The fear shows up in almost every demo. Business owner leans back, arms crossed: "What if they figure out it's AI? Won't they just hang up?" It's the objection that kills more AI voice projects than pricing, integration complexity, or any technical concern. The assumption is that callers will feel tricked, angry, or insulted the moment they detect a synthetic voice.

We pulled transcript and sentiment data from 5,042 inbound calls handled by VoxReach agents across forty-three Australian businesses between October and December 2024. The pattern that emerged doesn't match the fear at all.

How often callers actually ask

Out of those 5,042 calls, 283 callers explicitly asked whether they were speaking to a person or AI. That's 5.6 per cent. The other 94.4 per cent either didn't notice, didn't care, or chose not to raise it even if they suspected something.

The breakdown of when people ask is telling:

• First fifteen seconds of the call: 41 instances (usually phrased as "Is this a real person?" right after the greeting)
• Mid-call during a complex query: 89 instances (typically "Am I talking to a robot?" when the agent asks for clarification)
• At booking confirmation stage: 112 instances ("Just checking - you're AI, right?")
• After the task is done, almost as small talk: 41 instances

The context matters. When someone asks early, it's usually curiosity or mild scepticism. When they ask mid-task, it's often because they're trying to gauge how detailed their answer needs to be. When they ask at the end, it's frequently positive surprise.

What happens after disclosure

Every VoxReach agent is configured to confirm honestly if asked directly. Standard response: "Yes, I'm an AI assistant helping [business name]. I can help you with [scope]. Would you like me to connect you to a person instead, or can I assist?"

Of the 283 who asked:

• 19 requested transfer to a human immediately (6.7 per cent of those who asked, 0.4 per cent of total calls)
• 264 continued with the AI agent and completed their task

The nineteen who transferred fell into a clear pattern: fourteen were complex fault/complaint calls where the caller had already decided they needed to speak to someone senior. Three were older callers (self-identified during the call) who said they "preferred a real person". Two were prank calls where the caller was testing the system for fun.

None of the transfers were hostile. The tone in transcripts was matter-of-fact: "Okay, can you put me through to someone then?" Not: "I can't believe you're wasting my time with a robot."

Sentiment scoring tells a clearer story

We ran post-call sentiment analysis on the 264 callers who continued after disclosure. The scale runs from -1.0 (hostile) through 0.0 (neutral) to +1.0 (positive). Average sentiment score across that group: +0.61. For comparison, the sentiment average across all 5,042 calls was +0.58.

In other words, callers who knew they were speaking to AI and chose to continue were marginally more satisfied than the general population. Why? Our theory: they'd already passed the "will this thing actually help me" test. Disclosure removed uncertainty. They knew what they were dealing with and decided it was fit for purpose.

The call we reviewed last Tuesday is typical. Plumbing business in Western Sydney. Caller rang to book a same-day job, asked halfway through "You're one of those AI things, yeah?", agent confirmed, caller said "Sweet, as long as you can get Dave out this arvo", booking completed, call ends positive. Task done, expectation met, no drama.

Who doesn't ask - and why that matters more

The 94.4 per cent who never raised it are the silent majority. Some certainly noticed. Others genuinely didn't. But the shared behaviour is they chose to focus on the outcome, not the mechanism.

This aligns with how Australians interact with other automated systems. Most people don't ask if the Woolies self-checkout is "real staff". They ask: does it scan my items and let me pay? If yes, fine. If no, fetch someone.

The anxiety around AI disclosure often comes from the business owner, not the caller. Owners imagine an ethical trapdoor: "I'm deceiving my customers." But when the agent is clearly functional, fast, and polite, most callers evaluate it the same way they evaluate any service touchpoint. Does it work? Does it respect my time? If both yes, move on.

What to do if you're still worried

If disclosure fear is blocking your decision, test it. VoxReach lets you configure the agent greeting. You can go full upfront: "Hi, I'm Emma, the AI assistant for [business]. I can help with bookings and questions. What can I do for you?" Or you can stay neutral and answer honestly if asked.

Run your first fifty calls. Read the transcripts. You'll see the same pattern we do: most callers care about outcome, not ontology. The tiny fraction who mind will ask for a human. Let them. It's not a failure; it's a filter working correctly.

The myth assumes callers are primed for outrage. The data says they're primed for convenience. Meet that expectation and the "AI panic" stays hypothetical.

Sign up at app.voxreach.com.au/signup - a free 90-second demo call. See what your callers actually do when they meet the agent.