AI Apps Earn More, Retain Less: The Revenue Paradox in RevenueCat's 2026 Data
In 18 months, OpenEvidence grew from 3 million to 18 million monthly clinical consultations and became the AI tool used by more American physicians than all competitors combined. A breakdown of the GTM strategy, trust-building mechanics, and vertical AI dominance playbook.
In December 2025, OpenEvidence facilitated 18 million clinical consultations. Twelve months earlier, the number was approximately 3 million. In a market where healthcare AI companies routinely take three to five years to achieve meaningful clinical adoption, OpenEvidence grew sixfold in one calendar year.
By January 2026, when the company announced a $250 million Series D at a $12 billion valuation, CEO Daniel Nadler told CNBC that OpenEvidence was used by 40% of U.S. physicians — a claim that independent analysis supports. More specifically: OpenEvidence is used by more American physicians than all other AI tools for physicians combined.
This is not a story about a healthcare AI company that raised money. There are hundreds of those. This is a story about a healthcare AI company that achieved genuine vertical dominance in a market that is notoriously resistant to technology adoption. Understanding how they did it is the most important case study in vertical AI GTM of the past two years.
The Numbers That Define the Company
Before the strategy, the data. OpenEvidence's publicly reported metrics as of early 2026:
| Metric | Value |
|---|---|
| Verified physician signups | 757,000+ |
| U.S. physician penetration | ~40% |
| Hospital and health system partners | 10,000+ |
| Monthly clinical consultations | 18 million (December 2025) |
| YoY consultation growth | ~500% |
| Annual revenue | $100M+ |
| Valuation | $12 billion |
| Funding round | $250M Series D |
The comparison to competitors is the data point that matters most strategically: OpenEvidence is used by more physicians than all other physician AI tools combined. In a market with significant competition from both AI-native startups and large incumbents like Epic, Microsoft, and Google, that concentration is unusual. When a single product achieves over 50% market share in a professional services category, it typically indicates either regulatory protection (which OpenEvidence does not have) or a product-experience gap so large that competitors cannot close it through feature parity. OpenEvidence appears to be the latter.
Why Clinical Documentation Was the Right Entry Point
The healthcare AI landscape in 2023 and 2024 was full of ambitious companies targeting the highest-value clinical applications: diagnostic imaging AI, drug discovery, surgical robotics, precision oncology. These markets are real and large. They are also brutally regulated, high-liability, and organizationally resistant to adoption. An FDA De Novo clearance takes 18 to 36 months and costs $2 to $5 million before a product can legally be used in U.S. clinical settings. An enterprise contract with a major health system involves CIO approval, CMO approval, clinical champion identification, IT security review, data governance review, EHR integration testing, and a clinical pilot. From first meeting to signed contract, the timeline is commonly 18 months.
OpenEvidence took a fundamentally different route.
Clinical documentation and literature synthesis sit in a category of AI-assisted physician tools that do not require FDA clearance under current guidance because they support physician decision-making rather than making autonomous clinical decisions. The physician remains the decision-maker; the AI provides context, synthesis, and documentation assistance. This is a genuine medical-ethical distinction, not regulatory gamesmanship — the physician's judgment is the final clinical authority, and the AI's role is to make that judgment faster and better-informed.
The practical consequence: OpenEvidence could go from product to market to adoption without waiting for regulatory approval. While competitors were in FDA review, OpenEvidence was acquiring physicians.
The underlying workflow problem is also enormous. U.S. physicians spend an average of 2.6 hours per day on documentation, more than they spend with patients in many specialties. Physician burnout is at record levels, with documentation burden routinely cited as the primary driver. Any tool that meaningfully reduces documentation time while improving clinical accuracy addresses a pain that physicians experience dozens of times every working day. That frequency of use is the habit formation prerequisite that drives retention — which is why OpenEvidence retains physicians at rates that general-purpose AI apps cannot match.
