Blog

Should an AI receptionist tell callers it is an AI?

By Anand Gohel · October 11, 2026

Yes. An AI receptionist should say it is an AI in its first sentence, before the caller has told it anything.

Modern voice agents sound close enough to a person that a practice can choose not to mention it. Some vendors treat that as a feature: give the agent a friendly name, a warm voice, and let callers assume. On a medical practice's phone line we think that is a mistake, and this is why.

Callers choose what to say based on who is listening

A patient calling their doctor's office is about to talk about their health. How much they say, and how they phrase it, depends on who they think they are talking to. Someone who believes they have reached a nurse may describe symptoms and expect a judgement. Someone who knows they have reached an AI assistant gives a name, a reason for the call, and asks to be put through. The second call is the one an AI receptionist is built for.

Telling the caller up front is how you get the second call. It also keeps the agent out of trouble: a caller who knows it is an AI is far less likely to ask it for medical advice it should not give.

Being found out costs more than being upfront

Most callers work it out anyway. The agent answers a little too evenly, or handles an interruption in a way a person would not. A caller who was told at the start shrugs. A caller who was not told feels tricked, and the practice is the one that tricked them. That is a poor start to a call that may be about something frightening.

This matters more with older callers and their families. Many are already wary of the phone because of scam calls. A voice that turns out to be something other than it claimed confirms the worry.

What the greeting should say

Three things, in one or two short sentences:

• Who the caller has reached. The practice's name, so they know they dialled correctly.
• That it is an AI assistant. In those words. "Virtual assistant" and "automated system" are vaguer and callers hear them as a phone tree.
• What it can do for them. Usually: find out what they need and connect them to the right person.

For example: "Thank you for calling Maple Street Medical. I'm an AI assistant, and I can get you to the right person. What are you calling about?"

A name for the agent is fine. A name instead of the disclosure is not. "This is Jenny" tells the caller they have reached a person.

When the caller asks for a person

Some will, as soon as they hear the word AI. The agent should connect them, the first time they ask, without trying to talk them out of it. An agent that argues with "I can help you with that" three times in a row has turned an honest greeting into an obstacle.

If the caller asks "am I talking to a real person?" later in the call, the answer is the same every time: no, it is an AI assistant, and it can put them through to staff. This should be a fixed rule, not something left to the model to decide in the moment.

What about the law?

We are not lawyers and this is not legal advice. Rules on automated and AI callers differ by state and by the purpose of the call, and they are changing. Your counsel can tell you what applies to your practice. Our view is that it should not change what you do: an agent that discloses in its first sentence is on the right side of the question wherever the line ends up being drawn.

The same rule for outbound calls

Disclosure is not only for patients. When an AI agent phones an insurance company for a billing team, the representative who answers should also know what they are talking to. See how payer calling works.

The question to ask a vendor

Ask to hear the first five seconds of a real call. Then ask whether the disclosure can be switched off. If it can, ask who is able to switch it off, and why anyone would. It is one of the nine questions we suggest asking any voice AI vendor.

Common questions