CPD: When a little knowledge is dangerous – countering the financial ‘guidance’ your clients are getting from AI

The growing use of AI for financial guidance is creating a major consumer protection challenge and reshaping adviser-client engagement models.

Move over, Google

The expression “a little knowledge is a dangerous thing” seems particularly appropriate to financial decisions, where the amounts at stake and the inherent complexity make the risk of consumer harm significant.

More than a decade ago, search engines democratised access to knowledge, and many advisers will have stories of clients who came to appointments preloaded with misunderstandings and misguided questions, courtesy of time spent with ‘Dr Google’.

Today, generative AI tools make Dr Google look like an intern, as AI platforms can provide an immediate, confident, and seemingly personalised response to almost any financial question you can think of. The danger – of consumers acting on incorrect or inappropriate information – has been amplified, and advisers are in the impact zone.

In early 2026, Vanguard’s global chief economist, Joe Davis, suggested ChatGPT may now be the ‘largest provider of financial advice in the world’ – at least in the US^[1]. In Australia, there is also clear evidence of a shift in where Australians seek financial information. Around 9.1 million Australians visited ASIC’s Moneysmart website in 2025–26, down around 17 per cent from the frequently reported 11 million^[2]. ASIC, in explaining the decrease, said changing search behaviour, including the use of AI platforms and AI-generated search summaries, was affecting how consumers access information^[3].

Separately, research^[4] from McCrindle, released in September 2026, suggests more than three million Australians are already using AI to help manage their financial activity. Among Gen Z, the proportion rises to one in four.

This does not mean consumers have stopped valuing professional advice. On the contrary, the McCrindle study found that 42 per cent of Australians were very or extremely likely to act on advice from a paid, in-person financial adviser, compared with just 18 per cent for an AI platform^[5]. But it does suggest the sequence of engagement is changing – clients may consult AI before seeking advice, then bring its conclusions into adviser conversations.

The resulting consumer protection challenge extends beyond obviously false information. A response can be broadly correct but still produce a poor outcome for an individual when applied without the necessary context.

This article examines why AI-generated financial guidance can expose consumers to harm, including the blurred distinction between general information and personal advice and the lack of accountability when an AI-generated answer causes loss. It then considers how advisers can respond constructively to the assumptions clients bring into the advice process, including the privacy implications of using AI tools.

Information, guidance or financial advice?

While the distinction between financial information, general advice, and personal advice is seared into the brain of advisers, consumers are largely oblivious to the difference, especially when they are reading through an impressively detailed ‘financial plan’ from ChatGPT or Claude. This is problematic on a number of levels, including the extent to which that information seems accurate and personalised.

As a spokesperson for ASIC’s Moneysmart program explained: “AI is not bound by the same rules as a licensed financial adviser who has a duty to consider your circumstances before offering personal advice.”^[6]

The regulatory framework for personal advice thus becomes a critical dividing line, and accountability is one of the key differentiators between those sitting inside the framework and those sitting outside.

Under the Corporations Act 2001, a communication can constitute financial product advice if it is intended, or could reasonably be regarded as intended, to influence a decision about a financial product. It may be personal advice where the provider has considered the person’s objectives, financial situation or needs, or where a reasonable person might expect those matters to have been considered.

In Australia, licensed digital advice providers such as Otivo operate inside that framework, providing personal advice under an Australian financial services licence and producing statements of advice. General-purpose AI platforms do not generally operate within that model. ASIC’s Moneysmart warns that AI platforms are generally not licensed to provide personal financial advice. Otivo itself draws the same distinction in positioning its own service, noting that general-purpose tools “such as ChatGPT, Claude and Gemini are explicitly restricted from giving personal financial advice.”^[7]

To Pivot Wealth founder Ben Nash, the moment the line is crossed into personal advice is easy to spot: “[It’s] the moment a tool connects to someone’s bank data and starts answering ‘should I put the spare $500 a month into super or the mortgage.’”^[8]

The right answer to the wrong question

As Brisbane homebuyer Georgina Doll told the ABC^[9], when she used ChatGPT to model repayments across different home loan interest rates, she knew it was not always “100 per cent accurate”. But casually talking about margins for error – as if they were normal – significantly understates the potential scale of those errors. AI hallucinations are well documented, and despite the fact that models are learning and improving, the spectre of false information remains real.

