AI Disclosure: What Australian Businesses Should Tell Customers (2026)
Last updated: August 2026.
Most business owners deploying an AI agent ask the same nervous question: do we have to tell people? The commercial worry underneath it is that disclosure will cost conversions. In practice the opposite is closer to the truth.
On This Page
Written by Dr Priya Jaganathan — Go High Level Certified Admin, Certified AI Tech Stack Consultant and keynote speaker. This is an operations guide, not legal advice — confirm your specific obligations with a qualified adviser.
Why disclosure helps rather than hurts
The fear is that a customer told they are talking to AI will disengage. What actually happens is more interesting.
Undisclosed automation creates a low-grade uncertainty. The customer half-suspects, starts testing the system with odd questions, and interprets any imperfection as evidence of deception. The interaction becomes adversarial.
Disclosed automation sets expectations. The customer knows what they are dealing with, asks straightforward questions, and judges it on whether it solved their problem. They are not disappointed that it is a machine, because nobody told them it was a person.
There is also a simpler commercial argument. A customer who discovers the deception later does not just distrust the chatbot — they distrust you. That is an expensive way to save a sentence.
Where the law currently sits
Australia does not currently have a single general law requiring you to announce that a customer is talking to AI in every context. That does not mean disclosure is optional in practice.
Several existing obligations bear on it. Australian Consumer Law prohibits misleading or deceptive conduct — actively representing an automated system as a human employee sits uncomfortably against that. Sector regulators have their own expectations, particularly in health, legal and financial services.
And there is a specific, dated obligation approaching. From 10 December 2026, under amendments made by the Privacy and Other Legislation Amendment Act 2024, APP entities using personal information in automated decision-making that could affect an individual's rights or interests must set out in their privacy policy the kinds of personal information used, the kinds of decisions made, and how someone can seek meaningful human review. The OAIC indicated it intended to publish guidance by September 2026.
Note the distinction: that obligation is about privacy policy transparency for automated decisions. It is not the same as telling someone mid-conversation that they are chatting with a bot. Sensible businesses do both.
How to word it
Short, plain, and at the start. The goal is that nobody is surprised later.
What works: "Hi, you're chatting with our automated assistant. I can check availability and book you in — say 'human' any time and I'll pass you to the team."
What does not work: burying it in a terms link, using vague phrases like "smart assistant" that most people read as a person, or disclosing only after someone asks directly.
Pair the disclosure with a capability statement and an exit. Telling someone what the system can actually do, and how to reach a person, converts the disclosure from a warning into useful information.
| Situation | Disclose? | How |
|---|---|---|
| Website chat agent | Yes | First message, plus human option |
| Voice agent answering calls | Yes | Opening line, before questions |
| Automated SMS reminders | Light touch | Clear it's automated; reply path |
| AI-assisted email a human sends | Generally not needed | A person reviewed and sent it |
| Automated decision affecting a person | Yes — plus privacy policy | Dec 2026 obligation applies |
| Health, legal or financial context | Yes, prominently | Plus easy human escalation |
If you want your disclosure wording and escalation paths reviewed, book a CRM transition call.
When to disclose more than once
One disclosure at the start covers most interactions. Some situations warrant repeating it.
If a conversation runs long, or a customer returns days later into the same thread, a brief reminder is sensible — people forget, and a returning customer may not have seen the opening message.
If the customer appears distressed or the topic turns sensitive, disclosure should be paired with an immediate offer of a person. This is particularly important in health, legal, disability and financial contexts, where someone may be sharing more than they would with a machine if they remembered it was one.
And if a customer directly asks whether they are talking to a person, the answer is immediate and unambiguous. Any evasion at that point converts a minor question into a trust failure.
Related reading: Australian Privacy Principles and AI agents covers the December obligation in detail, and AI and Australian Consumer Law covers the misleading conduct angle.
Mistakes businesses make
Giving the agent a human name and no disclosure. "Hi, I'm Sarah" from a bot is the case most likely to feel deceptive in hindsight.
Hiding it in terms and conditions. Technically disclosed, practically not. Nobody reads it.
Disclosing but hiding the human option. Telling someone they are talking to a machine while making a person unreachable is worse than not disclosing.
Evading the direct question. An agent that deflects "are you a real person?" will be screenshotted.
Assuming it is only a legal question. Even where no specific rule applies, the commercial case for disclosure stands on its own.
Frequently Asked Questions
Do Australian businesses have to disclose they're using AI?
There is no single general law requiring disclosure in every context, but several obligations bear on it — including Australian Consumer Law's prohibition on misleading conduct and sector-specific regulator expectations. From 10 December 2026 there is also a specific privacy policy transparency obligation for automated decision-making. Take advice on your situation.
Will disclosure reduce our conversion rate?
Generally not. Customers respond to whether the interaction was fast and useful rather than to whether a machine handled it. What reliably damages trust is a customer realising afterwards that they were not told.
How should we word the disclosure?
Plainly and at the start — that they are chatting with an automated assistant, what it can do for them, and how to reach a person. Pairing disclosure with a capability statement turns it into useful information rather than a caveat.
Can we give our AI agent a human name?
You can, but pair it with clear disclosure. A human name with no indication the assistant is automated is the scenario most likely to be perceived as deliberate deception once discovered.
What changes on 10 December 2026?
APP entities that use personal information in automated decision-making capable of affecting someone's rights or interests must disclose in their privacy policy what personal information is used, what kinds of decisions are made, and how a person can request meaningful human review.
Do we need to disclose on automated SMS reminders?
A full disclosure is rarely necessary for a simple appointment reminder, since customers understand these are automated. Make sure it is obviously a system message and that there is a clear way to reach a person if they reply with a real question.
What if a customer asks directly whether they're talking to a bot?
Answer immediately and honestly. Evasion at that point turns a minor question into a trust problem, and the exchange is easily screenshotted and shared.
If you're unsure whether your setup discloses adequately, it's worth a review before December. Book a CRM transition call, or see how we work at Pivot 2 Thrive.
Related Articles
Services
Related guides
- Australian Privacy Principles and AI Agents
- AI and Australian Consumer Law
- What to Do When Your AI Agent Gets It Wrong
- AI Voice Agents vs Chatbots for Australian Business
More