The Trust Architecture: Why It Works for Physicians
The critical design decision at OpenEvidence is that every AI response cites its sources. This is not standard practice in consumer AI — ChatGPT, Gemini, and Claude regularly provide answers without visible source attribution. In consumer contexts, users generally accept this. In clinical contexts, physicians do not.
A physician asking "What is the current evidence on anticoagulation management in atrial fibrillation patients with CKD stage 3?" cannot act on an AI answer they cannot verify. The liability exposure, the professional obligation, and the patient safety requirement all demand that claims be traceable to evidence. OpenEvidence's citation architecture turns what is a product design choice into a trust foundation: every answer comes with the journals, guidelines, and studies that support it, with recency dates and evidence quality grades where available.
This design constraint also creates a distribution moat. Building a product that synthesizes current medical literature accurately, cites sources reliably, and updates continuously as new evidence is published requires significant ongoing investment in data acquisition, curation, and quality assurance. General-purpose AI companies can theoretically replicate the feature; they cannot easily replicate the systematic investment in medical literature accuracy that makes the citations trustworthy rather than plausible-sounding.
The trust architecture extends to the product's error behavior. When OpenEvidence is uncertain, it says so. When a question falls outside its knowledge base or involves clinical scenarios with limited evidence, it flags the uncertainty explicitly rather than generating a confident-sounding but unreliable answer. This is the opposite of most AI product design, which optimizes for apparent confidence. For physicians, apparent confidence without evidentiary grounding is not a feature — it is a liability.
The Bottom-Up GTM: How Physicians Distributed the Product
The dominant model for enterprise healthcare software sales is top-down: identify the economic buyer (typically a health system CIO or CMO), navigate procurement, close an enterprise contract, then attempt physician adoption through mandatory training and institutional policy. This model works for EMR systems because they are infrastructure that requires institutional mandate. It fails for clinical tools because physicians resist software that was chosen for them rather than by them.
OpenEvidence built its distribution around the opposite dynamic. The product launched with a free tier for individual physician accounts. Any physician could sign up, verify their credentials, and start using the product within minutes — no institutional approval, no IT integration, no procurement process. The initial revenue model was not individual physician subscriptions; it was enterprise contracts with hospitals and health systems that were formalized after organic adoption had already occurred within those institutions.
This inversion of the enterprise sales sequence — adoption before procurement rather than procurement before adoption — is the most strategically significant element of OpenEvidence's GTM. By the time a health system's procurement committee was evaluating an enterprise contract, 35 to 40% of its physicians were already using the free tier. The enterprise sale became a formalization of existing behavior rather than a change management initiative. The physician adoption evidence also provided the sales team with the strongest possible enterprise argument: your physicians are already using this product and they want to keep using it.
The peer network dynamics in medicine amplify bottom-up distribution in ways that are hard to replicate in most other professional categories. Physicians share clinical tools at morning rounds, in specialty group chats, at grand rounds presentations, and at professional conferences. The research on community-led growth mechanics shows that professional communities with high trust, strong peer networks, and a shared mission create the fastest-growing distribution channels for tools that actually solve shared problems. Medicine has all three properties.
What the Regulatory Positioning Got Right
OpenEvidence's regulatory positioning deserves extended analysis because it reflects a strategic choice that most healthcare AI founders do not make explicitly.
The highest-value clinical AI applications — diagnostic radiology, pathology, cardiology monitoring — require FDA clearance. That clearance creates a regulatory moat for incumbents who have survived the process, but it also creates a multi-year delay before any company can start generating revenue. The companies that raised large Series A and B rounds in 2022 and 2023 to pursue FDA-regulated diagnostic AI are mostly still in clearance processes in 2026.
OpenEvidence chose a different part of the market: the large, painful, daily-use clinical workflow that does not require FDA clearance. Clinical documentation assistance and evidence synthesis are genuinely useful, genuinely painful to physicians, and genuinely achievable with current AI capabilities — and they sit outside the primary FDA regulatory pathway.
