Do you have to tell callers they're speaking to AI?

Consumer group Choice has released survey results that any business putting AI on its phones should read. According to AAP’s report of the survey, published on 18 October, more than 1,000 people were asked about AI in customer service. Just over half said AI chatbots had made their customer service experiences worse. Fewer than one in 10 felt AI had made products and services better. And 70 per cent said businesses should not be trusted to use AI as they thought best.

That’s the mood an AI receptionist answers into. It also sharpens a question every business should settle before switching one on: do I actually have to tell callers they’re talking to AI?

This article sets out what Australian law and official guidance say at the time of writing (October 2026), where the rules are firmer (call recording in particular), and how to disclose in a way that is honest without making every call longer. It is general information, not legal advice.

The short answer

At the time of writing (October 2026), we’re not aware of a single Australian law that says, in so many words, that a business must tell an inbound caller they are speaking to an AI agent.

That doesn’t mean you can stay quiet about it. Several existing laws and regulators’ guidance point the same way:

• Consumer law prohibits misleading or deceptive conduct, and that applies to what your AI agent says.
• The privacy regulator’s guidance says public-facing AI tools should be clearly identified as AI.
• Call recording rules, which vary between states and territories, generally make it wise to tell callers before you record.
• The federal government’s guidance for businesses adopting AI says people should know when they’re interacting with AI.

Put together, the practical answer is yes: tell callers. It costs one sentence, and it removes a category of risk that is hard to justify carrying.

What the rules say at the time of writing

There is no AI-specific disclosure law, for now

For a while, Australia looked likely to introduce mandatory “guardrails” for high-risk AI. That changed with the National AI Plan, launched on 2 December 2025. As Piper Alderman’s summary of the Plan explains, the government abandoned its intention to introduce mandatory guardrails. Instead it chose to “uplift and clarify existing technology-neutral laws”, such as the Privacy Act and the Australian Consumer Law, to set up an AI Safety Institute, and to release more guidance for businesses.

In other words, AI on your phones is governed by the same laws as everything else your business does, at least for now. Choice is pushing for more. In the same report, Choice’s senior campaigns and policy advisor Jordan Cornelius is quoted saying that without “new, stronger laws”, Australia’s consumer protection framework will not keep pace with AI-related harms. The rules could change, so treat this section as a snapshot.

Consumer law: don’t mislead

The Australian Consumer Law prohibits misleading or deceptive conduct. It doesn’t care whether the misleading statement came from a staff member, a brochure or an AI agent.

In October 2025, Treasury’s review of AI and the Australian Consumer Law concluded that the existing law is broadly fit to handle AI, because it is principles-based and technology-neutral, as reported by BABL AI. That cuts both ways. There is no special AI rule to follow, but there’s also no special AI exemption.

Lander & Rogers points to the ACCC’s view that businesses using chatbots “must ensure that the information provided to consumers is consistent with their obligations” under the ACL, and warns that “AI-generated responses that mislead may expose businesses to regulatory action”. The firm also recommends that any limitations in a chatbot’s capability are clearly disclosed to consumers.

How this applies to disclosure specifically is a question for your own lawyer. But consider the obvious scenario. A caller asks, “Am I talking to a real person?” and the agent says yes. That’s a false statement made on your behalf, and there is no upside to allowing it. Whatever else you decide, your agent should never claim to be human.

Privacy: what the OAIC expects

The Office of the Australian Information Commissioner (OAIC) published guidance on privacy and the use of commercially available AI products in October 2024. It is one of the clearest official statements on disclosure. Among its recommendations is that “any public facing AI tools (such as chatbots) are clearly identified as such to external users such as customers”, so that people know they are interacting with an AI system rather than a human.

The guidance is written for organisations covered by the Privacy Act. Not every small business is. The OAIC explains that a small business with an annual turnover of $3 million or less is generally exempt, with exceptions that include health service providers and businesses that trade in personal information. If you’re unsure whether you’re covered, that is worth settling first.

For covered businesses, APP 5 already requires reasonable steps to tell people about the collection of their personal information. The OAIC’s APP 5 guidelines say that for telephone collection, this can mean explaining the relevant matters at the start of the call, “perhaps following a template script or using an automated message”, or soon afterwards, for example by directing the caller to a notice on your website. An AI receptionist’s greeting is a natural place to do both jobs at once.

One more change is close. From 10 December 2026, privacy policies of covered businesses will need to describe certain automated decisions. According to Allens’ summary of the OAIC’s final guidance, released on 30 September 2026, this applies where a computer program makes, or does something substantially and directly related to making, a decision that could reasonably be expected to significantly affect someone’s rights or interests. Allens notes that it’s a privacy policy obligation, not a requirement to notify each individual directly. Whether an AI receptionist is caught will depend on what it actually decides: an agent that only answers questions or takes messages is a different case from one that decides who gets an appointment or a service. We’ll cover that change in detail in a separate article before it commences.

Call recording: the firmer ground

Many AI receptionists record calls or keep a transcript. Here the rules are stricter, and they are not the same across the country.

Recording a phone call can engage federal telecommunications interception law as well as state and territory surveillance or listening devices laws. Sprintlaw’s guide to business call recording notes that some jurisdictions focus heavily on whether all parties consent, while others allow a participant to record in some circumstances but restrict how the recording is used or shared afterwards. Its advice is not to build your approach around the most relaxed state’s rule, and instead to use one approach nationally: clear notice before recording starts, consent by the caller choosing to continue, and another way to get help for anyone who objects.

That last point matters for AI agents. If a caller says they don’t want to be recorded, the agent needs somewhere to send them, such as a message without recording, a call-back from a person, or an email address.

