Can you trust AI chatbots for insurance questions? Benefits, risks, limits, and when to see a licensed adviser

AI chatbots for insurance at a glance: benefits, limits, and how an adviser closes the gap

The table below separates what AI chatbots for insurance handle well from what needs a person to check. It also shows what a licensed adviser adds on top: verification against your own policy, your own health history, and settings that have moved since the model was trained.

What are chatbots for insurance?

Chatbots for insurance come in two kinds: a general AI assistant such as ChatGPT, Gemini or Claude. The other is a chatbot embedded on an insurer's or adviser's own website or app to handle that company's enquiries. The first has no sight of your policy, while the second only answers within the scope its owner set.

General AI assistants like ChatGPT, Gemini and Claude

General AI assistants are large language models, and their major strength is predicting the next most likely word. They are not looking anything up. They put words together based on what you tell them and what they have previously been given.

As per Health New Zealand, these tools appear to be good at summarising information and describing it in language that makes sense to people. That said, the same resource also noted they can generate convincing texts while also giving inaccurate or misleading information.

Both halves are true at once, and New Zealanders are asking AI assistants questions anyway. OpenAI says New Zealanders send almost 10 million ChatGPT messages a day. But this pertains to general assistant use rather than a figure specific to insurance.

AI chatbots on an insurer's or adviser's own website

Some chatbots on insurance websites are AI-backed, and some are not.

An older rule-based bot uses conditional if-then logic and offers a fixed menu, so it fails on anything not listed. An AI-backed bot handles free text and can ask a clarifying question when it is unsure what you mean. A menu with no free-text box is often a sign you are dealing with the older, rule-based kind.

At least one New Zealand financial adviser has built and embedded her own chatbot on her practice website.

Benefits of AI chatbots for insurance: what they get right

AI chatbots are good at four things in insurance, starting with translating jargon into plain language and summarising a document you already hold. Health New Zealand's own qualifier applies to the summarising: good at it, but not reliable.

The Financial Markets Authority describes itself as technology-neutral and pro-innovation, and wants firms to use AI to improve customers' experience.

• Explaining how cover works: These tools are good at describing information in language that makes sense to people. Separately, 58% of New Zealanders say they are comfortable with AI explaining financial products clearly. But take note that this pertains to comfort level rather than an accuracy measure.
• Summarising a document you already hold: These tools are good at summarising information, though not reliably. In fact, about two-thirds of writing requests to ChatGPT ask it to edit, critique or translate text the user already has.
• Drafting a complaint: An insurance ombudsman scheme in New Zealand names AI-drafted complaints as useful, particularly for people facing language barriers. They accept the same use, while noting you do not need AI to complain at all.
• Preparing for a meeting: An adviser may ask about your situation, assets, debts, income, expenses, and existing cover. Running through that list with a chatbot first is one way to arrive at the meeting prepared.

Risks and limitations of using AI chatbots for insurance questions

The insurance ombudsman warns that AI can oversimplify complex policy wording, miss key exclusions, and rely on overseas information that does not apply in New Zealand. Invented cases and out-of-date figures complete the set.

Answers built on overseas insurance, not New Zealand's

An AI assistant trained mostly on overseas material answers in overseas vocabulary. New Zealand courts' guidance makes the same point about generative AI: its answers may be based on overseas law that does not apply here.

Four differences catch New Zealand readers out.

• Excess, usually not deductible. Here, the amount you pay towards a claim is the excess. The Insurance Council of New Zealand notes it is sometimes called a 'deductible'.
• Co-payment, not coinsurance. Some New Zealand health policies can apply a percentage co-payment to certain benefits, a mechanic at least one insurer introduced in November 2025.
• ACC has no overseas equivalent. ACC covers injury by accident, not illness or conditions related to ageing, and consumer guidance places ACC and income protection side by side as related cover. That is why ACC and income protection are considered together here.
• Pharmac has no overseas equivalent either. New Zealand is the only country with an agency like it.

