Nuance/Microsoft DAX: adjacent (ambient documentation), potential future convergence into a full clinical AI assistant.
Market size: US physician-directed pharma marketing alone is ~$20B+; add med-device, CME, and international, and the monetizable pool is large enough for venture-scale outcomes.
Traction & business signal (public only)
- Founded by Daniel Nadler (previously founded Kensho, sold to S&P Global for ~$550M) — a repeat founder with a real exit.
- Publicly reported to be used by a large share of US physicians, with the company claiming rapid organic growth; press reports in 2024–2025 cited usage by clinicians at a majority of US hospitals and claims of tens of thousands of new clinician registrations per month.
- Raised significant venture funding at a reported unicorn-plus valuation (Sequoia led a 2025 round; later reports put the valuation in the multi-billion range).
- Content partnerships with NEJM Group and the American Medical Association's JAMA Network — a meaningful trust and moat signal.
- Revenue: unknown. Ad revenue scale, advertiser count, CPMs, and retention/DAU figures are not publicly disclosed. Claims of usage penetration are company-reported and unaudited.
Risks (the three that kill the deal)
- The ad model poisons the trust that created the product. The entire asset is physician trust. The moment doctors perceive answers as influenced by — or even adjacent to — pharma money, virality reverses. Doximity threads this needle because it's a network, not an oracle; putting drug ads next to clinical answers is categorically more dangerous. One credible story of biased output near a sponsored placement could be existential.
- Commoditization by frontier models + incumbents. The core function — literature-grounded medical Q&A — gets better in every base model release. If OpenAI ships a verified-clinician mode, or Wolters Kluwer makes UpToDate conversational, OpenEvidence's product edge narrows to UX and content licenses. Content deals help, but publishers are non-exclusive-minded and will license to whoever pays.
- Regulatory/liability reclassification. If FDA treats point-of-care AI answers as clinical decision support requiring oversight, or a malpractice case attaches liability to an OpenEvidence answer, the free-viral model faces compliance drag that the ad revenue may not yet cover. Pharma ad compliance (fair balance, off-label adjacency) is also brutally hard next to dynamically generated content.
Verdict
Invest — but only at a valuation that hasn't already priced in Doximity-scale success, and with the answer to one question first:
> What is actual advertiser revenue and repeat-buy behavior — do pharma brands renew, and at what CPM relative to Medscape/Doximity — and does physician engagement (weekly retention, queries/user) hold up when ads are present?
Usage virality is demonstrated; the monetization-without-trust-erosion loop is not. If early ad cohorts show pharma renewals and stable clinician retention, this is a potential fund-returner and I'd stretch on price. If revenue is still pilots and press releases, the deal is a bet on Nadler and distribution alone — attractive, but at Series A economics only.
Thesis
OpenEvidence could be a fund-returner if it becomes the default information layer for clinicians: a product used multiple times per shift, trusted enough to influence treatment decisions, and monetized without charging physicians directly. The opportunity is attractive because clinical information retrieval is both high-frequency and economically valuable. Doctors routinely search guidelines, drug interactions, trial data, and differential diagnoses, yet existing tools are often slow, expensive, poorly designed, or disconnected from the moment of care.
The upside is not merely “an AI chatbot for doctors.” If OpenEvidence achieves habitual use and builds a trusted answer-and-attribution layer, it could expand into pharmaceutical education, clinical decision support, workflow software, and eventually health-system contracts. An ad-supported model could accelerate distribution: physicians face no adoption friction, while pharmaceutical companies and other healthcare advertisers have a large incentive to reach clinicians in a context related to diagnosis and treatment.
The important caveat is that virality among physicians is not the same as durable enterprise value. The company must prove that usage is frequent, clinically trusted, and monetizable without compromising credibility.
Product & wedge
OpenEvidence provides AI-generated clinical answers, reportedly grounded in medical literature and guidelines, for physicians. Its initial wedge is straightforward: ask a question in natural language and receive a concise answer with supporting sources, rather than searching across UpToDate, PubMed, guidelines, and general web results.
