If you want to reduce patient no-shows with AI reminder calls, start by accepting an uncomfortable fact: your practice almost certainly reminds patients already, and the reminders are not the problem. The problem is what happens when a patient hears the reminder and realises they cannot make it.
If the only options are “press 1 to confirm” or “call us back during office hours”, many of those patients simply do not show up. The slot dies empty. An AI reminder call earns its cost when it turns that moment into a cancellation with notice, or better, a new appointment booked on the same call.
This guide is for practice administrators and operations leads at multi-provider groups: the 2025–2026 data, the research, the rebooking loop we would build, a cost model, the FCC conditions and vendor questions.
What the 2025–2026 numbers say
Four data points from the Medical Group Management Association (MGMA) frame the problem.
- No-shows are rising again for a third of groups. In an August 11, 2026 MGMA Stat poll (190 applicable responses), 32% of medical groups reported higher no-show rates year to date than in 2025, 58% about the same and 10% lower. The share reporting higher rates was up 5 points from the August 2025 poll (MGMA).
- It is the top access priority. Asked in December 2025 for their main patient access focus in 2026, 27% of leaders chose no-shows, ahead of online scheduling (24%), phone access (22%) and wait times (21%) (MGMA, 236 responses).
- The baseline is back to pre-pandemic levels. MGMA DataDive puts the single-specialty aggregate no-show rate at 6.81% in 2023, against 7% in 2019 and 5% in 2021 and 2022 (MGMA, January 2025).
- Fees are the other lever. In the same January 2025 poll (622 responses), 42% of groups said they charge a no-show fee.
The detail that matters most sits in the comments of the 2026 poll. Groups whose no-show rates went up described the same tactics as everyone else: text and phone reminders, confirmation calls, automated messages and fees. MGMA's own advice in that piece is telling: let patients cancel or choose a new slot from the reminder when the visit type allows it, and track the share of missed visits rebooked within 30 days.
More reminders is not a strategy. A reminder that cannot rebook is only a notification.
What reminder research actually proves
The evidence that reminders help is real, but it is weaker and messier than vendor pages suggest. Two recent reviews are worth knowing.
- Werner et al., BMC Health Services Research, 2023. A systematic review of behavioural economic interventions against missed appointments included 61 studies, 56 of which tested reminders. Reminders reduced non-attendance across many settings, but results by channel were mixed: telephone studies were inconsistent, and three studies comparing automated and staff calls pointed in different directions. The authors suggest message content may matter as much as the channel (BMC Health Services Research).
- Al-Turbag, Mooney and Corry, Journal of Hospital Management and Health Policy, March 2026. A meta-analysis of 10 trials (8,236 participants) of hospital outpatient reminders found a pooled 11% relative improvement in attendance (RR 1.11). Telephone reminders were statistically significant (RR 1.11, 5 studies), SMS was not (RR 1.14, 95% CI 0.99–1.31). Heterogeneity was high, and none of the included studies reported cost-effectiveness (JHMHP).
Two practical conclusions. First, most of those trials started from no-show rates well above a 7% practice baseline, so do not expect the same relative gain if your rate is already low. Second, nobody has proven the cost case for you. The economics come less from the reminder itself than from what happens to the slot afterwards, which is where a conversational call differs from a text.
How to reduce patient no-shows with AI reminder calls: the rebooking loop
A conversational reminder call can do what a one-way message cannot: hear “I can't make Thursday” and offer Monday at 10 or Tuesday at 3. Here is the loop we would design.
1. Decide who gets a call, not just a text
Calls cost more and annoy more than texts. Reserve them for appointments where an empty slot hurts: new patients, long visits, procedures and patients with a history of missed visits.
2. Call with the schedule in hand
The agent needs read access to open slots for that visit type and provider before it dials. A call that says “someone will call you to reschedule” recreates the problem.
3. Confirm, cancel or move in one conversation
Write-back to the schedule should happen inside your scheduling rules, after the identity check, and only for visit types you have cleared. Anything clinical (“should I still come if my symptoms changed?”) goes to staff, never to the model.
4. Backfill the freed slot
A cancellation with notice is only valuable if someone fills it. Connect the cancellation to a waitlist or a list of patients who asked for an earlier date, and measure refills.
5. Answer the callback line
Every reminder generates return calls. If they hit a busy front desk, you have moved the no-show to the hold queue. This is where an AI receptionist for medical practices earns its place: answering the inbound side of the loop around the clock.
