The Easiest Business for AI to Book Might Not Be the Best One

The easiest senior living community to tour isn't necessarily the right place for your mother to live.

If you're looking for memory care, arranging a tour may be the straightforward part. You need to know whether the community can meet her particular needs, how staffing works, what care costs as those needs change, and what happens if she requires more support. A calendar full of available tour times answers none of that.

Now give the assignment to a personal AI agent: find three appropriate communities and arrange visits for next week. The agent has two jobs wrapped into one request. It has to identify the right places, then find a way to contact or schedule them. And as these systems become better at the second job, I wonder how much it will influence the first.

That's an important question for any business that depends on being chosen for its experience, qualifications, or fit rather than simply being available. It isn't hypothetical that agents can now take actions on behalf of customers. What's still largely untested is whether the easiest path to completing an action influences which businesses they put forward.

The business an agent knows about and the business it can book

In August, Google announced new connected apps for Gemini. Among them: Angi and Thumbtack for finding home professionals, and Zocdoc for booking medical appointments. The announcement describes different capabilities for different industries, which is worth paying attention to. Finding a contractor is not the same thing as hiring one.

Zocdoc said its Gemini integration could surface real-time appointments from a network of more than 200,000 providers, and that providers already on Zocdoc would be enabled through that connection without additional setup. Then, in September, Zocdoc expanded that same infrastructure across a broader network that includes Amazon Health AI, Gemini, Yelp, and other services.

Think about what that means from the perspective of a medical practice. One scheduling connection can now put its available appointments in front of people using several different products. That's a meaningful distribution advantage. It also means those particular appointment-booking experiences depend on which providers and inventory the connected service can access.

But there's a question hiding inside the convenience. Suppose a patient needs a specialist with experience treating a particular condition and an appointment within two weeks. One practice is an excellent clinical match but doesn't participate in the scheduling network. Another accepts bookings through the network, though its relevant experience is harder to establish. Which practice should the agent recommend?

I don't know how often that situation occurs, or how Gemini resolves it. The example is illustrative, not an observed result. Being included in a booking integration is not proof that an AI platform favors you in its general recommendations. But if an agent is asked to complete the appointment, the inventory it can access plainly matters to what it can finish through that route.

The issue isn't whether integrations are useful. They are. The issue is whether customers can tell the difference between these are the best matches I found and these are the matches I can book for you here.

There is more than one way for an agent to reach a business

It's tempting to picture a future in which a handful of marketplaces control every AI-mediated transaction. That would be a neat business story, but the actual technology is messier.

In May, Google described expanded agentic booking in Search, including direct booking links and AI-powered calls to businesses in categories such as home repair, beauty, and pet care. Google's business-calling capability can gather pricing and availability without requiring every shop to join a particular marketplace.

OpenAI has also developed agents that can navigate ordinary websites, rather than relying exclusively on a formal partner integration. And ChatGPT's restaurant reservation experience can surface inventory through providers such as OpenTable, Resy, and Yelp.

These approaches don't offer the same reliability or reach. Structured appointment inventory makes a certain kind of booking straightforward. A website might require several steps, a login, or human approval. A phone call can reach a business whose availability isn't published anywhere, but somebody has to answer and provide accurate information. Sometimes an agent can arrange a next step; sometimes it can only present options for a person to pursue.

That matters because a company doesn't have to be absent from the web to become difficult for an agent to engage. It may have clear, detailed information about what it does, yet no dependable way for the agent to check whether it has an opening or accepts a particular type of job.

The opposite is possible too: a business may be extraordinarily easy to schedule while providing relatively little information about the experience or capabilities that would make it a good choice.

A booked appointment can be the beginning of the decision

This difference becomes much more consequential outside relatively simple transactions.

For a restaurant, reserving a table is often the desired result. If the place meets your preferences and has availability, the task is largely complete. For a roofing contractor, scheduling an estimate doesn't establish that the company has worked on a similar roof, has the right crew, or will submit an appropriate proposal. A consultation at a law firm isn't an agreement to represent a client. A dental appointment isn't evidence that a practice is the best choice for a complicated procedure. A senior living tour certainly isn't a care decision.

Even Zocdoc's own analysis of patient choice suggests this is not a one-click decision: patients viewed an average of 21 provider profiles before booking on its platform, and respondents to a commissioned consumer survey ranked personal connection above convenience. Those figures don't tell us how an AI agent will behave, but they show how much can precede an appointment.

Agents will need to understand these distinctions, and so will anyone measuring their commercial impact. A reservation, a qualified inquiry, a consultation, and a completed service purchase are four very different outcomes. Calling all of them an agent-driven conversion creates the impression of more progress than the underlying activity warrants.

There is also a consumer-interest issue. If an assistant produces a list of three providers, does it tell the customer whether it assessed the broader market, only checked a partner's inventory, or used a mix of public research, websites, and phone calls? Were any otherwise suitable businesses excluded because it couldn't complete the requested action?

These aren't reasons to oppose agentic booking. They are reasons to expect more transparency from it. A good assistant should be able to say, in ordinary language, when it has found a strong candidate but can't schedule that business directly.

That may sound like a small product detail. For a consumer facing an expensive renovation, a legal dispute, or a decision about a parent's care, it could be the difference between a convenient shortlist and an informed one.

What service businesses should pay attention to now

For owners, I wouldn't start by trying to sign up for every marketplace or chasing another round of AI optimization tips. The sensible first step is to see what an agent can actually accomplish with your business today.

Can it determine whether you handle the customer's specific need? Can it establish the relevant credentials, service area, and practical limitations? Can it reach you through a functioning website, booking process, or phone number? If it tries to take the next step, does the interaction produce a useful response or a dead end?

Those questions apply whether you're operating a plumbing company, an accounting firm, a dental practice, or a group of senior living communities. They also reveal two different problems worth solving: whether a business's capabilities are understandable, and whether a customer can act on that understanding.

At TrueSignal, we spend a great deal of time thinking about the first problem: making a business's actual capabilities and operating history easier to examine. As agents become more capable, the second problem deserves just as much scrutiny. Neither one substitutes for the other.

There is a research question here that I think the AI visibility industry should take seriously. Take real consumer-facing agents, give them specific buyer requirements across several industries, and record the businesses they identify, the ones they consider suitable, the ones they can contact, and the ones they can actually schedule. Then check whether any qualified options were lost between those steps, and why.

To my knowledge, we don't yet have sufficiently broad, independent evidence to say that booking access routinely distorts AI recommendations. Most of the public research still measures appearances in answers, rather than what happens when an assistant attempts to finish the job. We should be careful not to turn a reasonable concern into an invented finding.

But the question is becoming urgent because the infrastructure is being built now. Booking networks are expanding their reach into AI products. Assistants are learning to work through websites and call businesses. And consumers are starting to hand over tasks that used to require their own judgment and follow-through.

The risk isn't simply that an agent fails to find a business. It's that it successfully completes a task using a narrower set of options than the customer realizes.

When an AI agent says it found the right business, we should be able to ask whether it found the best fit—or just the easiest next step.