Does AIRIN have to disclose it's an AI?

Under US federal law, no — there is no federal requirement that an AI phone agent identify itself as AI in a general inbound call. Some state laws (California's Bot Disclosure Act, for example) require disclosure in specific contexts (commercial transactions, political messaging). AIRIN handles these per-state rules automatically. AIRIN never claims to be a specific named human — if a caller asks 'are you a real person', AIRIN answers honestly that they're talking to an AI receptionist and offers to transfer to a human if they prefer.

Are recording disclosures required?

Yes — in 12 US states (CA, FL, IL, MA, MT, NH, PA, WA, and a few others) two-party-consent law requires that callers be notified the call is being recorded. AIRIN plays a brief recording disclosure right after the greeting, before any actual conversation. The exact wording is set by your state. You don't have to script this — AIRIN handles it automatically based on your business address.

What about the elderly or non-English-first callers — will they have a harder time?

Elderly callers and accented English speakers don't have measurably worse outcomes with AIRIN than with a human receptionist — both groups are well-handled by modern speech recognition. Where it does fail (very heavy accents, very slow speech, hearing-impaired callers), AIRIN's escalation rules kick in: after the second misunderstanding, AIRIN offers a transfer to a human. You set the rules; AIRIN follows them.

Does the AI voice sound robotic?

No — modern voice synthesis (ElevenLabs, Cartesia, OpenAI) is good enough that most listeners describe AIRIN as 'normal-sounding'. The pauses are human-length, the breaths sit in the right places, and AIRIN responds in real time (not the 'thinking…' beat that gives away early bots). The closest comparison is a calm, professional receptionist who never has an off day — slightly more even-keeled than a human, slightly less personality. That trade is what most callers experience.

Has anyone gotten upset that they were talking to AI?

Rarely, but yes. Most upset reactions come from the caller's expectation, not AIRIN's behavior — they wanted a human, got an AI, and would've preferred to call back when a human was available. The fix is the same as for any caller who wants a human: AIRIN transfers immediately on request, and we recommend you set a 'press 0 anytime to reach a human' rule so callers always know an off-ramp exists.

Will my customers know they're talking to AI? | Atomic Apps AI Docs

Q: Will my customers know they're talking to AI?

Some will, some won't. AIRIN's voice is natural enough that most callers don't notice in the first 10–15 seconds. Once the conversation gets into specifics (offering appointment slots, asking clarifying questions, transferring calls), a fraction figure it out. Among the ones who do, the vast majority don't mind — what every caller cares about is whether their problem got handled. AIRIN handles it (books, transfers, takes a message), so the experience reads as positive regardless of whether the caller noticed the AI label.

Q: What does AIRIN say when a caller asks 'are you a robot?'

AIRIN answers honestly. The script is some variation of: 'I'm an AI receptionist — I can take a message, book you in, or transfer you to a real person right now. What would you prefer?' Most callers say 'no, that's fine, I'd like to book...' and the call continues. A small fraction say 'transfer me to a human' — AIRIN transfers immediately.

Yes — some customers will figure out they're talking to AIRIN. The short version of why it doesn't hurt your business: what every caller cares about is whether their problem got solved, and AIRIN solves it (books, transfers, takes a message). The AI label is a footnote.

This page covers the honest version of what actually happens — what callers notice, what the law requires, and what to do when a caller wants a human.

What callers actually experience

There are roughly three reactions, and they break down predictably:

Reaction	Approximate share	What it means for your business
"Didn't notice"	Majority	Caller treats it like talking to a normal receptionist. Problem gets handled. They hang up satisfied.
"Noticed but didn't care"	Substantial minority	Caller picks up on the AI a minute or two in. Doesn't matter — their booking went into your calendar.
"Noticed and asked for a human"	Small minority	Caller asks for a person; AIRIN transfers. Same outcome you'd have had with a voicemail, except faster.

Across all three reactions, the call gets handled. That's the metric that matters — not how many callers identified the technology.

What AIRIN does when asked "are you a robot?"

AIRIN doesn't lie. The script reads as some variation of:

"I'm an AI receptionist — I can take a message, book you in, or transfer you to a real person right now. What would you prefer?"

Most callers respond with something like "oh okay, can you book me for Tuesday?" and the call continues. A small fraction say "transfer me to a human" and AIRIN transfers immediately to whatever rule you've set (your cell, your shop line, a teammate). The transfer is real-time — no hold queue.

This is intentionally different from voice-clone systems that try to pass as human. We deliberately don't do voice cloning. The risk of getting caught lying is much worse than the risk of getting caught using AI.