That wedge is well chosen. Physicians have strong incentives to reduce information-retrieval time, and a focused interface can be more useful than a general-purpose model. A free product also makes peer-to-peer distribution plausible, particularly among residents, specialists, and clinicians active in group chats or hospital networks.
The product’s defensibility, however, will not come from the underlying language model. Models are increasingly commoditized. Defensibility would need to arise from a trusted clinical corpus, superior citation and answer quality, physician-specific workflow integration, proprietary usage data, and a brand associated with reliable medical answers. The company may also benefit from a two-sided network: more physician attention attracts higher-value advertisers, and advertiser revenue funds a better free product.
Market & competition
The market is large but crowded. The incumbent most directly threatened is UpToDate, which has deep physician trust, extensive editorial content, and broad institutional distribution. DynaMed and BMJ Best Practice compete on evidence-based clinical reference. AMBOSS is strong with medical students and younger physicians, combining reference content with education and exam preparation. PubMed, specialty society guidelines, and hospital knowledge systems remain important sources.
AI-native and adjacent competitors include ChatGPT, Google’s medical search and AI products, Perplexity, Glass Health, and other clinical copilots. EHR vendors such as Epic and Oracle Health can embed answer generation directly into clinical workflow, potentially making standalone products less important. Pharmaceutical companies and medical-information providers could also build or sponsor competing tools.
OpenEvidence’s advantage is potentially distribution and simplicity: a free, focused product that doctors can try immediately. Its disadvantage is that competitors possess either stronger content brands, more workflow access, or vastly greater model and distribution resources.
Traction & business signal
Publicly, OpenEvidence has generated substantial attention and has been described as spreading rapidly among physicians. The company and media coverage have emphasized physician adoption and an advertising-supported model. That is an encouraging business signal because repeated use by clinicians would create valuable, intent-rich inventory for pharmaceutical and healthcare advertisers.
What is not publicly established, at least from reliable disclosed information, is more important: audited monthly active physicians, retention by specialty, queries per active user, percentage of answers used in care, revenue, advertising yield, gross margin, and the proportion of usage that is organic versus paid or institutionally distributed. The exact number of verified physician users is also unknown unless independently substantiated. Customer concentration, advertiser renewal rates, and whether health systems pay for the product are unknown.
The financing history and valuation should therefore be treated cautiously unless confirmed in company filings or direct diligence. A high valuation based primarily on attention would be dangerous; a valuation supported by retained clinical usage and recurring advertising revenue would be more compelling.
Risks
1. Trust and clinical liability could kill the company. A single visible hallucination, unsupported recommendation, or misleading citation can materially damage physician adoption. Clinical answers require not just fluency but calibrated uncertainty, current evidence, specialty context, and transparent sourcing. Liability, malpractice concerns, and regulatory scrutiny could force conservative product design or expensive controls.
2. Advertising may undermine the core asset: trust. Pharmaceutical advertising is valuable precisely because clinicians trust the environment. If sponsored content is perceived to bias answers, OpenEvidence could lose the credibility that makes its inventory valuable. Conversely, strict separation between ads and answers may reduce targeting and monetization. The company needs evidence that advertisers will pay without contaminating the product.
3. Distribution and moat risk are serious. ChatGPT, Google, EHR vendors, and incumbents can replicate an answer interface and distribute it through existing subscriptions or workflows. If OpenEvidence remains a standalone destination, it may be vulnerable to platform bundling. The company must establish proprietary clinical data, workflow integration, or a brand-level trust advantage before larger players converge.
Verdict
Invest at a hypothetical Series A, but only at a valuation disciplined by verified usage and monetization. The combination of a painful workflow, free viral distribution, and potentially valuable clinician advertising is sufficiently asymmetric to merit investment. The one question I would need answered first is:
What are the independently verifiable 6- and 12-month retention rates for verified physicians, segmented by specialty, and what percentage of revenue comes from repeat advertisers?
If retention is exceptional and advertiser demand is recurring, OpenEvidence may be building a category-defining clinical information network. If usage is primarily novelty-driven and revenue is unproven, the business is more likely an impressive AI feature than a venture-scale company.