Out-of-date figures, thresholds and product names

New Zealand insurance settings move, and a chatbot answers from the version it was trained on. Four examples show how fast the details move, and a model that missed an update still answers with confidence.

• ACC's gross maximum weekly compensation rose to $2,466.20 a week from 1 July 2026, and the adjustment indices behind it are reviewed every year.
• The IFSO Scheme's lump-sum compensation limit rose from $350,000 to $500,000 plus GST for complaints deadlocked on or after 18 July 2024.
• The Contracts of Insurance Act 2024 received Royal assent in November 2024 but is not yet in force. An answer describing it as current law today is wrong.
• Policy documents also move fast. When a New Zealand advice firm ran an AI tool across two real health policies, one of its errors was comparing an outdated version of a policy.

Invented examples, cases and statistics

AI chatbots invent things that look like evidence. The insurance ombudsman warns that AI-generated responses can give incorrect information about likely complaint outcomes and hallucinate previous cases, which leads to incorrect advice. The Banking Ombudsman reports made-up laws and case examples too.

In one example seen by the IFSO Scheme, a Google AI summary suggested claim decisions are "frequently overturned" when consumers complain. It also suggested that 80% to 90% of cases can succeed if people persist. The ombudsman describes that as misleading.

New Zealand's Supreme Court has seen it too. In Jones v Family Court at Whangārei, the Supreme Court found an applicant had cited authorities that appeared to have been hallucinated by an AI application. It warned that relying on false citations may in serious cases amount to contempt of court.

Policy wording summarised without its exclusions

A plain-English summary of a policy drops the parts that decide claims. A New Zealand advice firm uploaded two real health policy documents and asked a popular AI system to compare them, the way a member of the public might.

A researcher checking the output found eight errors. One of them, comparing an outdated version of a policy, is covered above; the rest included:

• missing a guaranteed wording promise about what cover cannot be downgraded to
• missing a separate document that actually decided whether a claim was paid
• getting a major difference in non-Pharmac medicine cover wrong
• leaving out a related benefit category the policy did include
• misapplying a policy-wide limit as though it covered only one small section
• saying a benefit lasted four years when it lasts three
• saying a benefit did not exist when it was listed under another name

Both sources are industry businesses: one an insurance advice firm, the other a research firm serving insurers and advisers. This is industry testing, not independent research. A pre-existing condition definition can turn on a symptom or a sign, with no diagnosis at all. Definitions also differ between providers.

What's safe to type into an insurance chatbot?

General questions about how cover works, policy wording or definitions are safe to put to an AI chatbot. But anything that identifies you is not.

Netsafe's consumer guidance is direct. Do not enter personal or private information such as contact details, passport numbers, or banking information into an AI tool. It may be re-generated as part of an answer for someone else.

New Zealand courts' guidance for people representing themselves sets the same line: enter nothing that is not already in the public domain. Nothing private, confidential, suppressed, or legally privileged should go in either. The insurance ombudsman advises caution about putting personal information into AI tools, particularly free ones.

Health information deserves more care again. Commenting on mental health chatbots specifically, the Privacy Commissioner has encouraged consumers to check what a chatbot collects and how it will be used before engaging with it.

That's particularly true where what they share is health or other sensitive information. The same caution fits health details you might mention to any chatbot, insurance included.

The guidance the Commissioner gives organisations goes further. It tells agencies not to put personal or confidential information into a generative AI tool unless it has been confirmed the input is not retained or disclosed. That expectation is set for organisations, not for you, and it still says something useful about the risk.

How to check a chatbot's insurance answer before you act on it

Check a chatbot's insurance answer in four steps before you act on it. The first two come straight from the insurance ombudsman's consumer checklist [IFSO Scheme, 2026a].

1. Read your own policy wording first. IFSO's first step is to read your full policy and ask your provider how it applies to your situation. The answer to most questions about your cover sits in your policy document and benefit schedule, not in a model's training data.