This is not the maximum-value application of AI in healthcare. Diagnostic AI that improves cancer detection or predicts sepsis has higher individual-patient impact. But for a company trying to achieve market scale, clinical documentation offered a path to 40% physician penetration that the diagnostic AI path does not.
The healthcare AI funding analysis documented the structural problem: the sector raised $18 billion in 2025 alone while the FDA approved only 12 AI products. The gap between investment pace and regulatory throughput creates a liquidity crisis for companies that bet on the regulated path. OpenEvidence bet on the unregulated-but-valuable path and captured the market while competitors waited.
The Six-Step Vertical AI Dominance Playbook
OpenEvidence's success is not accidental, and it is not unique to healthcare. The underlying strategic logic is a repeatable playbook for achieving vertical AI dominance in any professional services category where expert judgment is central and trust is the primary purchase criterion.
1. Identify the highest-frequency, highest-pain workflow that sits outside the primary regulatory or procurement barrier. In healthcare, this was clinical documentation and literature synthesis. In legal, the equivalent is contract review and case research. In accounting, it is tax code synthesis and audit preparation. These are not the highest-profile applications, but they are the ones that generate daily use, immediate workflow value, and peer sharing behavior.
2. Build for the expert's trust criteria, not the institution's procurement criteria. Physicians care about accuracy and citation transparency. Lawyers care about accuracy and jurisdiction specificity. Accountants care about accuracy and regulatory citation. The expert user will adopt a product that meets their professional trust standards even without institutional endorsement. The institution will formalize adoption of products that experts already trust.
3. Price the individual tier for adoption, not revenue. Free or near-free individual tiers in professional categories eliminate the financial barrier to expert adoption and allow peer networks to drive distribution. The monetization layer is enterprise contracts, which are signed after adoption evidence has been generated, not before.
4. Invest disproportionately in accuracy and reliability. In high-stakes professional domains, apparent confidence without demonstrated accuracy is a brand liability. A single high-profile AI error in a clinical, legal, or financial context can undo years of adoption. The products that win in professional vertical AI are those that invest in accuracy infrastructure — data quality, model evaluation, error case identification, and continuous improvement — as aggressively as they invest in product features.
5. Integrate into the institutional workflow as early as feasible. EHR integrations, legal practice management integrations, and accounting platform integrations embed the product into the daily workflow at the institutional level, creating retention that persists even when individual champions leave.
6. Use adoption data as enterprise sales collateral. When 40% of a health system's physicians are already using your free tier, the enterprise contract conversation is: "Would you like to manage this at the institutional level and get enterprise features?" Not: "Would you like to introduce this new technology to your clinical staff?" The first conversation is easy. The second conversation takes 18 months. The enterprise AI transformation research shows that physician and employee resistance to technology introduction is the primary reason enterprise AI deployments fail — OpenEvidence's model eliminates that resistance by making adoption voluntary and evidence-based before it is institutional.
The Competition and Why It Missed
The question that matters for incumbents and challengers is: why didn't the existing players achieve what OpenEvidence achieved?
Epic, the dominant EMR vendor with relationships in nearly every major U.S. health system, had both the distribution and the data advantages to build a dominant physician AI tool. It built AI features into its platform — ambient documentation, AI-assisted notes, predictive clinical tools — but these features are part of the Epic platform, which means they are subject to Epic's institutional procurement and rollout cycle. A physician at a hospital that has not yet deployed Epic's AI module cannot use it, regardless of how good the product is. OpenEvidence's distribution model — individual physician signups independent of institutional status — reaches physicians that EMR-integrated AI cannot.
Microsoft's DAX Copilot and related tools targeted a similar clinical documentation workflow and backed by Microsoft's enterprise relationships. The challenge for Microsoft is that its go-to-market is fundamentally top-down: enterprise contracts with health systems, followed by IT deployment, followed by physician adoption. The sequence creates exactly the friction that OpenEvidence bypassed.
