Will patients hate talking to an AI receptionist? What the data actually shows

This is the question almost every clinic owner asks first. Not "can AI answer the phone?" but "will my patients hate it?"

That fear is reasonable. A phone call is often the first human moment between a patient and a clinic. If the AI feels like a wall, a trick, or a cheap replacement for care, patients notice.

But that isn't what the data shows when AI is used narrowly for admin tasks. Patients don't hate AI on the phone. What frustrates them is four specific things: getting blocked from a human when they need one, an AI that clearly doesn't understand what they're asking, an AI that pretends to be a person, and an AI that tries to make a medical call it has no business making. Strip those four out, and the survey data is more positive than most clinic owners expect. In a KLAS Research and Luma Health survey of just over 1,000 US patients, 69% said they were "very" or "somewhat" comfortable with AI handling appointment scheduling — one of the highest comfort scores measured for any AI use case in a clinic.^[1]

That comfort isn't unconditional, and it shouldn't be treated as one. A separate Sogolytics survey of 1,012 adults found scheduling was still the single most accepted AI task on offer, but at a lower 52% — and acceptance collapsed to just 16% the moment AI made a live human harder to reach.^[2] That gap between "comfortable with AI" and "comfortable with AI that blocks a person" is the entire story of this article. Patients aren't rejecting AI. They're rejecting AI done badly.

Key takeaways

Scheduling and check-in are among the most accepted AI use cases patients report, not the least.
What patients actually reject is losing access to a human, not the presence of AI itself.
The strongest resistance in the data is about diagnosis and treatment decisions, not booking a visit.
Real clinic deployments show measurable time savings without a patient backlash, when the AI stays inside a tight scope.
There's a genuine limit: some patients, and some calls, still need a human first, every time.

The practical answer

If an AI receptionist is used as a wall, patients will hate it.

If it's used as a shortcut, many patients will accept it — especially when the alternative is voicemail, waiting on hold, or calling again tomorrow. That's the difference clinics should actually care about.

Patients usually aren't asking, "was that interaction powered by AI?" They're asking simpler questions:

Did someone answer?
Did it understand why I called?
Did I get the appointment, callback, or next step I needed?
Could I reach a human if the call became sensitive?

An AI receptionist fails when it breaks those expectations. It works when it protects them.

The four things that actually turn patients off

None of the research in this piece supports "patients hate robots." What it supports is something narrower and more useful: patients react badly to four specific failures, and every one of them is a design choice, not an inherent property of voice AI.

Getting blocked from a human. This is the biggest one by a wide margin, and the Sogolytics tradeoff data above proves it directly — acceptance falls from 52% to 16% the moment AI becomes a barrier rather than a shortcut.^[2] KLAS found the same pattern in a single quote from an older respondent that says more than any statistic: "I would expect to have a human confirm everything, so I don't see any benefits."^[3] That isn't a rejection of AI. It's a request for a human backstop, which is exactly what a properly scoped deployment provides.

Not understanding the caller. A JMIR qualitative study on patient attitudes toward AI support tools found people were broadly open to AI for administrative tasks, but named the absence of human oversight and a sense of profit-driven implementation as the two barriers that soured the experience^[4] — both of which trace back to an AI that feels like it's optimising for the business, not the caller in front of it.

Pretending to be human. Nothing in the sourced data measures this directly, but it sits underneath the trust concerns in both the KLAS and JMIR findings. An AI that doesn't disclose what it is turns a neutral interaction into a deceptive one the moment the patient figures it out — and patients do figure it out.

Making a medical call. This is where the data gets sharpest, and it's covered in its own section below, because it's the one failure mode that isn't really about phone design at all — it's about scope.

What the research says patients feel about AI scheduling

Set the four failure modes aside for a moment and look at the baseline number twice, because it's easy to misread. Two separate, independently run surveys — different countries, different samples, different methodology — both name appointment scheduling as the AI use case patients accept most readily in a clinical setting. KLAS + Luma put comfort at 69%.^[1] Sogolytics, testing a broader set of AI use cases against a stricter bar, still put scheduling on top at 52%.^[2] Neither number is a coincidence, and neither is the exception in the wider literature — a JMIR Human Factors survey on people with chronic conditions found under a third were willing to use a general health chatbot in the next twelve months, but interest in voice-based tools rose specifically for the low-friction, administrative tasks scheduling belongs to.^[5] The pattern holds every time it's tested: the more a task looks like "help me get an appointment," the more comfortable patients are letting AI handle it.