As an ASIC spokesperson warned: “It is also not uncommon for AI models to hallucinate, or invent, information that has no basis in fact, so it is important to not trust its output implicitly.”^[10]

The pace of legislative change in Australia, and the complexity of rules around superannuation and Centrelink, make these margins for error razor thin. When Otivo presented ChatGPT with a hypothetical Australian retirement planning scenario, the response included outdated superannuation contribution limits and missed recent changes to pension eligibility entirely.^[11]

But the issue is bigger than just accuracy. An AI response does not have to contain false information to cause harm. It may be factually correct but still highly unsuitable for the person receiving it.

As one adviser observed: ‘AI-generated financial information is often not incorrect “in a vacuum”‘.^[12] The problem often lies in what the platform has not been asked and therefore has not considered. A client asking whether to salary sacrifice more into super might receive a technically correct explanation of concessional tax treatment, with no way for the model to know that the client is saving for a home deposit or already close to their contribution cap. So, while the tax explanation may be technically right, the resulting strategy may still be wrong.

An experiment reported by academics writing in The Conversation reinforces this point^[13]. When testing ChatGPT, Claude and Perplexity across five vulnerable-consumer scenarios, the models produced structured, practical-sounding responses but did not adequately account for vulnerability, even when it was stated explicitly in the prompt. The scenarios included a pregnant woman planning for maternity leave and a single parent considering a relative’s cryptocurrency tip. In some cases, the suggestions could have worsened the underlying financial harm.^[14]

If a licensed adviser missed this context, they could be held accountable for it, and a trip to AFCA could be on the cards. But when an AI platform misses it, there is no one to be held responsible, leaving consumers in the lurch when something goes wrong.

Plenty of trust, but no accountability

The risk is heightened when consumers place confidence in AI-generated guidance without understanding that it sits outside the protections applying to licensed financial advice.

When personal advice is provided by an Australian financial services licensee or its representative, the provider is subject to many legal obligations and must have an internal dispute resolution process. Consumers may also be able to take an unresolved complaint to the official external body – the Australian Financial Complaints Authority (AFCA). General-purpose AI platforms offer no such mechanisms for recourse. If a consumer follows an unsuitable answer and suffers a loss, there is no licensed advice provider against which to complain, and no AFCA or Compensation Scheme of Last Resort (CSLR) pathways available. As ASIC’s Moneysmart puts it plainly: “AI can help you learn, but it can’t take responsibility for your decisions.”^[15]

Nevertheless, consumers – especially younger ones – seem happy to place their faith in AI. ASIC’s 2026 Moneysmart research^[16] found that 64 per cent of Gen Z respondents trusted AI platforms for financial information and guidance. Almost the same proportion, 63 per cent, expressed confidence in the accuracy and appropriateness of AI-provided financial guidance. Among respondents aged 29 and over, both figures were lower, at 36 per cent. Separate research^[17] commissioned by AustralianSuper found that one in four Australians using AI for financial matters had acted on its information without checking another source.

Research from the UK illustrates how easily this confidence can lead consumers to believe they are ‘protected’. A 2026 Financial Conduct Authority survey^[18] of investors aged 18 to 40 found that 44 per cent mistakenly believed AI-generated financial information was regulated. Almost one-third wrongly believed they could obtain compensation through the Financial Services Compensation Scheme or the Financial Ombudsman Service if AI advice caused them loss.

While these are UK findings, the underlying consumer protection issue seems equally relevant in Australia. Trust in advice-like output may lead consumers to assume there is a regulated provider standing behind it.

