Australians ask AI about super and money. What funds and advisers should make clear
Nearly one in five Australians use AI for money decisions. What super funds and advisers should publish and how the Privacy Act rewrite shapes client-facing AI.
The short answer
Make the facts people check before acting answerable: fees, investment options and their returns with dates, insurance cover, eligibility, licence details and how to reach a person. AustralianSuper's survey found 17% of Australians use AI for financial decisions, while 37% distrust it, mostly over where its information comes from. Clear, dated pages give AI answers a source worth trusting.
Key takeaways
- In AustralianSuper's survey of 1,009 Australians (April–May 2026), 17% used AI tools to inform financial decisions, rising to 28% of 18-to-29-year-olds.
- In the same survey, 37% did not trust AI for financial information, and 59% of that group cited not trusting where the information came from.
- Comparing products and understanding super and retirement were among the most common uses, which puts fee and option pages directly in the path of AI answers.
- In Semrush's US data (January–April 2026), the top three brands held 41.4% of AI visibility in finance, against 82.9% in news and media, so finance answers are less locked up than some sectors.
- The second tranche of Privacy Act reform, released as an exposure draft on 31 August 2026, would bring AI-generated inferences more clearly into scope, which matters for any client-facing agent a firm builds.
In this article
What the AustralianSuper survey found
A meaningful minority of Australians already use AI for money decisions, and they check what it tells them. AustralianSuper commissioned Quantum Market Research to survey 1,009 Australians online in April and May 2026, weighted to be nationally representative.1 AustralianSuper is a super fund, so it has an interest in the topic, but it does not sell AI tools.
In that sample, 17% used AI tools to inform financial decisions.1 Use was concentrated among younger people: 28% of 18-to-29-year-olds, against 6% of those aged 60 and over.1
Most users treated AI as a starting point, not the last word. Among them, 54% cross-checked AI answers against other sources and 32% consulted a financial professional.1 One in four, 25%, had acted on AI-generated financial information without checking it elsewhere.1
What Australians who use AI for money ask it to do
Do people trust the answer, or just start with it?
They start with it, and they want to know where it came from. In the AustralianSuper survey, 37% of Australians said they did not trust AI for financial information, and 59% of that group said it was because they did not trust where the information came from.1
Respondents said they would be more comfortable with AI guidance that drew on trusted, credible sources and reflected their own circumstances.1 Among 18-to-29-year-olds, 57% found securely personalised AI guidance appealing, against 34% across all ages.1
Our view: that distrust is an opening for funds and advisers. When an AI answer about fees or insurance in super cites a fund's own dated page, the shopper can check it in one click. When it cites a stale comparison article, the doubt the survey describes is justified. The firm that publishes the checkable source becomes the thing that settles the question.
Surveys in other markets show the same pattern. Our note on Gallup's US and Canadian data found the same mix of low trust and real use. Closer to home, our note on Australian car buyers runs the same kind of check with Australian prompts in another sector.
Why finance answers are still open to newcomers
AI answers in finance spread visibility across more brands than some sectors do, which gives mid-sized funds and advisory firms room to appear. Semrush, which sells AI visibility software, analysed 126 million US prompts from January to April 2026. The top three brands held 41.4% of AI visibility in finance, against 82.9% in news and media.2
The engines also differ in how many sources they draw on. In the same US data, ChatGPT cited about 15 sources per response and Gemini about three.2 An engine that cites fifteen sources has room for a smaller fund's fee page; one that cites three will tend to pick the most established. These are US figures, and Australian answers may differ, which is a reason to measure them directly.
Test the questions Australians actually ask. Prompts such as "which super fund has the lowest fees for a balance of 100,000 dollars", "is insurance through super worth keeping" or "how much do I need to retire comfortably in Australia" return different sources in different engines. Sigzen AI audits run each prompt on three separate days across six engines (ChatGPT, Gemini, Perplexity, Claude, Copilot and Google AI), because single runs vary.
What should funds and advisers make clear?
Publish the facts a member or client checks before acting, as dated HTML on your own site. Most firms already disclose them in product disclosure statements and PDFs that engines read poorly.
| Fact | Where it should live |
|---|---|
| Fees and costs, by option and balance | An HTML table with the date it applies from |
| Investment options and past returns | One page per option, periods and dates stated |
| Insurance in super: cover, cost, how to opt out | A plain-language page, linked from the fee table |
| Eligibility and who a product is not for | On each product page, near the top |
| Licence details and general advice warnings | Footer and product pages, matching the register |
| How to talk to a person | Every page that answers a decision question |
Illustrative. Check every page against your own disclosure obligations.
Two details make these pages more useful to an AI answer. State the date on every figure, so an engine quoting it carries the date with it. And write the short answer first: "Our administration fee is X% a year for balances up to Y" is easier to quote than three paragraphs that reach the number at the end.
How the Privacy Act rewrite affects a client-facing agent
If a fund or adviser builds its own AI assistant for members or clients, the proposed second tranche of Privacy Act reform is the rulebook to design for. As of late September 2026 it is an exposure draft, not law, and its final wording may change.
In a summary published in early September 2026, law firm Dentons reports that the Attorney-General's Department released an exposure draft bill on 31 August 2026, with submissions closing on 18 September.3 As Dentons describes it, the draft would broaden "personal information" to capture behavioural data and AI-generated inferences. It would also clarify that including personal information in a record counts as collecting it, whatever its source, which brings inferred data more clearly into scope.3
Dentons adds that the draft would replace the current collection, use and disclosure rules with a single "fair and reasonable" test, tighten consent, and set a 72-hour deadline to notify the regulator of an eligible data breach.3 For an assistant that infers a member's risk appetite or retirement readiness from a conversation, each of those points applies.
Our view: design any client-facing agent as if the draft will pass. Log what the agent infers and where it stores it, ask for consent in plain words before personalising, hand over to a person for anything that is advice, and keep the agent's sources to your own published, dated pages. This is not legal advice; take that from your own counsel.
Where to start
Start by finding out what AI answers already say about your fees and products, then fix the pages behind the errors. A baseline of twenty to thirty prompts across the main engines shows which facts are wrong and which sources the engines lean on.
Then convert the five or six facts members check most into dated HTML pages, and match them to your listings on comparison sites and the regulator's registers. Our Singapore playbook and our note on Germany cover the same work in other markets. Our page for financial services firms sets out how we audit it.