Where patients draw a hard line: admin tasks vs. medical judgement

The picture changes sharply once AI moves from booking into anything resembling a clinical decision. A JAMA Network Open study surveying 13,806 patients across 74 hospitals in 43 countries found 70.2% preferred "explainable" AI — a system whose reasoning they could follow — even at the cost of some accuracy, and 72.9% wanted a physician to make the final call in any AI-assisted diagnosis. Only 4.4% supported a fully autonomous diagnostic AI.^[6]

That's the line, drawn by patients themselves, not by a vendor's marketing copy: administrative help, yes; clinical authority, no. It's also exactly where Aimée draws it. When integrated with the clinic's calendar or CRM, Aimée can check real availability, book an approved slot, and read it back for confirmation. She doesn't diagnose, doesn't recommend treatment, and doesn't offer an opinion on symptom severity — because the data above says patients would trust her less, not more, if she tried.

Can AI actually handle scheduling without wasting the caller's time?

The scepticism here is fair, because scheduling is genuinely hard work. MGMA's 2026 research on practice phone workflows found scheduling is one of the most time-consuming call types precisely because it stacks high call volume with provider-specific rules, limited openings, and voicemail backlog — and the report makes a point of noting that conversational tools don't fix a broken workflow, they expose it.^[7] An AI bolted onto a messy scheduling process will feel exactly as slow as the process underneath it.

Where the process is actually well-integrated, the outcomes look different. A scoping review of 30 peer-reviewed studies on automated patient self-scheduling found consistent gains in staff time savings, patient satisfaction, and appointment attendance.^[8] A Frontiers in Digital Health study on online appointment scheduling in private practice found a measurably lower no-show rate and better slot utilisation after adoption, and noted that phone-only scheduling caps access to office hours and creates its own bottlenecks — the manual process isn't a neutral baseline, it has its own hidden cost.^[9]

The clearest real-world proof is UAMS's deployment with Luma Health, which automated 95% of after-hours cancellation calls, recovering more than 800 staff hours a year that had previously gone into listening to voicemails and manually updating cancellations — roughly 10,000 calls a year handled without staff involvement.^[10] NHS England's own pilot at Mid and South Essex NHS Foundation Trust cut missed appointments by around 30% over six months, preventing 377 no-shows and enabling 1,910 additional patients to be seen — using AI to improve scheduling and reminder logic, not to replace a receptionist wholesale.^[11]

Does it replace the front desk, or take the pressure off?

Every deployment with public numbers behind it points the same direction: augmentation, not replacement. TriState Health's case study with Luma describes freeing the equivalent of three full-time staff from printing, scanning, and phone tag, while cutting no-shows by 40% and estimating roughly $760,000 in annual impact. Their CIO put it plainly: "Luma lets us focus on the people in front of us."^[12] Banner Health reports 70% less manual message triaging, more than 2,300 conversations handled entirely by AI in a year, and patient responses arriving roughly six hours faster on average — with a digital product manager noting, "our patients are getting quick answers and a better consumer experience."^[13]

The honest caveat: some of these deployments also mention needing to fill fewer positions going forward, so it would be dishonest to promise a clinic that AI never affects headcount planning at all. The fairer claim, and the one the evidence actually supports today, is that the strongest pattern by far is AI as an overflow and after-hours layer that removes the least valuable parts of the job — voicemail cleanup, repetitive callbacks — so remaining staff spend more time with the patients in front of them, not fewer staff spending the same time.

The dental pattern: missed-call recovery

It's worth noting a dental-specific data point here, with a clear caveat. Peerlogic — a different company from Aimée, with no affiliation between the two despite a shared product name — publishes case studies on a dental missed-call AI tool. Across 26 dental practices, they report a 40% engagement rate on missed-call follow-up, 144 booked appointments, and $47,088 in recovered revenue in a single month; a separate Florida DSO case study reports 311 appointments and over $100,000 in 90 days.^[14] These are vendor-published numbers from a company we have no relationship with, so treat them as an industry pattern, not a benchmark for any specific product — but the pattern itself lines up with everything above: the calls that go unanswered are recoverable, and dental clinics specifically leak a measurable amount of revenue through them.

Where a human-only answering service still wins

Here's the honest limitation, because a page that only says "AI works" isn't telling the whole story. An answering service staffed by real people is still the better choice for some clinics: those with a large share of elderly patients, those whose patients have said plainly they want a human and only a human, and those whose calls are frequently emotional, complex, or insurance-related in a way that needs a person to confirm details on the spot.

