Article
Most big consumer companies are treating AI agents like intruders. Amazon, Delta Air Lines and major US banks have blocked people from using personal AI agents to shop, book flights or manage transactions. On October 28, Zocdoc did the opposite. It opened the front door and invited the bots in.
What happened
Zocdoc, the online doctor-booking platform, announced that AI agents can now find vetted providers, verify insurance and book confirmed appointments directly onto providers' calendars on behalf of patients. Agents can connect several ways, including a Model Context Protocol (MCP) server, a REST API and guidance published in an llms.txt file. Google Gemini and Amazon Health AI are already connected, and Instinct, a text- and call-based agent, is the first formal agent partner.
The rules are clear on paper. Agents must register and identify themselves, get explicit patient consent before booking, and protect patient and provider data. They can't train on, resell or republish the data they access.
Why Zocdoc is betting on bots
Because the traffic is already arriving. Zocdoc says agent traffic on its platform has doubled roughly every 10 days since early September and grown nearly 30x since July. Blocking that wave would push patients toward agents that guess, invent a provider or book the wrong slot. Opening a verified path keeps Zocdoc as the trusted layer between software and real appointments.
The business model follows. Zocdoc charges providers a one-time fee when it helps a patient find and book them, and it says providers pay when agent referrals bring in new patients. Eligible agent partners may also earn revenue.
What's still unproven
Zocdoc hasn't disclosed how many agent visits became bookings or how much revenue they generated. It also hasn't explained how it will enforce the consent rules on agents it doesn't control. And growth measured from a small starting point can look dramatic, so treat the 30x figure as a signal of direction rather than scale.
The bigger shift
Every business with a booking flow or a checkout now faces the choice Zocdoc just made. Block agents and hope customers keep arriving through the front door, or build a side door agents can use reliably. The companies that block are protecting control and data. The ones that open up are betting agents will become a major source of demand, and that being the easiest place for an agent to finish a task will become a competitive advantage, the way being easy to find on Google once was.
Healthcare is an especially telling place to make this bet. Booking a doctor involves insurance rules, provider credentials and real-time availability, which are exactly the details an agent is most likely to get wrong without trustworthy data behind it.
Your next customer might not be human
Founders running anything bookable or buyable should start preparing now, before agent traffic shows up uninvited. Check your analytics and server logs for agent visits, because they may already be there. Then make the basics machine-readable: clear service descriptions, live availability, honest pricing and a simple way to complete a transaction without guesswork. Publishing an llms.txt file or a simple API costs little compared with losing a customer whose agent couldn't make sense of your site. Just as important, decide your rules before you open the door, including consent, data use and who's responsible when an agent books the wrong thing, because trust is what agents are really shopping for.
Leave a comment