Should I put an AI chatbot on my business website?
An AI chat assistant can answer common questions outside business hours and take load off the phone. It can also make things up, quote the wrong price or promise a refund you don't offer, and customers will treat its answers as yours. Set it up with tight limits before it goes live. If you want help choosing and setting one up, see our AI for business service.
Last updated
The short answer
You can, but treat the chatbot as part of your website, because customers and regulators will. Under the Australian Consumer Law, anything a business tells customers about its products, prices or services must be accurate, and it doesn't matter whether the business meant to mislead. In a 2024 Canadian case, Moffatt v Air Canada, an airline had to pay a customer after its website chatbot gave wrong refund advice. No Australian court has ruled on a chatbot yet, but the Federal Court has already held Trivago responsible for misleading people through the way its website's algorithm ranked hotel prices, with penalties of $44.7 million. If the Privacy Act covers your business, the OAIC recommends saying in your privacy policy that you use AI and to label public chatbots clearly as AI. Keep the chatbot's scope narrow, base its answers on your own pages, hand anything tricky to a person and test it before launch.
What an AI chatbot can get wrong
Generative AI produces likely-sounding text. It doesn't check facts. Australia's cyber security agency, the ASD's ACSC, warns that generative AI systems are known to make up information that isn't true, which is often called hallucination. On a business website that can mean a wrong price, an opening hour you don't keep, a refund policy you never wrote or advice you're not qualified to give. The ACSC also warns about prompt injection, where someone types hidden instructions to make the AI ignore its rules. A visitor might trick your chatbot into offering a discount, saying something offensive or revealing information it was given to work with.
The main risks
Made-up answers
The chatbot fills gaps with confident guesses. If your website doesn't cover a question, it may invent an answer.
Wrong prices and policies
Old or misread information about prices, refunds, warranties or delivery can mislead customers, and you're responsible for what it says.
Prompt injection
A visitor types instructions designed to make the chatbot break its rules, change its tone or give away what it was told.
Personal information
People type names, phone numbers, health details and complaints into chat boxes. Those conversations are records you hold, and the AI provider may keep them too.
Provider settings
The OAIC notes that some commercial AI products have terms or settings that let the provider use what customers type to train its AI. Check this before you sign up.
Decisions about people
A chatbot that only answers questions is low risk. One that approves refunds, books or turns people away, or screens them, raises the stakes.
What the law says, in plain terms
- Consumer law. The ACCC says any information a business gives about its products or services must be accurate, truthful and based on reasonable grounds, including prices and anything on its website. It makes no difference whether the business intended to mislead. A chatbot answer is information from your business.
- The Air Canada case. In Moffatt v Air Canada (2024 BCCRT 149), a Canadian tribunal rejected the airline's argument that its chatbot was responsible for its own actions. The tribunal said the business is responsible for all the information on its website, whether it comes from a static page or a chatbot, and ordered the airline to pay the fare difference plus interest and fees. It's an overseas small claims decision, so it doesn't bind Australian courts, but it shows how a customer and a tribunal are likely to see it.
- The Australian example. No Australian court has ruled on a chatbot yet. The closest case is ACCC v Trivago: the Federal Court found Trivago misled consumers because its website's ranking algorithm gave prominence to hotel booking sites that paid it more, so the top offer often wasn't the cheapest. Trivago was ordered to pay $44.7 million in penalties in 2022. The business was held responsible for what its automated system showed customers.
- Privacy. If the Privacy Act covers your business, it applies to personal information typed into the chatbot and to anything the AI produces about a person, even if that output is wrong. The OAIC recommends updating your privacy policy and notices about your use of AI, clearly identifying public chatbots as AI, and not entering personal information, especially sensitive information, into publicly available AI tools. Businesses with annual turnover of $3 million or less are generally exempt, but some, such as health service providers, are covered whatever their size.
- The 10 December 2026 privacy policy rule. From 10 December 2026, covered businesses must explain in their privacy policy when a computer program uses personal information to make, or substantially and directly help make, decisions that could significantly affect people. A chatbot that only answers common questions probably isn't doing this. One that approves, refuses or screens people might be. Our guide to the automated decisions rule explains the three questions to ask.
- Digital Duty of Care. This is proposed and isn't law. The bill was introduced on 15 October 2026 and, as introduced, has no small business exemption. Law firm Allens notes the draft reached many website operators and AI tools that let users generate and share content. See our small business Digital Duty of Care guide.
Guardrails to set before it goes live
- Limit what it talks about. Decide which questions it should answer, such as hours, services, how to book and where you are. Tell it to decline everything else, including legal, medical, tax or financial advice.
