Do I Have to Tell My Customers I'm Using AI?

In most cases, no. There is currently no general Australian law requiring a business to tell customers it uses AI. The government reviewed exactly this question and concluded in late 2025 that existing consumer law is fit for purpose, recommending no new disclosure rule. But that is not the end of the story. There are four situations where staying quiet can breach laws that already exist, and a solid business case for disclosure even when it is optional.

Where silence can break the law

1. Your chatbot pretends to be human. Section 18 of the Australian Consumer Law prohibits misleading or deceptive conduct, and it applies whether or not you meant to mislead. If a customer asks "am I talking to a real person?" and your chatbot says yes, you have a problem. The same section makes you responsible for what your chatbot tells customers.

The cautionary tale is Air Canada, ordered by a Canadian tribunal to compensate a customer after its chatbot gave wrong advice about refunds. Different country, but Australian law works the same way: the business owns the bot's promises.

2. AI-generated reviews or testimonials. A glowing review written by AI and presented as a real customer is misleading conduct, full stop. The same goes for AI-generated "photos" of work you never did or products that do not exist as shown.

3. Customer data goes into AI tools. If you feed personal information into AI systems, your privacy policy should say so. And be careful which tools: free tiers of some AI products may use what you type to train their models, which is a poor place for customer records regardless of what your policy says.

4. You are in a regulated profession. Tax practitioners must obtain client permission before putting client information into an AI tool, under guidance from the Tax Practitioners Board. Similar duties are emerging in law and health. If you hold a professional licence, check your body's position before assuming silence is fine.

What the voluntary standard says

Australia's Voluntary AI Safety Standard recommends telling people when they are interacting with AI or reading AI-generated content. It is guidance, not law. But with national AI legislation expected in early 2027, this guardrail is the most likely candidate to become mandatory. Businesses that adopt it now will not have to change anything later.

The business case for telling people anyway

Customers rarely punish a business for using AI. They punish concealment they discover later. A one-line disclosure costs nothing and buys trust: "Our chat assistant handles common questions instantly and hands you to a human when it can't." That sentence makes AI sound like what it is, a service improvement, rather than a secret. Concealment turns the same tool into a story about what else you might be hiding.

A practical checklist

• Label your chatbot as an assistant or bot. Never let it claim to be human, even as a joke.
• Give customers a path to a real person, and make the handover work.
• Disclose AI content where authenticity is the point: reviews, testimonials, case studies, images of your actual work.
• Skip the disclaimers where nobody expects them. AI-assisted email drafts, social captions and blog outlines that a human has reviewed and owned do not need labels.
• Update your privacy policy if personal information touches any AI tool, and keep customer data out of free-tier products entirely.
• A human signs off before AI output reaches a customer.

Not sure where your setup stands? Disclosure questions come up in most of the free 45-minute consultations we run for eligible Australian SMEs, usually alongside privacy and tool selection. Book a session and we will work through your specific case, or start with our AI risk and governance resources.

Frequently asked questions

Is it illegal to use AI to write my marketing content?

No. AI-assisted marketing content is legal and common. The line is misleading customers: AI-generated reviews presented as real, fake imagery of your products or work, or claims the content itself gets wrong.

Do I need to label AI-generated images?

Not in general, as of 2026. The exception is images that misrepresent something material, such as work you claim to have done or a product's appearance. Those can be misleading conduct regardless of how they were made.

Is my business liable for what my chatbot says?

Yes. Under the Australian Consumer Law, statements your chatbot makes to customers are treated as statements by your business, and the prohibition on misleading conduct applies regardless of intention.

Will AI disclosure become mandatory in Australia?

Possibly from 2027. Transparency is one of the ten guardrails in the Voluntary AI Safety Standard, and the national AI legislation expected in early 2027 may make some form of it binding. Labelling AI interactions now is cheap insurance.