For advisers, this becomes a practical consumer protection issue. If a client brings an AI-generated recommendation into a meeting, the first task is to establish how much weight they have placed on it and whether they have already acted. Then the adviser must diligently untangle the facts from the fallacies, and effectively reset the client, using every soft skill and every ounce of diplomacy at their disposal.

Correcting clients without condescension

Increasingly, AI will change the starting point for many advice conversations. Rather than arriving as a ‘blank canvas’, many new clients will arrive with the answer to a question they have already asked and expect the adviser to confirm. Others will use AI to critique recommendations put forward by advisers. This should not be treated as a reflection on the adviser’s experience or competence. A client’s willingness to learn and appetite to understand more should be embraced, and it augurs well for a long-term, engaged client relationship. But starting a relationship on the right footing is critical.

When responding to a client who has used AI to research their options, the first priority is to establish whether the client has already acted on that research. If they have, the adviser must assess the consequences and determine whether corrective action is needed before examining how the recommendation was produced.

If the client has not acted, or once any immediate concern has been addressed, the adviser can turn to the AI exchange itself. Analysing the complete ‘chat’ matters because AI output depends heavily on how the question was framed and what the platform was told. Asking what the client found persuasive can also reveal the concern behind the question.

There is no need to disparage the technology or the client’s decision to use it. After all, most advisers are likely using AI themselves. Instead, consider framing a response this way: “That gives us a useful starting point. Let’s see whether the answer changes when we include the parts of your situation the platform may not have considered.”

This allows the adviser to demonstrate the missing context. In the salary-sacrifice example referenced earlier, that might mean showing how the recommendation changes once the client’s contribution history and need for accessible savings are considered. The issue is no longer that the AI was simply ‘wrong’. It is that the answer was based on an incomplete version of the client.

The adviser’s value in this situation does not come from winning an argument against ChatGPT or Claude. It comes from turning a plausible general answer into a decision that reflects the client’s actual circumstances. The following steps provide a practical way to approach that task.

A practical response to AI-generated guidance

When a client brings AI-generated financial guidance into the advice process:

1. Establish whether the client has acted. This determines whether the conversation is preventative or whether corrective action may already be required.
2. Ask to see the complete exchange. Review the client’s original prompt and the full response, not a summary or isolated recommendation.
3. Find out what persuaded them. Understanding which part of the answer seemed useful helps reveal the client’s underlying concern.
4. Check the claims. Verify calculations, contribution limits and other factual content against current, authoritative Australian sources.
5. Apply the missing context. Test the proposed course of action against the client’s complete circumstances and existing strategy.
6. Explain what changes the answer. Show the client why a factually plausible response may not be appropriate for them, then agree on what happens next.
7. Record the discussion. Retain relevant AI output and document the advice provided in response.

A further question is what personal or financial information the client has already shared with the platform, an issue considered below.

Privacy – the personalisation paradox

If a lack of context explains why some AI guidance is inappropriate or incomplete, giving an AI platform more information should help it produce a more personalised response. The paradox is that the information needed to improve the response may expose the client to another form of harm: loss of privacy.

This concern is already affecting Australians’ willingness to use AI for financial purposes. Colonial First State research¹⁹ released in March 2026 identified data security and privacy as among the biggest barriers to greater use of AI in managing personal finances. While 42 per cent of respondents were comfortable using AI for everyday financial activities such as budgeting and comparing products, only 29 per cent were comfortable with its involvement in investment management.

ASIC’s Moneysmart warns consumers to consider how information entered into an AI platform will be collected, stored and used^[20]. It cautions against providing personal data such as date of birth and financial information – details which could easily find their way into a chat about life insurance sums insured or superannuation switching.

What happens to a person’s data once it has been entered can become harder to verify.

Advisers should therefore ask not only what the client asked the platform, but what information they provided in the process. Where identifying or sensitive information has been disclosed, the client may need to review the platform’s privacy controls and delete stored chats where possible. If login credentials were included, passwords should be changed immediately.

