What if the user is not the payer?
Most AI product founders assume a simple commercial rule: the person who uses the product pays for the product.
OpenEvidence grew in the opposite direction. It makes the product free for physicians, while media reports say it reached more than $100 million in annualized revenue within a year and completed a $250 million financing round in January 2026 at a $12 billion valuation.
This is not just a “ChatGPT for doctors” story. More accurately, OpenEvidence turns medical knowledge retrieval into a professional entry point: free for doctors, anchored by authoritative medical publications and societies, and monetized through advertising and professional distribution budgets.
For AI founders, the most useful question is not limited to healthcare. The broader question is this: when the core user has no procurement budget, faces slow purchasing processes, or is not the right person to charge directly, can an AI product still commercialize?
OpenEvidence’s answer is yes, but only if the product understands which side of the value chain actually has budget.
It solves point-of-care knowledge anxiety, not generic search
Doctors do not lack information.
The real problem is that there is too much information, it changes quickly, quality varies widely, and many questions happen in real clinical settings where there is no time to slowly search journals, read guidelines, and compare evidence.
That is where OpenEvidence positions itself. It serves verified medical professionals in the United States with AI medical search and clinical knowledge answers. Its public website shows content agreements or collaborations with organizations such as NEJM, JAMA, NCCN, and Cochrane, and emphasizes that verified U.S. healthcare professionals can use it for free.
The key productization point is not “AI can answer medical questions.”
The real productization is compressing work that physicians previously spread across journals, guidelines, search engines, and colleague discussions into one trusted entry point.
First, the product verifies the user’s professional identity, defining the use case as professional medicine rather than general health advice.
Second, it connects to authoritative medical content, reducing source-trust problems.
Third, it answers real clinical questions rather than producing generic health explanations.
Fourth, the free strategy lowers adoption friction on the physician side.
Medicine is an extremely high-trust field. In such a field, model capability is only the ticket to entry. Content rights, evidence sources, identity systems, and usage boundaries become part of the product.
That is what many general AI products underestimate when entering professional industries. Users do not only ask whether the model is smart. They ask whether they can use it inside their own responsibility and liability environment.
Free is not a subsidy if it creates the right distribution surface
OpenEvidence’s most counterintuitive move is making the physician side free.
From a traditional SaaS perspective, a professional tool for doctors seems like an obvious subscription product. Healthcare is a high-value industry. But OpenEvidence did not put revenue pressure directly on physicians first. It made the doctor-facing product free, low-friction, and suitable for frequent use.
New York Post reported in January 2026 that OpenEvidence had surpassed $100 million in annualized revenue, mainly from advertising. The report also said more than 40 percent of U.S. physicians used the product and 95 percent of new users came from referrals by other doctors. These are media-reported, company-disclosed numbers, not independently audited figures, so they should be treated as strong signals rather than final facts.
Even viewed conservatively, the structure is instructive.
Physicians may not be the best payer, but they are the best entry point.
In the medical knowledge distribution chain, the side with clear budgets may include pharmaceutical companies, medical-device companies, medical education, professional content, and healthcare marketing organizations. Physician attention, query intent, and professional context are scarce assets in that network.
So OpenEvidence’s commercial logic is not simply: “We save doctors time, so doctors pay us monthly.”
It is closer to: “We aggregate real, professional, high-frequency medical knowledge demand from physicians, then build a compliant distribution network around that entry point.”
That is the difference between free as a subsidy and free as strategy. If free only produces signups with no clear value-capture path, it is burn. If free gathers a high-value user group and the other side of the value chain has budget, it becomes distribution.
The moat is content, identity, and real queries
AI companies often reduce moat discussions to model quality. OpenEvidence shows that vertical AI moats often sit around the model.
The first layer is content. Public disclosures around NEJM, JAMA, NCCN, Cochrane, and related medical organizations suggest that OpenEvidence is not merely scraping open web pages. It is trying to secure formal access to trusted medical content. These are still official disclosures rather than independent audits, but the direction matters: in high-trust industries, content rights and source credibility become product capability.
