Illustrations by ZEBU
In a competitive property market, real estate professionals are freeing up more time in their day for valuable face-to-face interactions by letting artificial intelligence (AI) help create their property listings.
A report by the McKinsey Global Institute has stated that generative-AI could be worth billions of dollars for the real estate industry, yet “for all the hype that gen AI has received to date, many organisations are finding it difficult to implement and scale use cases, and have not yet seen the promised value creation”.
This could be because adoption of new technology that radically changes the way a real estate company works takes time. It’s also true that speed at which AI technology is advancing makes it difficult to predict how it’s going to change the property sector landscape long-term.
Then there’s the fact that AI isn’t a perfect solution. As anyone who has used generative AI such as ChatGPT, Gemini or Claude knows – it is prone to mistakes.
Often just innocuous grammatical mistakes, they can also be rather more significant factual errors, as happened in the case of Australian estate agency LJ Hooker. A New South Wales branch of the company received unwanted press coverage for publishing (and hastily deleting) a property listing that had been written by ChatGPT and not checked properly. The listing said: “One of the standout features of this property is proximity to excellent educational institutions … Farley Primary School and Farley High School.” The problem being that neither school exists.
This is known as an ‘AI hallucination’, when generative AI fabricates answers which sound plausible but are ultimately wrong. In 2023, an AI start-up called Vectara (founded by former Google employees) found that chatbots invent things between 3% and 27% of the time. Much like a writer needs a good editor, generative AI continues to need a good fact checker.
So how are property companies employing the services of AI with a watchful human eye managing proceedings, and are the listings actually any better or capable of generating more sales leads when they were written by a human? Parikshat Chawla, global head of operations at Chestertons, says the organisation is in the early stages of using a property-focused AI marketing platform called Coralytics. It can provide a written description for a property listing based on just a few photos, as well as enhancing the photos themselves if the lighting is poor or the colours aren’t right.
If the property is empty, the software can add furniture in a similar way to how a CG render is created of a yet-to-be-built home. Virtual staging (changing elements of the room), AI-generated 3D floor plans and customised social media content can all be created to improve the quality of the listings and attract potential buyers or tenants.
“We’ve found this is a great tool to save agents a lot of time, especially sales and lettings agents,” says Chawla. However, he is also keenly aware of the potential pitfalls of letting AI get carried away. “We also looked at it from an ethical perspective – we're not positioning the property in a manner that's unrealistic. We're not repainting everything or giving it an industrial look if it's an old cottage next to the forest.”
Legal implications of AI misinformation
With AI-generated content of all kinds rapidly proliferating online, there are questions being asked around whether it should always be clearly labelled as such by responsible organisations, be they news sites, or social media posts. The same questions apply to estate agents, suggests Adrian Tagg MRICS, associate professor in building surveying at the University of Reading.
“Where does the responsibility lie in terms of property misdescription? There's legislation in place to prevent that,” says Tagg.
He cites two pieces of UK legislation that could come into play if someone bought a property based on AI-generated information that turned out to be false. The Consumer Protection from Unfair Trading Regulations 2008 and the Business Protection from Misleading Marketing Regulations 2008 Acts.
Despite these two acts being passed into law some years before ChatGPT and its ilk existed, the principles still apply. “You must be clear if images have been altered or furniture has been AI-generated,” says Tagg. “If you started to mess around with the actual fixtures or the wall finishes for example, I think that would be misdescription territory.” And in the matter of AI-assisted building condition surveys, he points to tort law – “you can sue someone for giving incorrect advice,” he says.
“There’s definitely scope to enhance AI, but it’s still there to assist the agents and not replace them just yet.” Adrian Tagg MRICS, University of Reading
He adds that from a surveying point of view, AI can’t make the connections between a building’s age, a material component, its life cycle and the symptom of a defect. “Maybe it will get there one day, but then who’s liable and responsible for that computer generated advice? And of course, AI can't do the conveyancing. That's always a process that takes time.”
When it comes to the text in a property listing, Tagg says that if it is too obviously AI-generated that could create a level of mistrust. “I think you're more likely to be sceptical of something that was written by AI. The language has to be such that it reads like a real person's written it.”
Making AI-generated text sound human is an issue that Chesterton’s encountered early on with their new software. “The first batch of listings were starting to sound a little bit similar, a lot of phrases were being repeated, as were a lot of adjectives,” says Chawla. “We had to prompt the system to tell the AI to sound less like AI.”
Human oversight
Much of the media coverage around AI focuses on which jobs it’s going to make redundant as it grows in capability, but this includes plenty of speculation and hyperbole. Both Chawla and Tagg are of the opinion that human oversight will be needed for a long time yet and that in terms of the real estate industry, AI is more of an admin assistant than a job threat.
Because as Tagg points out, property is one of the only things left that you can’t buy online in a couple of clicks. “You can buy a £250,000 car online if you want, but you still can’t buy a house. There’s no genuine online option to do so – the buyer doesn't contact the seller online and make an offer, it always goes through an agency. You cannot replace the transaction process.”
He believes the value in AI lies in its ability to give an agent more time to meet with buyers, sellers, landlords and tenants. “That’s where they want to be, and if they can outsource the creation of property listings to AI then they have more time for those discussions. There’s definitely scope to enhance it, but it’s still there to assist the agents and not replace them just yet.”
“I personally don't think that AI is going to take away too many jobs,” says Chawla. “The smart businesses will figure out we're always going to need people, we might just need them in a different way.”
While AI can certainly write convincing property listings, the human element of a property transaction is still essential. And of course, AI still needs someone to check its work and make sure it hasn’t invented a south-facing garden before hitting the ‘publish’ button.
