Medical answering service guide for clinics, dentists and chiropractors

What does a medical answering service do?

A medical answering service answers your practice line when your staff cannot. That may be after hours, during lunch or when every line is busy. It follows a script your practice writes and approves.

Typical tasks include these.

• Book and change visits: It schedules, moves or cancels appointments in your calendar.
• Take messages: It records the patient's name, date of birth, callback number and reason for the call.
• Route urgent calls: It pages or calls the on-call doctor, dentist or nurse when a call meets your rules.
• Answer office questions: It gives hours, location, parking, insurance accepted and forms to bring.
• Pass refill and lab requests: It logs the request for clinical staff to review the next day.

How does this differ for dental and chiropractic offices?

A dental answering service hears a lot of after-hours toothache, swelling and broken tooth calls. Your dentist should decide in advance which of those warrant a same-night callback. A chiropractor's line is mostly booking and rescheduling, so an AI receptionist often covers most of the work.

What must a medical answering service never do?

This section matters more than any other on this page. Write these limits into the script and test them.

• No medical advice: The service must not tell a caller what is wrong, what medicine to take or whether a symptom is serious. Only a licensed clinician can do that.
• Emergencies go to 911: If a caller mentions chest pain, trouble breathing, heavy bleeding, signs of stroke, thoughts of self-harm or anything that sounds life-threatening, the service tells them to hang up and call 911 now. It does this before it takes any other details.
• No guessing: If the caller's need is unclear, the service passes the call to the on-call clinician. It never decides that a call can wait.
• No promises about care: It can say when someone will call back. It must not say a provider will prescribe or approve anything.

For an AI receptionist, these rules need hard limits in the system. The AI should have a fixed emergency script that it cannot change. It should refuse any medical question and offer a callback from staff. You should test it with dozens of tricky calls before it goes live, and keep testing after every change.

What does HIPAA require from an answering service?

HIPAA is the US law that protects patient health information. The US Department of Health and Human Services (HHS) explains who counts as a business associate. It is a person or company that does work for a healthcare provider involving protected health information. A patient's name plus the reason for their call is generally protected health information. So an answering service that takes patient messages acts as a business associate in most cases.

HHS says a provider may share that information with a business associate only after it gets written assurances, usually in a contract. That contract is the business associate agreement, or BAA. It must say how the vendor may use the information, ban other uses, and cover security and breach reporting. HHS also says business associates are directly liable for parts of HIPAA. A subcontractor that handles the data for your vendor counts as a business associate too.

In practice, check these points before you sign.

• Get a signed BAA: If a vendor will not sign one, do not send patient calls to it.
• Check AI coverage: Some vendors offer HIPAA options for live agents only. Abby Connect, for example, marks its HIPAA options as "Human only" on its pricing page. Its healthcare page says its AI receptionist is for calls that carry no patient health details.
• Ask about the AI's suppliers: An AI receptionist often uses outside voice and AI services. Those suppliers handle patient data too, so they need to be covered by the vendor's own agreements.
• Ask where recordings and transcripts are kept: Find out who can see them and how long they are kept.

Do any state laws cover AI messages to patients?

California does. Its Health and Safety Code section 1339.75 covers clinics, physician offices and group practices that use generative AI to create messages to patients about clinical information. Those messages must say they were created by AI. They must also tell the patient how to reach a human provider. For audio, the notice must be spoken at the start and the end. The rule does not apply if a licensed provider reads and reviews the message first, and it does not cover scheduling or billing. If your AI only books visits and takes messages, this rule is less likely to apply. Check with your own counsel if your AI does more than that.

How much does a medical answering service cost?

Here are starting prices from each vendor's own pricing page, as of October 2026. The last column shows what each vendor's own site says about HIPAA and a BAA.

Service	Starting price	What is included	HIPAA and BAA on the vendor's site
Ruby	$250 a month	50 minutes, 24/7 coverage	HIPAA service on request, and Ruby says it provides a BAA at no extra charge
Abby Connect	$164 a month	50 "Abby Minutes", AI or human answering	HIPAA call handling with human receptionists only, and no BAA is mentioned
AnswerConnect	$350 a month	200 minutes, 24/7 live answering	A HIPAA setting you turn on in its portal, which switches off call recording, and no BAA is mentioned
Smith.ai	$300 a month	30 calls, live agents 24/7	Its pricing and medical pages do not mention HIPAA or a BAA

Only Ruby says on its own site that it provides a BAA. For the others, ask for a signed BAA in writing before any patient calls go to them.

