Zocdoc Opens Its Scheduling System to AI Agents
Zocdoc has moved from tolerating AI traffic to actively inviting it in. The online scheduling company now allows registered AI agents to search for providers and book appointments on behalf of patients, a choice that places it on the more permissive side of a fast-moving debate over how much autonomy agentic AI should be given in healthcare. The company says agents must register with Zocdoc and obtain explicit patient consent before any appointment is booked or canceled, and it has already connected Google Gemini, Amazon Health AI, and the text- and call-based agent Instinct.
A nearly 30-fold rise in AI agent traffic since July appears to have been the commercial trigger. Rather than trying to block that demand, Zocdoc is formalizing it and turning a growing source of visits into a structured integration pathway. The company’s stance is unusual because much of the industry is still approaching agentic AI with caution, particularly when those systems can take multi-step actions with little human intervention. Oliver Kharraz, Zocdoc’s founder and CEO, called the decision a “contrarian move.”
The clinical relevance is not in diagnosis or treatment, but in access. Appointment booking is often the first bottleneck patients face, and in many settings the delay between intent and scheduled care determines whether a patient actually reaches a clinician. An AI agent that can sort through provider availability, verify insurance, and complete scheduling could reduce that friction if the workflow remains accurate and consent-driven. Zocdoc is betting that its platform can absorb that delegation without losing control of the underlying transaction.
Consent, Verification, and the Mechanics of Delegation
Zocdoc’s model depends on a sequence of guardrails. Approved agents plug into the company’s AI backend, which allows them to search its provider directory, check availability, verify insurance, and book an appointment after checking with the patient before any final decision. The company also says agents are not permitted to train on, resell, or republish patient and provider data, though Zocdoc did not specify how it intends to enforce those restrictions.
Those details matter because agentic AI changes the risk profile of digital health tools. A chatbot that answers questions can be limited to information retrieval. An agent that can act on a patient’s behalf enters a more sensitive operational space, where a misread insurance rule or a mistaken booking could create a direct access problem for both patient and provider. Zocdoc’s response is to require registration and consent, and to route agents through its own infrastructure rather than the open web. In practice, that means the company is favoring controlled integration over scraping-based automation, which should improve data accuracy and reduce dependence on brittle third-party inference.
“Our mission is to give power to the patient,” Kharraz said, framing delegation as a form of access rather than a surrender of control. The company is presenting the AI agent as an extension of patient choice, not as an autonomous substitute for it. A patient still needs to authorize the action, and the booking logic still runs through Zocdoc’s verified network of providers tied to electronic health record and scheduling integrations. The company argues that this structure improves precision while keeping the user closer to care.
There is a practical logic to that approach. Scheduling platforms are only as useful as their ability to reconcile multiple data streams: provider availability, insurance eligibility, and the patient’s preferred timing. If those elements can be assembled reliably by machine, the administrative burden falls. If they cannot, the patient inherits the error. Zocdoc is assuming the former can be made dependable enough to support scale.
A Wider Platform Strategy Built Around Interoperability
The move is part of a broader expansion strategy, and the timing is deliberate. In September, Zocdoc launched the Care Access Network, a platform designed to embed its scheduling technology into third-party apps, insurer websites, and AI tools. Earlier this month, the company acquired Sohar Health, a startup that provides automatic insurance benefits eligibility checks. That was the first acquisition in Zocdoc’s nearly 20-year history, signaling a sharper push to build the infrastructure around appointment access rather than simply operate a consumer-facing marketplace.
Taken in isolation, the AI-agent announcement might look like a feature update. In context, it is a platform play. Zocdoc is trying to become a scheduling layer that can sit inside other digital health products, and agentic AI expands the number of entry points through which patients can interact with that layer. The monetization opportunity is explicit: companies that run the agents can monetize the integration, creating an incentive for them to participate. Zocdoc is effectively asking AI developers to treat appointment booking as a valuable workflow endpoint, not just a convenience function.
The company’s embrace of agents also reflects a broader push across healthcare for connected systems that can exchange data more cleanly. Last year, CMS launched its health tech ecosystem initiative to increase the availability of digital health tools and improve interoperability. Zocdoc participates in that ecosystem and has pledged to provide personalized AI-powered support to patients. The new agent policy fits that posture, but it also raises the standard for governance. If agents are going to act inside healthcare workflows, consent cannot be a decorative feature; it becomes the central control point.
What the Market Gets Wrong About Autonomous Scheduling
The most striking element of Zocdoc’s move is not that it uses AI, but that it is treating AI as a distribution channel. The company has concluded that patients may be willing to delegate a narrow administrative task if the system can do it faster and with less friction than a manual search. That assumption is plausible because the work of booking care is repetitive, data-heavy, and often frustrating. It is also the type of task that can appear simple until it collides with real-world variation in clinician availability, insurance requirements, and patient preference.
Kharraz described Zocdoc’s existing infrastructure as already “primed” for this transition because much of it predates the AI era and operates in a deterministic way. That framing is important. The company is not asking agents to invent decisions; it is asking them to traverse a controlled scheduling system already linked to verified providers. The architecture is meant to limit improvisation, which is one reason Zocdoc sees scraping the open web as inferior to direct platform access.
Still, the company’s own safeguards leave open a critical operational question: how those agreements will be enforced in practice. Zocdoc says agents must protect patient and provider data, but it does not spell out the technical mechanisms behind that oversight. The article also makes clear that the industry remains divided, with some companies embracing agentic AI and others warning about privacy and governance risks. Zocdoc has chosen the permissive route, and the next test is whether its controls are sufficient once booking volume, data flows, and the number of connected agents begin to rise together.