The second layer is identity. OpenEvidence is free for verified U.S. medical professionals. Verification may look like a registration step, but it actually defines the network. This is not a general health-answer community. It is a work entry point for clinicians.
The third layer is real queries. A June 2026 arXiv preprint used real clinical queries from the OpenEvidence platform as part of the Real-POCQi evaluation set. The paper said the evaluation included 620 real point-of-care questions from 30 specialties, with 149 practicing physicians participating in comparative review. A preprint is not final peer-reviewed proof, but it signals something important: OpenEvidence has accumulated enough real physician questions that the questions themselves can be used in AI medical evaluation.
That matters for product strategy. Every real query is not only an inference call. It is also a record of demand distribution. Over time, the product learns what physicians truly ask, which specialties ask what, where evidence needs to be stronger, and which scenarios repeat.
That data cannot be replicated by simply crawling the web. It comes from the real workflow.
Doctors recommend an answer entry point, not an AI trick
Many AI products spread through spectacle. Users share the output because the model seems amazing.
OpenEvidence likely spreads in a different way. Doctors are probably not telling each other, “This AI is cool.” They are more likely saying, “You can look up that question there.”
Those two forms of spread are very different.
Spectacle spreads fast and fades easily. Workflow spread is slower, but when it takes hold, it becomes closer to infrastructure. In professional communities such as physicians, lawyers, investment analysts, and engineers, durable word of mouth usually comes from repeated peer confirmation that a tool is reliable.
The reported figure that 95 percent of new users come from physician referrals, if it remains true, suggests OpenEvidence is not only enjoying consumer-style virality. It is benefiting from trust transfer inside a professional group.
That is why it should not be treated only as medical search. Search solves information location. OpenEvidence is competing to become the default knowledge entry point for a professional industry. If that entry point forms, it can extend into specialty modules, guideline updates, medical education, drug information, clinical pathways, and more professional content distribution.
Four lessons for AI founders
First, do not rush to charge the end user. If your product serves a high-value user group with difficult procurement, direct subscription may not be the best first model. Ask whether the user is the best payer, and whether another side of the value chain will pay for reach, distribution, insight, or outcomes.
Second, vertical AI must turn trusted sources into product capability. In medicine, law, finance, and engineering, the answer is not enough. Users need source visibility, boundaries, responsibility, and traceability. The product that builds these into the experience has a better chance of moving from demo to production.
Third, a free strategy must earn high-value assets. Free registration alone is not valuable. OpenEvidence’s free physician access can earn real queries, professional word of mouth, usage frequency, and industry entry-point status. Those assets can improve the product, distribution, and commercial model.
Fourth, a vertical entry point can be easier to commercialize than a general assistant. A general assistant can do many things, but it is often hard to explain why one professional group must use it. A vertical entry point does one thing, but it can build a clearer loop around identity, content, context, and budget.
The risks are obvious
OpenEvidence’s model is not light. Medical advertising is sensitive, and the responsibility boundary around clinical answers must be clear. There is also a long-term trust challenge between free physician access and advertising monetization. If commercial incentives influence answer ranking or presentation in a way that users perceive as biased, professional trust can disappear quickly.
Large model companies such as OpenAI and Anthropic are also moving into healthcare. OpenEvidence’s ability to keep its advantage depends on whether its content rights, physician network, real query data, and product experience become deep enough. Being “more medical” today is not enough.
But the risk is exactly why the case is useful. It proves that AI commercialization is not limited to subscription plus seat pricing.
In professional industries, the best business is sometimes not selling AI to the final user. It is becoming the default work entry point for that industry’s knowledge flow.
OpenEvidence may not really be selling a doctor Q&A tool. It may be selling a new distribution position for medical knowledge in the AI era.