A patient call needs several details, such as date of birth and insurance. So minute-based plans can run out fast. Count your average call length before you choose a plan size.

For comparison, the US Bureau of Labor Statistics puts the median wage for receptionists in healthcare and social assistance at $19.00 an hour, as of May 2025. An answering service adds coverage without adding a full-time hire.

Should a practice use AI or live agents?

Live agents suit practices where many calls are from anxious patients or need judgment. Examples are pediatrics, oncology and any practice with a high share of older patients. A person can slow down, repeat things and notice distress.

AI suits practices where most calls are routine. Booking, rescheduling, directions and insurance questions make up most of the calls at many dental, chiropractic and physical therapy offices. AI can handle those at any hour and can answer many calls at once.

Many practices use both. The AI books routine visits and a live agent or the on-call clinician gets anything clinical. If you go this way, make sure your vendor's BAA covers both parts.

What other rules apply to patient calls?

Call recording is one. Some states require every person on a call to agree to a recording. Justia's survey lists California, Florida, Maryland, Massachusetts, New Hampshire, Pennsylvania and Washington, and Nevada applies the rule to phone calls. Tell every patient at the start that the call may be recorded.

Outbound calls are another. The FCC ruled in February 2024 that AI-generated voices are "artificial" under the Telephone Consumer Protection Act. If you plan to use an AI voice for reminder or recall calls, check the consent rules with your counsel first.

What mistakes do practices make when they set this up?

Most of the real risk in a medical answering service setup is not the vendor, it is how the practice configures and checks it. A few mistakes come up again and again.

• Assuming "HIPAA compliant" covers the AI channel too. Some vendors, Abby Connect among them, offer HIPAA call handling for live human agents only, not for their AI receptionist. A practice that signs up for a HIPAA plan and assumes every feature is covered can end up routing patient details through a channel the vendor never agreed to protect. Ask specifically, in writing, which parts of the service the BAA actually covers.
• Assuming the BAA covers the AI's own suppliers. An AI receptionist is often built on top of outside voice and language model providers. Those suppliers touch patient data too, and they need to be named or covered in the chain of agreements, not just the vendor you signed with.
• Leaving the emergency rule to AI judgment instead of a hard-coded script. "Use good judgment" is not a safety rule. The emergency script, what counts as an emergency and the exact words the AI says, needs to be fixed and untestable to talk around, not something the model decides fresh on every call.
• Testing only the obvious emergency phrasing. A caller rarely says "I am having a medical emergency." They say something vaguer, that it hurts to breathe, that their face feels tight, that the pain will not stop. Test the escalation rule with the vague, real-sounding version, not just the textbook one.
• Treating dental and chiropractic calls as automatically lower-risk. A toothache call can still describe facial swelling that is spreading or trouble swallowing, both signs that need the same urgent escalation as any medical emergency. The emergency rule should not quietly relax just because the practice type sounds less serious.

When does a custom AI receptionist make sense for a practice?

A ready-made HIPAA service is the right choice for most single-location practices. It is faster to start and easier to cancel.

A custom build makes sense when the AI must work inside your own systems. Examples include booking straight into your practice management software, checking a patient's insurance on file, or following booking rules that differ by provider and room. It also fits groups with many locations, where per-call fees grow fast. If you use dental software, see my guide to dental practice management software for what those systems can already do.

How does Blur Studio approach a medical AI receptionist?

I start with the safety rules, before any booking features. I write the emergency script, the list of topics the AI must refuse and the handoff steps together with the practice, and I test those first, with the vague, real-sounding version of an emergency call, not just the textbook one.

Once that is solid, I connect the AI to the practice's calendar and records system, with a signed BAA in place for every vendor in the chain, including whichever voice or AI model providers sit underneath it. The practice owns the code and the accounts, which means it controls where recordings and transcripts actually live, not a vendor.