Queensland property investors who relied on artificial intelligence (AI) were at risk of losing hundreds of thousands of dollars as groundbreaking new research found chatbots were widely missing the mark in the state’s dynamic property market.
An inaugural report from financial services firm MCG Number Surveyors exposed major flaws in popular large-language AI systems such as ChatGPT.
Tasked with providing hyper-local investment tips for the best returns, the national study revealed that AI remained woefully inept at simply gathering information, leading to questionable suggestions for locations and property types within a $1 million budget.
![](data:image/svg+xml,%3Csvg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 1280 720"%3E%3C/svg%3E)
ChatGPT returned flawed data for half of its ownership choices
When ChatGPT was asked to list suburbs that met key investment criteria, it returned flawed data for half of its choices, both nationally and for Queensland.
The findings are significant given the rapid updating of AI-powered tools. World Economic Forum polls show that 48 percent of investors of all ages in emerging markets are open to AI-enabled decisions, with half of Gen Z and Millennial users using it for investment planning.
The study, which used both a ‘basic’ prompt and a data-rich ‘deep research’ workflow, found that the latter amplified rather than corrected the model’s bias.
MCG founder Mike Mortlock
MCG founder Mike Mortlock said ChatGPT Pro returned errors in key metrics for Brisbane picks Bowen Hills, Woolloongabba/Dutton Park and Redcliffe.
“In Brisbane, the main issues were less about poor choices in the suburbs and more about the way ChatGPT interpreted the data,” Mortlock said.
“It consistently selected units despite a $1 million budget that could easily have created a detached house on the same sites.
“That preference for higher-yielding assets meant that stronger opportunities for long-term capital growth were overlooked.”
He said the AI used incorrect or mismatched data for details such as the number of days on market and the 12-month price change, and placed too much weight on rental yields over fundamentals such as land values and supply pressures.
A three-bedroom house priced under $1 million for sale at 143 Scarborough Rd, Redcliffe
“In short, ChatGPT didn’t pick bad suburbs. It picked the wrong type of assets. That’s the difference between an AI reading a spreadsheet and a human reading the market.”
Mr Mortlock said the model was being overtaken by Brisbane’s fast-moving entry-level market, where prices had risen by $50,000 to $150,000 in just a month due to huge demand for the government’s expanded first home buyer program.
“For houses under $1 million, you’re looking more at the edges of Moreton Bay, Ipswich and the like.”
Paradoxically, the simplest, free version of ChatGPT delivered better insights for Brisbane, with fewer errors in the output than for other cities.
The basic choices in the suburbs were Woolloongabba (1-2 bedroom units), Chermside (2 bedroom units) and Albion (two bedroom units).
Buyer’s agent Lauren Jones said investing is an art and a science
Brisbane buyer’s agent Lauren Jones said property investing required a nuanced approach that AI couldn’t replicate.
“I always say that real estate investing is both an art and a science,” said Ms. Jones. “If you take away the ‘art’ part you run the risk of buying a property that is not suitable or in a location that may have problems such as social challenges.”
She emphasized the importance of local understanding: “Being able to walk around the neighborhood and understand what people want in this location is a crucial part of a good investment.”
Despite its shortcomings, Ms Jones said AI was useful for gathering basic information about demographics, employment centres, local schools and public transport.
Woolloongbbabba units were recommended by both the Pro and Basic versions of the LLM
MORE NEWS
Hopes of a rate cut over Christmas faded as banks postponed relief
Australian job that destroys relationships
‘Ridiculously underestimated’: investor buys sixty houses in six months
“If you still want to rely on AI, give it information from credible sources, rather than just scanning the entire internet,” she said.
“Also connect what the tech advised with a human element, for example by walking the streets of the neighborhood or going to the local shops.”
PropTech founder Aaron Scott said AI was limited to published or historical information and tended to make assumptions about user requirements, potentially leading investors astray.
Many benefits, but with potential risks
Mr Scott of property comparison platform bRight Agent said property transactions inherently involve human interaction, motivations and negotiation – factors that AI had not yet mastered.
“AI needs to really understand the human asking for advice before handing out the advice itself,” he said, suggesting that future “AI wrappers” or smart forms could build user profiles.
“To get better at providing specific real estate investment advice, AI must follow up a human prompt with an AI prompt back to the human before looking for the ‘right’ answer,” Scott said.
However, he remained “incredibly optimistic about AI in the Australian property sector, and sees many benefits for everyday homeowners over time”.
Two bedroom unit for sale at 1003/50-54 Hudson Rd, Albion
Antonia Mercorella, CEO of the Real Estate Institute of Queensland (REIQ), viewed AI as a “useful tool” for industry professionals, streamlining communications, reviewing investigations and assisting with marketing.
But she issued a strong warning about which tasks are outsourced, as real estate is one of the most highly regulated professions in Australia.
“The use of AI in the real estate context carries the significant risk of not correctly interpreting or considering the nuances in the law, potentially leading to unintentional violations of the law,” said Ms. Mercorella.
REIQ CEO Antonia Mercorella: Don’t take claims at face value
The warning also applied to buyers and investors.
“It is imperative that you understand the source, parameters, methodology and broader context of the information presented.
“Don’t take claims at face value and ensure statements are verifiable, substantiated and relevant to your specific goals and circumstances.
“As for predicting real estate performance, remember that no one, not even AI, has a crystal ball that can predict future outcomes with 100 percent accuracy,” she said.
Mr Mortlock concluded with a stark warning: “Investors who treat AI as an autopilot risk buying the wrong asset in the right zip code. Keep people informed, validate every number and use AI to speed up due diligence, not replace it.”
#AIs #milliondollar #blunder #Chatbots #fail #Qld #real #estate #investors #realestate.com.au