Do clients trust AI for therapy intake and booking? What we found

Every practice owner weighing an AI receptionist eventually lands on the same worry: clients might not accept it. So we asked 50 people who are currently in therapy, or actively looking for a therapist, whether they would trust AI for therapy intake and booking. Their answers were more decisive than we expected, and they say a lot about where the front office of behavioral health is heading.

This post lays out what booking a first session actually costs today, what our respondents told us, why attitudes have shifted so quickly over the last two years, and how the technology has answered the two objections that used to stop practices from acting: task complexity and patient privacy.

What booking one first session actually costs

Walk through the mechanics of a typical first booking and the friction becomes obvious. The client emails or calls, the practice replies with questions about insurance and preferred times, the client answers, and the practice comes back with an offer that may no longer work for anyone. In our experience that loop settles only after three or four email exchanges spread across several days. Handled by phone instead, the same work takes twenty minutes or more of live conversation, and it only happens if the client calls during office hours and someone is actually free to pick up.

Picture how that plays out from the client's side. Someone finally decides on a Monday evening to start therapy, finds three practices through their insurer's directory, and sends the same inquiry to all of them. One practice replies Wednesday afternoon asking about insurance. Another never replies at all. The third answers within a minute, confirms eligibility during the same conversation, and offers a Thursday slot. That client's search is over, and the other two practices never learn they were in the running.

The friction would be survivable if clients waited politely in one queue, but a person searching for a therapist is usually contacting several practices in the same sitting. The practice that answers first and offers the fastest route to a first session usually wins the client, and everyone else remains on the list as a silent no. None of that reflects staff effort. Availability lives in the calendar, eligibility lives with the payer, and intake lives in a form system, so a human has to bridge all three between everything else the day demands. We covered how much high-intent demand leaks out of this gap in our guide to getting more clients for your practice.

We asked 50 people: would you trust AI with intake and booking?

Plenty has been written about what practices think of AI, so we wanted the other side of the desk. We ran a simple survey of 50 people, every one of them either currently in therapy or actively looking to start, and put the same question to each: would you trust an AI to complete your scheduling and intake? Three patterns came through clearly in the answers.

• Respondents between 16 and 40 gave a strong, consistent yes. For this group, an AI handling booking and paperwork registered as a convenience upgrade, roughly the way online check-in for a flight does. Several volunteered that they would actively prefer it to leaving a voicemail and waiting.
• Older respondents showed mild apprehension, but still leaned toward yes. Their hesitation was about backup rather than the AI itself. They wanted confidence that a real person is reachable if something goes wrong, and with that assurance in place the objection largely dissolved.
• Every age group drew the same line in the same place. Scheduling, insurance questions, and intake paperwork were fine to hand to an AI. The therapy itself was not, and nobody suggested it should be.

An honest caveat belongs here. Fifty respondents is a small, informal sample, and we present it as a directional signal rather than a formal study. We decided it was worth publishing because it matches what we hear from practices and their clients closely enough that the pattern seems real.

What changed: talking to AI became a daily habit

Attitudes did not shift because anyone in healthcare persuaded the public. They shifted because consumer AI moved into daily life. The person filling out your intake form at 11 pm has probably already talked to ChatGPT, Claude, or Siri several times that day, for directions, for work, for homework, or for dinner ideas. Conversing with software stopped feeling novel somewhere in the last two years, and trust followed familiarity the way it always does.

The same period reshaped expectations about speed. Patience for hold music, phone tag, and two-day email replies has fallen sharply, because most other services in a client's life now resolve questions within seconds. That matters more in behavioral health than almost anywhere else. Reaching out for therapy often takes a person weeks of working up to it, and the window of readiness is short. When the response to that hard-won message is silence, the silence does not read as a busy practice. It reads as a closed door, and many people quietly move on rather than knock twice.

Practices are moving for their own reason: every missed call is a drop-off

The pull is not only coming from clients. Healthcare organizations are actively shopping for AI that answers calls in parallel, because a front desk with two hands can hold exactly one phone. When two prospective clients call at the same time, one of them reaches voicemail, and a voicemail during an active search for a therapist usually means that caller books with someone else. An AI receptionist answers every caller simultaneously, at 1 pm and at 1 am, so the practice never has to trade one new client for another.

Capability caught up over the same stretch. Two years ago, language models could hold a pleasant conversation but stumbled on multi-step work. Today they reliably run the full sequence a first booking requires: verify insurance eligibility, read real calendar availability, offer concrete times, confirm the appointment, and walk the client through intake questions before the first session. That sequence is precisely what our Front Office Agent and Intake Agent run for behavioral health practices every day, inside rules the practice defines.

Trust in the outcome, though, depends on the agent getting the hard parts right, and that reliability is not a given for every product on the market. An agent that fumbles a coverage question or misroutes a distressed caller erodes the very trust these respondents extended. The difference between an agent that resolves those moments and one that constantly hands them back to a human comes down to how specifically it was built for behavioral health, which we cover in our comparison of why a behavioral health AI receptionist beats a generic voice bot and in our guide to how accurate an AI front desk really is.

The PHI question deserves a real answer

The strongest objection we hear from owners concerns privacy, and it deserves more than reassurance. Intake conversations contain protected health information, and no practice should hand PHI to a system that quietly stores it. The infrastructure answer has matured considerably. ElevenLabs, whose voice technology powers many healthcare agents, now offers a HIPAA-compliant configuration with a zero retention mode: PHI in a conversation is redacted before anything is stored, raw audio is never uploaded, language models that fail the compliance bar are blocked from use, and the whole arrangement sits under a signed business associate agreement.

Vendor controls are half of the answer, and scope is the other half. If you are evaluating any AI for your front office, four questions separate serious platforms from risky ones.

1. Will the vendor sign a business associate agreement that covers conversation content?
2. Is there a zero retention or PHI redaction mode, and is it actually enabled for your account?
3. Where does call audio go, and is anything raw retained after the conversation ends?
4. What happens when a conversation turns clinical, and how quickly does a human take over?

That last question is the one we consider non-negotiable. An intake and booking agent should never perform clinical work. At Vybz, every agent operates inside guardrails the practice sets, with a clinician in the loop on anything clinical, so the AI collects paperwork and books sessions while humans keep the care.

What this means for your practice

Put the findings together and the conclusion is plain. Clients across age groups are ready to complete intake and booking with an AI, and they consistently reward whichever practice gets them to a first session fastest. Meanwhile, most practices still take days to finish a loop that software now completes in a single conversation. The gap between what clients accept and what practices offer has quietly inverted: the risk is no longer that an AI receptionist puts clients off, it is that a slow front office does.

The practical move is to adopt at the administrative boundary, where client trust is already established and the payoff arrives immediately. Let an AI answer every inquiry the moment it lands, complete eligibility and intake before the first session, and hand your team a booked, prepared client instead of a message slip. If you want to judge the experience the way your clients will, the fastest path is to talk to one of our live agent demos and time how long a booking actually takes.

Also read

• What hiring an AI co-worker for your practice operations looks like: clients trust the front of the funnel, and this covers the rest, from inbox triage and EHR tasks to a written end-of-day report.
• How Luna's AI memory and self-learning work: the memory system that keeps that trust earned, and why one practice's confidential details never leak to another.
• How to set up an AI receptionist for a therapy practice: the practical setup steps for putting these findings to work at your own front desk.
• Real-time insurance eligibility verification: a simple secret to more clients: the deeper treatment of the eligibility step this survey's fastest practice won with, answered in seconds instead of days.