The same caution applies when advisers use AI in their own work. The Office of the Australian Information Commissioner says the Privacy Act applies to uses of AI involving personal information. As a matter of best practice, it recommends that organisations do not enter personal information, particularly sensitive information, into publicly available generative AI tools because of the privacy risks involved.^[21]

Advisers should of course only use AI tools approved by their licensee and in accordance with their practice’s privacy and information-security policies.

Conclusion – a growing trend for advisers to adapt to

AI will increasingly be the first place consumers turn when they have a financial question. The consumer protection challenge is not to stop them from using it, but to prevent general information from being mistaken for personal advice, particularly when an incomplete answer could expose consumers to poor decisions, unsuitable products or financial loss.

This growing uptake of AI is reshaping the sequence of adviser-client engagement, far more dramatically than the arrival of Google ever did. A growing number of clients may consult AI before approaching an adviser and arrive with conclusions already formed. Advisers will need to adapt by treating those exchanges as part of the advice conversation, rather than a personal affront. They will need to become adept at untangling the information provided to a platform and the answers it provided. Rather than being a threat, this can actually be the basis for a strong demonstration of the value of professional advice – advice that is accurate, tailored, and subject to regulation and consumer recourse mechanisms.

If a little knowledge is dangerous when confidence runs ahead of understanding, then helping clients close that gap will become a vital role of professional advice and a critical driver of financial consumer protection.

Take the FAAA accredited quiz to earn 0.5 CPD hour:

CPD Quiz

The following CPD quiz is accredited by the FAAA at 0.5 hour.

Legislated CPD Area: Client Care & Practice (0.25 hrs) and Regulatory Compliance & Consumer Protection (0.25 hrs)

ASIC Knowledge Requirements: Client Engagement (0.25 hrs) and Regulatory Environment (0.25 hrs)

please log in to start this quiz

———–

References:
[1] https://www.ifa.com.au/chatgpt-the-largest-provider-financial-advice-in-the-world
[2] https://www.asic.gov.au/about-asic/news-centre/find-a-media-release/2026-releases/26-074mr-from-anxiety-to-action-helping-australians-to-plan-for-their-financial-future
[3] https://download.asic.gov.au/media/hotigt5a/asic-corporate-plan-2026-27-published-26-august-2026.pdf
[4] https://www.moneymanagement.com.au/financial-advice-meets-the-diy-generation
[5] Ibid.
[6] https://www.abc.net.au/news/2025-10-21/using-ai-chatgpt-for-financial-advice/105882042
[7] https://www.moneymanagement.com.au/otivo-unveils-personalised-digital-advice-app/
[8] Ibid.
[9] https://www.abc.net.au/news/2025-10-21/using-ai-chatgpt-for-financial-advice/105882042
[10] Ibid.
[11] https://www.professionalplanner.com.au/2026/04/people-are-already-using-ai-for-financial-advice-so-lets-make-it-safe/
[12] https://www.ifa.com.au/iadviser-the-invisible-risk-of-ai-financial-advice/
[13]https://theconversation.com/we-asked-chatgpt-claude-and-perplexity-for-financial-advice-what-we-got-was-practical-but-with-big-blind-spots-288948
[14] Ibid.
[15] https://moneysmart.gov.au/online-safety/ai-and-money-decisions
[16] https://download.asic.gov.au/media/3l1l0xpc/26-049mr-asic-moneysmart-gen-z-financial-behaviours-report-2026.pdf
[17] https://www.cyberdaily.au/digital-transformation/13928-australiansuper-finds-aussies-are-using-ai-for-financial-advice
[18] https://www.fca.org.uk/news/press-releases/young-investors-trust-ai-more-tv-or-celebrities
[19]https://www.cfs.com.au/about-us/media/cfs-tech-AI
[20] https://moneysmart.gov.au/online-safety/ai-and-money-decisions
[21]https://www.oaic.gov.au/privacy/privacy-guidance-for-organisations-and-government-agencies/guidance-on-privacy-and-the-use-of-commercially-available-ai-products