The lesson for vertical AI challengers in any professional category is that incumbents' distribution advantages (enterprise relationships, integration access, procurement familiarity) are also distribution constraints. They cannot go direct to the expert user without disrupting their existing enterprise business model. A challenger that goes direct to the expert user first, and formalizes institutional relationships second, operates in a structural space that incumbents find difficult to occupy.
What Comes Next
OpenEvidence's stated 2026 priority is international expansion. The company has focused almost entirely on U.S. physicians to date; according to PYMNTS, it reached approximately 860,000 clinicians by mid-2026. The international physician market is larger than the domestic market, and the competitive position in major international markets is weaker — creating an opportunity to replicate the U.S. playbook in markets where the bottom-up adoption model is similarly viable.
The vertical AI second-mover analysis shows that second movers in vertical AI markets frequently outgrow pioneers, often because they benefit from pioneers' market education work without carrying the early-market risk. OpenEvidence is now the pioneer in physician AI. The competitive risk for its next phase is whether a well-funded second mover, operating from a position of knowing what the market validated, can apply the same playbook in specific specialties or international markets faster than OpenEvidence can defend its position.
The defense is the same thing that drove the offense: accuracy, citation quality, physician trust, and workflow depth. Products that have invested four years in medical literature data quality are hard to catch from a standing start. The moat is real. Whether it is durable enough to hold against both AI foundation model companies entering the space with superior general-purpose models and specialist challengers with deeper domain focus in particular medical specialties is the competitive question OpenEvidence's next two years will answer.
Takeaway: OpenEvidence's growth from 3 million to 18 million monthly clinical consultations in 12 months is the clearest available proof of what vertical AI dominance looks like when a company sequences its strategy correctly: find the high-frequency workflow that sits outside the primary regulatory barrier, build for expert trust rather than institutional procurement, distribute through the free individual tier, and formalize enterprise contracts after adoption evidence is established rather than before. The $12 billion valuation reflects not just current revenue but the compounding structural advantage that comes when a product achieves physician-level trust at scale. Trust, once earned in high-stakes professional domains, compounds in ways that pricing advantages and feature advantages do not. That is the real moat OpenEvidence has built — and the real lesson for every vertical AI company that is still trying to sell top-down into institutions that haven't asked for the product yet.
Frequently Asked Questions
What is OpenEvidence and how does it work for physicians?
OpenEvidence is an AI-powered clinical decision support platform that physicians use to answer clinical questions in real time — at the point of care, while seeing patients, or during documentation. The product ingests and synthesizes medical literature, clinical guidelines, drug interactions, and real-world outcomes data to provide evidence-backed answers to questions like 'What is the first-line treatment for this presentation in a patient with these comorbidities?' Unlike general-purpose AI assistants, OpenEvidence is trained on and indexed against the current medical literature and designed to cite its sources transparently — a critical trust requirement in a domain where a wrong answer can harm patients. The product operates in the clinical documentation assistance category, which means it supports physician decision-making and documentation workflows rather than making autonomous diagnostic or prescribing decisions. This positioning keeps it out of the FDA regulatory pathway that would otherwise create a multi-year barrier to market entry. As of early 2026, more than 757,000 verified physicians have signed up for the platform, and the company reports over 40% of U.S. physicians use it regularly.
How did OpenEvidence grow to reach 40% of U.S. physicians?
OpenEvidence's growth trajectory is exceptional even by AI-era standards. The company grew monthly clinical consultations from approximately 3 million per month in late 2024 to 18 million per month in December 2025 — a 6x increase in 12 months. That growth was driven by three interconnected mechanisms. First, physician peer networks: physicians are a highly connected professional community that relies heavily on collegial recommendations. When a physician finds a tool that genuinely saves time and improves accuracy in a workflow they perform dozens of times daily, they share it with peers at rounds, conferences, and in department channels. The peer recommendation flywheel in medicine is more powerful than in almost any other professional category. Second, hospital system integrations: OpenEvidence secured partnerships and EHR integrations that brought the product into clinical workflows through the institution rather than requiring individual physician signups. Third, the free tier for individual physicians created zero-friction adoption that allowed the peer flywheel to operate without financial barriers at the individual level. Revenue comes from enterprise hospital contracts and institutional licenses, not from charging individual physicians — a GTM structure that maximizes adoption speed while monetizing through the procurement channel most appropriate for healthcare enterprise sales.