2. Check it against New Zealand sources. IFSO's own instruction is to double-check any AI-generated information against trusted, New Zealand-based sources. ACC, Pharmac, the Financial Markets Authority and the IFSO Scheme itself are good places to start, and an overseas-shaped answer may not match them.

3. Go back to the original source. If an AI answer refers to a law, policy or obligation, check that source for relevance and credibility. Check what the answer left out, not only what it got wrong.

4. Ask a person how it applies to you. This is the step a chatbot cannot take for you, and it is where Policywise comes in. Policywise's advisers read the wording against your own health history. There is usually no upfront fee, because the adviser is paid a commission built into the premium.

For an automated decision rather than general information, one more test helps: ask who made it, and whether a human will review it.

When to move from AI research to expert insurance advice

Move from AI research to a licensed adviser when your question changes: from how something works to what you should do about your own cover. Recommending or giving an opinion on a specific financial product is regulated financial advice in New Zealand, while describing how products work is not.

All advisers must have their own Financial Advice Provider licence, or be covered by an entity that has one, and robo-advice is not exempt. An adviser owes you the care, diligence, and skill a prudent adviser would exercise.

As trade press reports, AI outputs often feel authoritative, yet carry no statutory duty to act in a client's best interests.

Five moments call for stopping your research and making the call:

• You have a pre-existing condition: Having a pre-existing condition doesn't mean you can't get cover at all; an insurer may still offer it after reviewing further medical information, though usually at a higher premium. The definition can turn on a symptom alone, with no diagnosis needed, so read health insurance for pre-existing conditions and life insurance with pre-existing conditions before you apply.
• A claim has been declined: Taking a dispute to a scheme involves no court hearing and costs you nothing, and an adviser can pass it straight to the scheme.
• You are replacing existing cover: A new policy may require a new qualifying period, and existing health problems may be treated as pre-existing, so a claim could be declined that your original policy would have accepted. It's worth keeping the old policy running until the new one is in force.
• Switching products: An adviser recommending a change has to explain whether the new product suits you and whether it carries exclusions or limitations.
• You want quotes compared for your situation: Premiums are calculated for each applicant from what they disclose. For life insurance specifically, Consumer NZ suggests getting at least three quotes, and the same habit applies when comparing other types of cover. A recommendation about which suits you is regulated advice.

The two work best in that order. Let AI explain the concepts and help draft your questions, then bring those questions to someone who can answer them for your circumstances. Expect an adviser to explain their recommendations, how they are paid, and which dispute resolution scheme they belong to. Reviewing your cover regularly is also part of an adviser's role.

Policywise is an independent adviser, not tied to a single insurer. It compares across the providers it's disclosed with, helps where there is a health history to explain to an underwriter, reviews cover as life changes, and stays with clients at claim time. Compare health insurance, life insurance, income protection or trauma cover with us, or book a callback.

Whether AI will replace brokers altogether is a bigger question, answered in full by Policywise's companion article 'Will AI replace insurance brokers in NZ'.

RECOMMENDED READINGS

How AI is changing health insurance underwriting and claims in NZ

Turn insurance questions into a clear plan with Policywise

Policywise is a 100% free service which tells you which health, life, and disability insurance provider best fits your needs. We offer fast, comprehensive, and easy-to-understand comparisons of all leading providers, and a simple summary clearly recommending which insurer is best for your situation.

Not all insurance policies are the same. Policywise can help you sort out the duds, avoid the lemons, understand the fine print and exclusions, and get the right insurance for you and your family.

We make the important decision of where to buy your insurance super easy. We’ll answer your questions, provide experienced advice and quotes, and manage all the back and forth throughout the application process. Taking out your cover through us means you'll have our lifetime support and claims advocacy, and we'll help you negotiate a positive outcome at claim time. We can also take care of lodging any claims on your behalf and back you up if the going gets tough.

Check out the reviews on our homepage for how other New Zealanders have found our service, because now is the time to get your personal, family, or business insurance sorted. Give your family or someone you love the most outstanding financial support possible. Book a 5-minute callback with Policywise today; our service is fast and free.