- Base its answers on your own content. Point it at your website pages and approved documents rather than letting it answer from general knowledge. Keep those pages current, especially prices, refunds and warranties, and have it link to the page it used.
- Make it easy to reach a person. Give a clear way to call, email or leave a message, and have the chatbot offer it whenever a question is about money, complaints, refunds or anything it isn't sure of.
- Say it's AI. Label the chat window as an AI assistant, as the OAIC recommends, and add a short note that answers may be wrong and your written policies apply.
- Don't ask for sensitive information. Tell visitors not to type health, financial or identity details, and don't design the chatbot to collect them. Check where the provider stores conversations, how long it keeps them and whether it trains on them.
- Keep logs and read them. Keep conversation records so you can see what it told customers, fix wrong answers and deal with complaints. Treat them as personal information, limit who can see them and delete them on a schedule.
- Test it, then keep testing. Before launch, ask it awkward questions, try to trick it into discounts or rude replies, and check every price and policy answer. The OAIC says AI products shouldn't be set and forgotten, so review it regularly and after any change to your prices or policies.
- Update your privacy policy. Mention the chatbot, what it collects, which provider runs it and whether information goes overseas. Ask your lawyer to check the wording.
How WaTo can help
- Through our AI for business service we can help you choose a chatbot, limit it to your own content, set up the hand-off to a person and test it before it goes live.
- Our AI policy and training service covers the written rules and staff training that sit behind it, such as who reviews the logs and what the chatbot may answer.
- Our website security and compliance audit lists the chatbot, forms and tracking tools on your site so you or your lawyer can update the privacy policy.
- This page is general information, not legal advice. We aren't lawyers, so for advice on your situation talk to a lawyer.
AI chatbot questions
Is my business responsible for what its AI chatbot says?
In practice, yes. The ACCC says information a business provides about its products and services must be accurate, whatever the channel, and intent doesn't matter. In Moffatt v Air Canada, a Canadian tribunal held the airline responsible for its chatbot's wrong refund advice. In Australia, the Federal Court held Trivago responsible for misleading results produced by its website's ranking algorithm.
Do I have to tell people they're talking to an AI?
The OAIC recommends that public chatbots are clearly identified as AI, and that businesses covered by the Privacy Act update their privacy policy and notices about their use of AI. It's also simple good practice, because customers are less likely to rely on a labelled AI answer as a firm promise.
Does my chatbot need to be in my privacy policy?
If the Privacy Act covers your business and the chatbot handles personal information, the OAIC recommends that your privacy policy explains your use of AI. From 10 December 2026, a chatbot that makes or substantially helps make significant decisions about people also needs to be described under the new automated decisions rule.
What is prompt injection?
It's when someone types instructions designed to make an AI ignore its rules. Australia's ACSC describes it as an attack that can hijack the AI's output and get around its safeguards. Limiting what the chatbot can see and do keeps the damage small if someone tries it.
Can a chatbot stop making things up?
Not completely. Basing answers on your own pages, telling it to say it doesn't know, and handing tricky questions to a person all reduce the problem. You still need to test it and read the logs.
Does the Digital Duty of Care apply to my chatbot?
It's proposed and not yet law. As introduced, the bill has no small business exemption and could reach many websites, including AI tools that let users generate and share content. Wait for the final law and eSafety's guidance, and keep a simple list of what your chatbot does.
How WaTo can help with this
Guides on this topic
- The privacy policy change coming on 10 December 2026
- Practical AI governance for small businesses
- Does the Digital Duty of Care apply to my small business?
- Does my website need a privacy policy?
Sources & references
- OAIC: Guidance on privacy and the use of commercially available AI products
- OAIC: APP Guidelines, Chapter 1, APP 1 open and transparent management of personal information
- OAIC: Privacy for small business
- ACCC: False or misleading statements
- ACCC: Trivago to pay $44.7 million in penalties for misleading consumers over hotel room rates
- Civil Resolution Tribunal (British Columbia): Moffatt v. Air Canada, 2024 BCCRT 149
- Australian Signals Directorate, ACSC: Engaging with artificial intelligence
- Department of Infrastructure: Digital Duty of Care
- Allens: Australia's digital duty of care, what the draft bill means for online services (September 2026)
WaTo Consulting. Last reviewed 2026-11-13. Review after 10 December 2026, when the Digital Duty of Care bill passes or changes, and when the OAIC or ACCC publish new AI guidance.
Thinking about a chatbot for your website?
Book a call and we'll talk through what it should answer, what it shouldn't, and how to set it up so it doesn't make promises for you.
Mon to Fri, 8:30am to 5pm