Why is clinical documentation AI different from diagnostic AI for regulatory purposes?
The FDA regulates AI as a medical device when it is intended to diagnose, treat, cure, or prevent disease — a definition that applies to AI that provides diagnostic conclusions, recommends specific treatments, or interprets medical images for diagnostic purposes. Clinical documentation and decision support tools that assist physicians without making autonomous clinical decisions fall outside the primary medical device regulatory pathway under the current FDA framework, though this regulatory landscape is evolving. OpenEvidence is designed and positioned as a documentation assistance and evidence synthesis tool: it gives physicians access to synthesized medical literature and surfaces relevant evidence for the physician to apply using their own clinical judgment. The physician makes the clinical decision; OpenEvidence provides the evidentiary context. This positioning is deliberate and has allowed OpenEvidence to reach market and scale without the 18- to 36-month FDA clearance timelines that diagnostic AI companies face. The clinical documentation market is also, practically speaking, enormous: U.S. physicians spend an average of 2.6 hours per day on documentation, and reducing that burden has measurable impact on physician burnout, patient throughput, and system cost without requiring the regulatory approval needed for clinical decision-making AI.
What makes OpenEvidence's GTM strategy different from other healthcare AI companies?
Most healthcare AI companies in 2024 and 2025 attempted to sell top-down into hospital systems — approaching CIOs, CMOs, and procurement committees with enterprise contracts, multi-month pilots, and committee approvals. OpenEvidence inverted this model. It launched with a free tier for individual physicians, optimized the product for speed and accuracy to the point where it was genuinely faster and more reliable than manual literature searches, and let physician peer networks do the distribution work. By the time hospital procurement committees were evaluating enterprise contracts, OpenEvidence had already achieved organic adoption rates of 30 to 40 percent within those hospitals. The enterprise sale became a formalization of existing behavior rather than a behavior change initiative — one of the most favorable sales dynamics in enterprise software. This bottom-up penetration strategy requires accepting low revenue per physician during the growth phase, which explains why OpenEvidence needed venture capital backing at a significant scale. But it eliminates the primary obstacle that kills healthcare AI startups: the years-long gap between regulatory clearance, hospital IT integration approval, clinical champion identification, and actual physician adoption. OpenEvidence short-circuited all of those obstacles by making the individual physician experience exceptional before worrying about institutional contracts.
What is the vertical AI dominance playbook that OpenEvidence demonstrates?
OpenEvidence's growth illustrates a repeatable vertical AI dominance playbook that applies across professional services categories where expert judgment is central. The six principles are: First, find the workflow that is simultaneously the most painful and the most frequent — for physicians, clinical literature search and documentation were both. Second, build for the expert, not the institution — physicians evaluate tools by accuracy and speed, not by vendor pedigree. Third, position in the regulatory safe lane — clinical documentation does not require FDA approval, while diagnostic AI does. Fourth, use the free individual tier as a distribution channel, with institutional contracts as the monetization layer. Fifth, invest in accuracy and citation transparency above all other product features — in high-stakes professional domains, trust is the moat. Sixth, build EHR integrations early to embed into the institutional workflow before competitors can match your adoption numbers. Each of these principles is domain-specific in its implementation but domain-agnostic in its logic — the same framework applies to legal AI (Clio, Harvey), financial services AI (BloombergGPT applications), and accounting AI (various QuickBooks AI integrations). The common thread is using a genuinely superior individual user experience to establish adoption before competitors can engage institutional procurement.