The short answer: AI is genuinely useful for understanding how super works, and genuinely unreliable for the numbers that decide what you can actually do. The concepts it explains well. The caps, thresholds and rules it confidently gets wrong, because they change faster than the models learn.
We’re not anti-AI at Wealthlab. Plenty of our clients arrive at meetings better informed because they spent an evening asking a chatbot how the assets test works, and that’s a good thing. Scott and Phil made exactly this point wrapping up the year on the podcast: AI is great for ideas, but outdated on super, and the nuances are where the real money sits. Here’s what that means in practice.
The problem isn’t intelligence. It’s the calendar.
AI models learn from information with a cutoff date, then answer with total confidence as if nothing has changed since. Australian super and pension rules, meanwhile, change constantly. Not occasionally. Constantly.
Look at 2026 alone. The concessional cap rose to $32,500 and the non-concessional cap to $130,000 on 1 July. The transfer balance cap moved to $2.1 million. Deeming rates changed in March and again on 20 September. Age Pension rates and thresholds moved in March, July and September. Payday super commenced. Division 296 started. New SMSF residential borrowing was banned from 10 August. The CGT discount and negative gearing rules were rewritten, commencing 1 July 2027.
An AI trained even twelve months ago misses most of that list, and it won’t tell you it’s missing anything. It will quote last year’s cap with the same confident tone it uses for this year’s. That’s the trap: the errors don’t look like errors.
Where the nuance bites
Even when the figures are current, the interpretation is where things go wrong. Phil’s example from our episode on free DIY super tools applies just as much to AI answers: “A balanced fund is not a true balanced fund with most of these funds these days. They are every day of the week a growth fund that they slap the name balanced on.”
An AI comparing your fund options takes the labels at face value. It doesn’t know that two funds both called “balanced” can hold wildly different growth allocations, or that the comparison you asked for is quietly comparing apples with oranges. We see the same pattern with catch-up contributions (eligibility hinges on your total super balance at a specific date), bring-forward contributions (the available amount depends on where your balance sat at 30 June), and Age Pension questions (where the answer changes with homeownership, relationship status and which test binds). The general explanation is right. The application to an actual person is where it slides.
The bigger gap: AI can’t give you advice, and isn’t allowed to
There’s a structural point underneath the accuracy one. Personal financial advice in Australia is a licensed activity for a reason: a licensed adviser has a legal best-interests duty, professional standards, and accountability if the advice is wrong. An AI chatbot has none of those. It isn’t licensed, it owes you nothing, and if its confident answer costs you money, there’s no complaints scheme to call.
That’s not a criticism of the technology. It’s a category difference. General information, which AI can provide, tells you how the rules work. Personal advice tells you what to do given your income, balance, relationship, health and goals, and takes responsibility for it. The gap between those two is exactly where retirement mistakes live, and it’s the same gap that call centres marketing “just use your super” exploit from the other direction.
How we’d suggest using AI well
Used for the right jobs, AI genuinely helps. It’s good for learning concepts at your own pace, in plain language, without feeling silly for asking. It’s good for preparing questions before a meeting with an adviser or accountant, so the paid hour goes further. It’s good for translating jargon in a statement or a product disclosure document into English.
Where the discipline matters is the numbers. We generally suggest treating every figure an AI gives you as a hypothesis to verify, not a fact to act on. The authoritative sources are free: the ATO for caps and tax, Services Australia for pension rates and tests, and MoneySmart for general guidance. All three publish current figures with effective dates, which is precisely what AI can’t guarantee.
And for anything that ends in a decision, the checking step is a human who’s licensed to be wrong-proofed: your accountant for tax, a financial adviser for strategy. Ironically, the year-in-review lesson Scott and Phil kept coming back to on the 2025 wins and losses episode was that the clients who did best weren’t the ones with the cleverest information sources. They were the ones who acted early and got the sequencing right, which is judgement, not data retrieval.
If you want a starting point that’s built on current Australian figures, the free Wealthlab super calculator is maintained against the actual rules, and it takes two minutes.
FAQ
Can AI give financial advice in Australia? No. Providing personal financial advice is a licensed activity under Australian law, and AI chatbots aren’t licensed, don’t owe you a best-interests duty, and carry no accountability for errors. They can provide general information about how rules work, which is a different category from advice about what to do in your specific situation.
Is ChatGPT accurate about Australian super rules? Often for concepts, unreliably for figures. Super caps, pension rates, deeming rates and thresholds change multiple times a year, while AI models have training cutoffs, so they frequently quote superseded numbers with full confidence. Any figure matters enough to act on is worth verifying against the ATO or Services Australia.
What changed in super rules recently that AI might miss? In 2026 alone: contribution caps rose on 1 July, deeming rates changed twice, Age Pension rates and thresholds moved three times, payday super and Division 296 commenced, new SMSF residential borrowing was banned, and the CGT and negative gearing rules were rewritten. A model trained before these changes answers as if none of them happened.
What is AI actually good for in retirement planning? Learning how the system works, preparing sharper questions for professionals, and translating jargon. It compresses the education phase. The verification and decision phases still belong with current official sources and licensed professionals.
Where can I check current super and pension figures for free? The ATO (ato.gov.au) for contribution caps and super tax, Services Australia (servicesaustralia.gov.au) for Age Pension rates, tests and deeming, and ASIC’s MoneySmart (moneysmart.gov.au) for general guidance. All publish current figures with effective dates.
Why do super rules change so often? Most caps, thresholds and payment rates are indexed to inflation or wages, with scheduled updates in March, July and September, and governments layer policy changes on top. It’s a feature of the system, and it’s exactly why any static source, human or AI, drifts out of date.
Bring us your AI homework
Genuinely. If you’ve been researching your retirement with a chatbot and you’ve arrived with a list of ideas and half-answers, that’s a great starting point for a conversation, and we’d rather check the numbers with you than have you act on last year’s caps. Book a free chat with the Wealthlab team, or take the free Wealthlab retirement quiz for a general snapshot built on current figures. Our retirement planning process starts wherever your research got you to.