Aimée isn't built to close 100% of a clinic's calls, and she shouldn't be sold that way. She's not built to pretend to be a person either — every call opens with a disclosure, not a script designed to fool anyone. Where she's genuinely strong is fast booking and rescheduling, 24/7 coverage, live access to the clinic's actual schedule, and taking the missed-call and after-hours load off a front desk that's already stretched. Human handoff stays mandatory for anything outside that scope. That's not a hedge — it's the design.

What this means if you're evaluating an AI receptionist

If you're worried patients will hate this, the data says you're asking the wrong question. The real question is whether the specific deployment you're looking at avoids the four failure modes above: does it block access to a human, misunderstand callers, hide what it is, or reach into clinical judgement? If the answer to all four is no, the evidence — not just ours, independently gathered across KLAS, JAMA, NHS England, and half a dozen operational case studies — suggests patients are often willing to accept this for narrow admin tasks.

See the honest, dimension-by-dimension breakdown of AI receptionist vs. answering service, or work out what missed calls are actually costing your clinic before deciding either way.

FAQ

Do patients actually complain when a clinic switches to an AI receptionist?

Some do, and pretending otherwise would be dishonest. But the complaints cluster around the same four failure modes covered above, not around AI existing in the first place — being unable to reach a person, feeling misunderstood, or discovering they were talking to AI without being told. A clinic that discloses upfront, hands off fast, and stays inside booking and admin tasks sees far less friction than one that tries to stretch AI into everything.

Is it OK not to tell patients they're talking to AI?

No, and it's a bad idea even where it's technically legal. Every Aimée call opens with a short disclosure — "Hi, this is Aimée, the AI assistant for [Clinic Name]…" — because the trust research is consistent on this point: patients tolerate AI far better when they know what they're dealing with than when they find out later.

Will older patients refuse to use an AI phone system?

Some will prefer a human, and the KLAS research specifically flags this as a real segment, not a myth.^[3] The fix isn't to avoid AI for every clinic with an older patient base — it's to keep a fast, easy route to a human for anyone who asks, and to never make that route feel hidden or difficult to find.

What happens if a patient gets upset or demands a human?

They get one. Aimée hands off the moment a caller asks for a person directly, the moment anything sounds urgent or health-related, or the moment a question falls outside what the clinic's own knowledge base covers. None of those triggers are left to guesswork mid-call.

Does using AI for calls make a clinic look impersonal or cheap?

The evidence points the other way. Patients consistently rank "nobody answered" and "I got voicemail" as worse experiences than a disclosed AI that solved their problem quickly.^[2] An AI that books the appointment in one call reads as more competent than a phone line that goes unanswered, not less.

Is there a type of clinic where an AI receptionist just doesn't work well?

Yes, and it's worth saying plainly. Clinics with a heavily elderly patient base, clinics whose patients have explicitly asked for a human-only line, and clinics where calls are routinely emotional, complex, or insurance-heavy are better served, at least in part, by a human answering service. A well-run AI receptionist can still help with overflow and after-hours coverage in those clinics — it just shouldn't be the only option on the line.

Sources

KLAS Research + Luma Health, Patient Perspectives on AI for Healthcare (2025): klasresearch.com
Sogolytics, Healthcare AI Adoption and Trust Report (2026): sogolytics.com
KLAS Research + Luma Health (2025), as above.
JMIR, qualitative study on patient attitudes toward AI support tools (2025): jmir.org
JMIR Human Factors, survey on chronic-condition patients and health chatbots (2024).
JAMA Network Open, international patient survey on AI in healthcare (2025): jamanetwork.com
MGMA Stat, Phones Are Still a Backlog Costing Medical Practices Time (2026): mgma.com
Scoping review of automated patient self-scheduling, PMC: pmc.ncbi.nlm.nih.gov
Frontiers in Digital Health, online appointment scheduling study (2025): frontiersin.org
Luma Health, UAMS case study: lumahealth.io
NHS England, AI expansion announcement (2024): england.nhs.uk
Luma Health, TriState Health case study: lumahealth.io
Luma Health, Banner Health case study: lumahealth.io
Peerlogic, Aimee product case studies (different company, no affiliation): peerlogic.com/aimee

Will Patients Hate Talking to an AI Receptionist? What 15,000+ Patients Actually Said