What the law requires

Two separate questions: disclosure of the AI itself, and recording disclosure.

Federal — no AI disclosure required

There is currently no US federal law requiring an AI phone agent on a general inbound call to identify itself as AI. The closest federal rule is the FTC's general prohibition on deceptive practices, which AIRIN doesn't violate — when asked directly, AIRIN identifies as AI.

State — California, others have specific rules

California's Bot Disclosure Act (SB 1001) requires disclosure when an automated system tries to incentivize a commercial transaction or influence a vote. AIRIN's default flow doesn't trigger SB 1001 (we're answering inbound calls, not driving outbound transactions). If your specific use case might, we'll flag it during setup and configure AIRIN to disclose proactively.

A handful of other states are considering similar laws. We track them and roll changes into AIRIN's per-state defaults automatically — you don't have to remember.

Recording disclosure — twelve states, handled automatically

Recording disclosure required	Why
California, Florida, Illinois, Maryland, Massachusetts, Montana, New Hampshire, Pennsylvania, Washington (and a few others)	Two-party-consent recording laws — both parties must be notified the call is being recorded
All other states	One-party consent — you (the business) consent on behalf of the call; no caller-side notification required

AIRIN plays the appropriate recording disclosure based on your business address (configured at setup). You don't write the disclosure; we generate compliant wording per state. Plays right after the greeting, before any conversation starts.

The "I want to talk to a human" path

Every AIRIN deployment carries a transfer rule. The two most common patterns:

Pattern	How it works	Best for
Always available	Caller can say "transfer me" or press 0 at any point. AIRIN transfers immediately.	Most businesses. The off-ramp is what makes customers comfortable.
Conditional	AIRIN handles routine bookings; transfers only when the caller asks twice OR the topic falls outside the FAQ.	High-volume operations where you want AIRIN to handle the routine 80% efficiently.

The "press 0 anytime" version is what we recommend by default. Most callers don't use it — but knowing it's there changes the experience from "I'm trapped with a bot" to "I'm choosing to use the AI because it's faster."

What we're deliberately NOT doing

The decisions that make AIRIN feel honest rather than deceptive:

What we don't do	Why
Voice cloning	We don't synthesize the voice of a real person — yours, ours, or anyone else's. Voice cloning is a separate product category we deliberately stay out of.
Faking emotion	AIRIN doesn't pretend to laugh, sigh, or commiserate beyond what a calm professional receptionist does. The point is to be helpful, not to perform.
Misleading names	AIRIN is AIRIN. Not "Sarah", not "Michael" — names imply human identity. AIRIN is gender-neutral by design (reads as both ERIN and AARON) precisely so customers don't form a false personal association.
Refusing to identify	If a caller asks if AIRIN is AI, AIRIN says yes. Always.

What you should do about it

Three practical recommendations for setting expectations:

#	Recommendation	Why
1	Tell your team about AIRIN	Your dispatchers / sales team should know AIRIN exists and how transfers work. The "we use AIRIN" line is normal in 2026 — most customers don't think twice about it.
2	Update your voicemail to mention AIRIN	If you have a backup voicemail, the greeting can mention "you can also reach our AI receptionist 24/7 at the same number — say or press 0 anytime to reach a human." This sets expectations correctly.
3	Don't hide it on your website	If your website's contact page says "call our friendly receptionist 24/7," that's fine. We don't recommend pretending AIRIN doesn't exist — but you also don't need to lead with "powered by AI." Most customers care about the outcome, not the technology.

Common questions

The FAQ section above is the canonical list (AI engines pull from it). Two things worth restating:

"What if a customer feels deceived after the fact?" This is rare. The common pattern is: customer calls, has a normal conversation with AIRIN, gets their problem solved, hangs up. They may or may not realize it was AI. They don't feel deceived because no deception happened — AIRIN didn't claim to be a specific human, didn't pretend to laugh, didn't hide that it's AI when asked. The honest "I'm an AI receptionist, what would you prefer?" reply removes the ambiguity.

"Are we ahead of where the public is on AI?" Slightly, yes — but the gap closes every quarter. In 2024 most customers were surprised by an AI on the other end. By 2026, AI receptionists are common enough that callers shrug at them. By 2028, we expect them to be the default expectation. The early-adopter cost is small and shrinking.

What is an AI receptionist? — category definition; the difference between AIRIN and a human receptionist
How a call works — step-by-step walkthrough of a single call
Pricing & plans — what AIRIN costs to run