Outbound calls are a different set of rules

Everything above is about inbound calls. If you’re thinking about an AI agent that makes outbound marketing calls, the telemarketing rules apply too. The ACMA’s guidance on telemarketing sets out permitted calling hours, a requirement to say who is calling and why, and caller ID that displays a return number, along with the Do Not Call Register.

Herbert Smith Freehills’ summary of the industry standard, written when it was made in 2017, notes that a call using “a recorded or synthetic voice” doesn’t have to give a caller’s personal name, but must still name the business and the purpose of the call. Outbound AI calling raises more questions than we can cover here. Get specific advice before you start.

If you have callers in Europe

Some Australian businesses take calls from customers overseas. The EU AI Act includes a transparency rule, Article 50(1), that applies from 2 August 2026. McCann FitzGerald explains that providers of AI systems that interact directly with people, including voice assistants, must design them so users are informed they are interacting with AI, unless that’s obvious to a reasonably well-informed person. The obligation is framed around the provider of the system. If a meaningful share of your callers are in the EU, ask your provider how they handle it.

What official guidance recommends

The federal government’s Guidance for AI adoption isn’t legislation, but it is the clearest statement of what the government expects from businesses. It says plainly: “People should know when they’re interacting with AI and understand when AI decisions affect them.” It gives the example of a customer who should know that a chatbot is not a human specialist.

The same guidance says people affected by AI should be able to raise complaints and get help. On the phone, that means a working route to a person, which we covered in what happens when an AI receptionist can’t help?

Why telling callers is good for business

There are also practical reasons to disclose that have nothing to do with compliance.

Callers notice anyway. Modern AI voices are good, but most callers work it out eventually: a slightly too-even pace, a pause before an unexpected question, an answer that misses the point. A caller who discovers it mid-call, after not being told, tends to feel tricked. A caller who was told at the start just gets on with it.

It sets expectations. A caller who knows they’re speaking to an AI agent adjusts how they talk. They are more likely to give a clear answer to a clear question, and less likely to launch into a long story the agent has to untangle.

It makes the handover easier. “I’m the AI assistant, so I’ll put you through to someone who can sort that out” is a natural sentence. The same handover from an agent pretending to be a person is awkward, and the pretence usually unravels at exactly that point.

It meets people where trust is. The Choice figures above suggest many Australians are sceptical about how businesses use AI. Being upfront is one of the few things a small business can do on every call to show it isn’t one of the ones people are worried about.

How to disclose without slowing the call

Done well, disclosure takes a few seconds and doubles as a useful introduction.

Say it in the greeting

The simplest approach is to put it in the first sentence the caller hears:

“Thanks for calling Harbour Plumbing. I’m the AI assistant, and I can book a job, give you an update or take a message. How can I help?”

That tells the caller what they’re talking to and what it can do, in roughly the time a receptionist would spend saying good morning. (Harbour Plumbing is an invented example.)

Combine it with the recording notice

If you record calls, you can cover both in one breath:

“I’m the AI assistant for Harbour Plumbing. This call is recorded so we can follow up on your request. If you’d rather not be recorded, just say so and I’ll arrange for someone to call you back.”

Check the exact wording with whoever advises you on privacy and recording, and make sure recording actually starts after the notice, not before.

Answer honestly if asked

Callers will sometimes ask partway through, “Is this a real person?” The agent’s instructions should say to answer plainly, yes it’s an AI assistant, and to offer a person if the caller would prefer one. Test this. It’s one of the questions worth asking your agent in several different ways before you go live.

Giving the agent a name is fine, a backstory isn’t

Giving an AI receptionist a name is reasonable, as long as the caller is told it’s an AI. What causes trouble is building a fictional person: claiming to be “in the office today”, or describing a weekend or a family. Keep the agent’s persona friendly and honest about what it is.

Offer a way to a person

Disclosure lands better when it comes with a choice. Most callers will happily continue with the agent if they know they could ask for someone. A few won’t, and they should be able to say so early without being argued with.

Back it up in writing

Your privacy policy and website should say that you use an AI agent to answer calls, what it collects, where recordings and transcripts are stored and for how long, and how to contact a person instead. That gives the agent somewhere to point callers who want detail, and it’s the document the OAIC’s APP 5 guidance expects you to be able to direct people to. Our article on whether business data is safe with AI agents covers storage and retention.

A quick checklist

Before your AI receptionist takes real calls, check that:

• the greeting says the caller is speaking to an AI assistant
• the agent never claims to be human, including when asked directly
• if calls are recorded, the caller hears a recording notice before recording starts
• callers who object to AI or to recording have another way to get help
• the agent can transfer to a person or take a message on request
• your privacy policy and website describe the AI agent and what it collects
• you know whether the Privacy Act applies to your business
• you’ve had advice on recording rules in each state or territory you take calls from
• outbound calling, if any, has been checked against the telemarketing rules
• someone in the business reviews a sample of calls each week, including how disclosure questions are handled

The short version

At the time of writing (October 2026), Australia has no law that names AI disclosure on phone calls specifically, but consumer law, privacy guidance, recording rules and the government’s own AI guidance all point the same way. The safest and simplest approach is to tell callers at the start, never let the agent claim to be human, give notice before recording, and always offer a way to a person.

It’s also what callers want. With many Australians wary of how businesses use AI, a clear one-line introduction is a small way to start each call on the right foot.

This is general information, not legal advice. The rules described here can change, and how they apply depends on your business, your industry and where your callers are. Get advice on your own circumstances, particularly if you record calls, work in health or another regulated industry, or plan to make outbound calls.

If you’re weighing up an AI receptionist, the 3xRMD Voice page explains how we approach greetings, disclosure and handover. For a broader look at where voice agents fit, see is AI voice suitable for Australian businesses?