Which channel for which appointment
| Channel | What it can do | Best for | Watch out for |
|---|---|---|---|
| One-way text | Remind, link to portal | Routine, short visits; patients who self-serve | No rebooking unless the portal allows it |
| Keypad (IVR) reminder call | Remind, confirm or cancel by key press | Patients without smartphones | Cancels without rebooking; freed slot often wasted |
| Conversational AI reminder call | Confirm, cancel or rebook live; answer logistics questions | New patients, long visits, procedures, repeat no-shows | Needs schedule write access, identity check and strict content rules |
| Staff call | Everything, with judgement | Complex prep, sensitive visits, patients who asked for a person | Most expensive; competes with the front desk queue |
An illustrative cost model
This is a worked example, not a benchmark. Replace every input with your own figures. Assume a 10-provider group with 20 scheduled visits per provider per day over 250 days, and the 6.81% MGMA aggregate rate as a starting point.
| Input (your numbers go here) | Example |
|---|---|
| Scheduled visits per year | 10 × 20 × 250 = 50,000 |
| No-show rate | 6.81% → about 3,400 empty slots |
| Average collected revenue per visit (assumption) | $150 |
| Revenue tied up in no-shows | about $510,000 |
| Share of would-be no-shows the call converts into a cancellation with notice (assumption) | 25% → 850 slots freed early |
| Share of freed slots refilled from waitlist or rebooked (assumption) | 50% → about 425 visits recovered |
| Recovered revenue per year | about $64,000 |
Two things jump out. The headline “cost of no-shows” is large, but only a slice of it is recoverable, so be wary of any vendor ROI built on the full figure. And the biggest lever in the model is not the reminder conversion rate, it is the refill rate. A practice without a working waitlist will recover half as much from the same calls. Run the sensitivity before you sign anything: if your refill rate drops to 25%, recovered revenue falls to about $32,000.
Compare that against the full cost of the program: software or managed service, integration, staff time to review transcripts, and the calls themselves. For reference, Unicall prices its managed service from $1,000 per month, billed on ticket volume rather than per seat, with implementation quoted after a free call-flow audit; the pricing page explains the model.
Rules to build in before the first call
Automated reminder calls to patients' mobile phones fall under the Telephone Consumer Protection Act. Two FCC decisions matter. In February 2024, the FCC confirmed that AI-generated voices count as an “artificial or prerecorded voice” under the TCPA (FCC 24-17). And its 2015 order exempted certain free-to-the-patient healthcare calls to wireless numbers, including appointment and exam confirmations and reminders, subject to conditions (FCC 15-72, para. 146–147). Those conditions translate directly into system rules:
Notice the tension with the rebooking loop: a one-minute guideline and a live rescheduling conversation do not sit comfortably together. The cleaner design is a short reminder that offers to reschedule, with the patient choosing to continue, and a counsel-approved position on how that is handled. As we argued in Confidently Wrong, rules like these belong in code, not in the prompt: frequency caps, call windows and opt-out lists should be enforced by the dialler, where the model cannot talk its way past them. This section is general information, not legal advice; your counsel should confirm which rules apply to your practice and your patients' states.
The scorecard to track from week one
- No-show rate, separated from late cancellations. MGMA warns that a combined “missed visit” rate hides which problem is growing.
- Cancellation lead time. Did the calls move cancellations earlier, to a point where the slot can still be filled?
- Rebooked within 30 days. The share of cancelled and missed visits that became a completed visit.
- Slot refill rate. Freed slots filled by another patient. This is where the money is.
- Reach rate and opt-out rate. Calls answered by the patient, and how many asked you to stop.
- Callback answer rate. Inbound calls generated by reminders that were answered without a hold.
Questions to ask any vendor
For how voice AI providers differ on delivery and accountability, our comparison hub uses the same ten buyer criteria for each vendor, from public information only.
FAQ
Do AI reminder calls reduce patient no-shows?
Reminders in general do. A 2026 meta-analysis of hospital outpatient trials found telephone reminders improved attendance by about 11% in relative terms, with high variation between studies. There is little independent research yet on AI voice calls specifically, so measure your own results: no-shows, cancellation lead time and slots refilled.
Are AI reminder calls better than text messages?
Not for every appointment. Texts are cheaper and work for routine visits. A conversational call is worth its cost where an empty slot is expensive, such as new patients, long visits or procedures, because it can rebook the patient on the spot instead of only recording a cancellation.
Do AI appointment reminder calls need patient consent under the TCPA?
The FCC treats AI-generated voices as artificial voices under the TCPA. Its 2015 order exempts free appointment reminders to the wireless number a patient provided, if conditions are met: provider identified at the start, no billing or marketing content, generally one minute or less, at most one message a day and three a week, and an easy opt-out honoured immediately. Confirm your position with counsel; this is not legal advice.
What should an AI reminder say on voicemail?
Keep it minimal: the practice name, that it is about an upcoming appointment, a callback number and how to opt out. Leave out the reason for the visit or any clinical detail, since HIPAA privacy rules govern the content and someone other than the patient may hear it.
How should a practice measure a no-show reduction program?
Track the no-show rate separately from late cancellations, cancellation lead time, the share of missed or cancelled visits rebooked within 30 days, and the share of freed slots filled by another patient. Compare a few months before and after, by visit type, rather than one blended rate.
Unicall.ai builds and operates done-for-you AI voice agents. Our live deployments today are in customer care, led by e-commerce. For a clinic, we start with a free call-flow audit of your appointment calls, and the scope, integrations with practice systems and data-processing terms are agreed with your team before any contract is signed.