Should an AI caller say it's AI? What Australian guidance says about AI call disclosure

If you're putting an AI voice agent on the phones, you'll soon have to decide whether it tells people it's AI. My view is yes, early in the call and in plain words. The official Australian guidance on AI call disclosure points the same way. The privacy regulator and the National AI Centre both say people should know when they're dealing with AI, especially when it could be mistaken for a person.

To be clear about what this article claims: the guidance below is guidance. I'm not saying a statute requires the exact words "I'm an AI". Separate telemarketing and privacy rules do apply to AI calls, though, and they already shape how a call has to open. Here's what the sources say and how to build a disclosure that sounds natural.

What the OAIC says about AI disclosure

The Office of the Australian Information Commissioner (OAIC) has published guidance on using commercially available AI products. Two points matter for voice agents.

First, the OAIC's AI guidance says that when an organisation collects personal information through a public-facing AI system such as a chatbot, it should make sure people know they are interacting with an AI system rather than a human. The guidance uses a chatbot as its example, but a voice agent that books appointments and collects names, phone numbers and details of what someone needs fits the same description.

Second, the OAIC recommends that businesses update their privacy policies and notifications with clear information about their use of AI, including clearly identifying public-facing AI tools to customers. So disclosure isn't only something said on the call. It also belongs in your written privacy material.

What the National AI Centre says

The National AI Centre's Guidance for AI adoption is written for industry. It says people should know when they are interacting with AI. Its example is a customer getting services from a chatbot, who should know it isn't a human specialist.

The same guidance recommends making disclosure of AI use standard practice, especially where AI makes or influences decisions, generates content that can meaningfully affect people, or can be mistaken for a human being. That last condition is the important one for voice. Modern synthetic voices are good. If your agent sounds like a person, it can be mistaken for a person, and that's exactly the case the guidance points to.

The telemarketing rules that already apply to AI calls

Disclosure guidance sits on top of rules that already cover how a call has to open. If your agent makes commercial calls, treat these as the baseline.

• The Do Not Call Register Act 2006 defines a voice call to include a call that involves a recorded or synthetic voice.
• The ACMA states that commercial calls made with an auto-dialler, or using pre-recorded or AI voices, are all telemarketing calls.
• Under the Telecommunications (Telemarketing and Research Calls) Industry Standard 2017, as soon as a telemarketing call starts the caller must give their given name (unless the call is made solely using a recorded or synthetic voice), the business name of their employer, the name of the person causing the call to be made, and the purpose of the call.
• The same Industry Standard says that where a call involves recorded or synthetic voices, the caller must provide a way during the call for the recipient to request the caller's contact and complaint-handling details, for example by pressing a button to talk to an operator.

There's a practical point in that list. The opening of an AI call already has to cover who is calling, for whom and why, and there has to be a route to a person or to contact details. Adding "I'm an AI assistant" to an opening you already need costs almost nothing.

Privacy obligations alongside disclosure

Whether the Privacy Act applies depends on your business. Most small businesses with an annual turnover of $3 million or less are not covered by the Privacy Act 1988, but some are regardless of turnover, including health service providers and businesses trading in personal information.

If you are covered, a few obligations connect directly to how your agent talks:

• Under APP 5, you must take reasonable steps, at or before collecting personal information or as soon as practicable afterwards, to tell the person matters such as your identity and contact details and the purposes of collection.
• The OAIC says APP 7 on direct marketing doesn't apply to communications covered by the Do Not Call Register Act or the Spam Act, but generally applies to direct marketing calls to numbers that aren't on the Register.
• When APP 7 applies, you must give people a simple way to opt out and, if asked, stop using or disclosing their information for direct marketing within a reasonable period.
• From 10 December 2026, organisations covered by the Privacy Act that use personal information in automated decision-making that could affect people's rights or interests must explain in their privacy policies the kinds of personal information used and the kinds of decisions made.

The stakes have also gone up. The Privacy and Other Legislation Amendment Bill 2024, passed in November 2024, introduced a statutory tort for serious invasions of privacy and gave the OAIC new tiers of civil penalties and the power to issue infringement notices. If you're unsure where your business sits, check with the ACMA or a privacy lawyer before you launch.

How to disclose naturally at the start of a call

A good disclosure is short, comes early and doesn't apologise. It tells the person what they're talking to, why you're calling and how to reach a human, then gets on with being useful.

A sample opening

Here's a template for a call following up a new enquiry. Swap in your own details:

"Hi, is that [first name]? This is the AI assistant calling from [business name] about the enquiry you made on our website today about [service]. I can help you book a time now. If you'd rather speak with someone on our team at any point, just say so."

In a couple of sentences, that covers:

• The business name and who is causing the call.
• The purpose of the call.
• That the caller is an AI assistant, said plainly.
• A route to a human.

If you use an assistant name, keep the AI label right next to it, for example "This is Ava, [business name]'s AI assistant". A human-sounding name with no label is how people end up mistaking the agent for a person.

What to avoid

• Burying the disclosure in a later sentence or saying it only when asked.
• Vague wording like "virtual team member" or "digital concierge" that some people won't recognise as AI.
• Long legal preambles that make people hang up before you've said why you're calling.
• Adding filler sounds or fake typing noises so the agent seems more human.
• Saying "I'm a person" or dodging the question when someone asks.

When someone asks "Am I talking to a real person?"

The agent should answer honestly every time, even if it already disclosed at the start, and offer the human option again. For example: "No, I'm an AI assistant for [business name]. I can keep helping with your booking, or I can arrange for someone from the team to call you back. Which would you prefer?" Write this into the agent's instructions and test it with odd phrasings ("are you a robot?", "is this a recording?") before you go live.

Opt-outs and complaints

Build the "stop calling me" path into the opening, not as an afterthought. The agent should accept an opt-out in whatever words the person uses, confirm it, and make sure it reaches your records so nobody calls again. If someone asks who you are or how to complain, the agent should give your contact and complaint-handling details, or put them through to someone who can.

Questions to ask before you go live

• Does the opening say who is calling, for which business, and why, before anything else?
• Does it say plainly that the caller is an AI assistant?
• Can the person reach a human or get your contact and complaint details during the call?
• Does the agent answer honestly when asked if it's a person?
• Does your privacy policy mention your use of AI and identify the voice agent?
• If you're covered by the Privacy Act, does the call give the collection information APP 5 expects?
• Are opt-outs captured and acted on, and are you only calling people whose consent is current?
• Have you checked your setup with the ACMA's guidance or a lawyer?

Disclosure is a trust decision

Saying "I'm an AI assistant" early is what the OAIC and the National AI Centre recommend, and it also helps the call go well. People who know what they're talking to stay calmer, ask clearer questions, and are less likely to feel tricked halfway through. The agents that do well on the phone are the ones that are upfront and actually useful.

That's how we built Dialessa. It's an AI voice agent that calls new enquiries within a minute, inside your calling hours, asks your qualifying questions and books the appointment into your calendar. Whatever tool you use, decide your disclosure wording before launch, test it properly, and keep it in the first few seconds of every call.