Property agents are turning to AI-generated images to sell homes in Singapore

SINGAPORE – When Mr Szabolcs, 40, was browsing online directory PropertyGuru for a new place to rent in Singapore, one listing captured his attention.

Describing it as “too good to be true”, he shared that listing in a Reddit post published in October, writing: “The angles are unreal and the walls’ surfaces are perfect, proportions are non-existent.”

Netizens reacted to the images with criticism over the use of generative artificial intelligence (AI) in real estate. One user commented that it seemed “unethical”. Another poked fun at how a ceiling beam had apparently disappeared.

Alongside the AI-generated renderings in the property listing are images of what appear to be the same apartment in an unfurnished state, making the discrepancies easier to spot.

“I understand ‘concept art’ for purchases and I think it makes sense, if you are a potential buyer, to see the possibilities a unit has to offer,” the Hungarian software engineer, who declined to give his full name, tells The Straits Times.

“However, as someone looking to rent, it doesn’t make any sense,” he adds, noting that tenants often face strict restrictions around how they can change a space, and have little financial interest in renovating a rental unit.

As the listing did not disclose its use of AI, “it hit me more as a scammy tactic”, says Mr Szabolcs.

The property agent who created the listing that drew Mr Szabolcs’ attention declined to comment when contacted by ST.

This is no isolated incident.

Realtors and experts tell ST that it is reflective of how Singapore’s property landscape often affords agents little budget for sprucing up a home before selling it, creating a reliance on cheaper tools like generative AI.

Why is AI being used?

What is behind the rise of AI-generated images in property listings?

“That’s probably still because of us,” says Mr Raphael Yee, founder of AI start-up Renderspace. His company developed a generative AI tool for virtually remodelling homes, which is used by realtors at real estate agencies PropNex and ERA.

Mr Yee no longer actively runs Renderspace, but he says its AI tool is built into the PropNex app used by agents. It is just one of many such tools available to realtors.

In his view, market pressures are the reason behind the rise of AI in real estate.

“Home owners treat agents like trash,” he says. “They want them to sell quickly and sell for higher, but they don’t want to pack up and make the house look neater, or are unwilling to present the house in a nice condition.”

As an agent often competes for listings, presenting home owners with an aesthetically pleasing AI-generated rendering is one way to show he or she might be able to sell.

Ms Michele Cabasug, senior associate vice-president at Sotheby’s International Realty, concurs that market pressures are a key driver.

“Let’s say an owner does stage the home. Sometimes they expect the property agent to pay for it,” she says. “But they’re also not loyal. If I spend $2,000 to $5,000 to stage the space, they could turn around next month and say, ‘Okay, I’ll give the listing to somebody else.’

“Or there are already five other agents listing the same property,” says the realtor, who does not use AI-generated images in her listings.

With exclusivity at times being a luxury among realtors, and many working with small budgets, AI has become an affordable and attractive option for those looking to give their listings a visual boost.

AI-generated homes on the rise?

When queried by ST about examples of AI-generated images on PropertyGuru, the company says it reviewed those listings and reached out to the agents involved to take corrective action where necessary.

PropertyGuru’s chief technology officer Manav Kamboj says his platform does not allow “fully AI-generated photos” on the platform.

He adds that the firm uses an in-house AI-powered moderation tool to conduct automated monitoring of all listing images.

When queried about the use of AI in its listings, PropNex declined to comment.

Mr Luqman Hakim, chief data officer of property directory 99.co, says his platform has not yet observed a significant trend of AI-generated renders.

However, he acknowledges that the availability of increasingly sophisticated AI tools makes it challenging to distinguish between AI-generated creations and traditional renders and photos.

Meanwhile, Mr Eugene Lim, key executive officer of real estate agency ERA Singapore, opines that AI renderings remain uncommon in real estate agency work. This is because the Council for Estate Agencies (CEA) requires the use of actual photographs that accurately reflect a property’s true condition and appearance.

One workaround that some realtors have chosen appears to be presenting AI-generated renderings alongside artists’ impressions and actual photos of the property, without stating which is which.

And the use of AI extends beyond photos, with some agencies using the technology to make transactions more efficient.

In 2023, the CEA lauded two local real estate agencies for incorporating AI into their workflows, by cutting down on the time spent writing up listing descriptions or by using chatbots to process invoices.

Still, even as realtors acknowledge a growing trend of AI-generated images, those speaking to ST say they believe the images are used on only a small fraction of listings.

Ms Cabasug says the use of AI is less common here than in other property markets like New York, where such renderings are common in advertisements and property directories.

Whether this will change remains unclear. Of late, she says, she has been bombarded with advertisements from AI interior design businesses eager to cash in on the trend.

A casual scroll through property listings reveals how difficult it has become to distinguish AI-generated images from traditional photography.

Some telltale signs exist, like odd visual elements, furniture that vanishes into thin air or layouts that shift in unfeasible ways between photos.

Another sign is that when downloaded, images from a listing have file titles like “ChatGPT_image”.

However, in many cases, the absence of disclosure means one can never be certain. 

Enhancement or deception?

For renters like Mr Szabolcs, the rise of AI-generated property listings adds to the complexities one must navigate when looking for a home in Singapore.

He notes that while some photos are edited with “beauty filters”, others change the proportions and layout of a room.

Editing to touch up photos and artists’ impressions has long been a part of selling real estate. What makes the use of AI complicated is that it blurs the lines between ideal representation and deception.

Mr Shen Jiaming, senior group division director at ERA, says: “It should be used to help, not to deceive.”

He uses AI editing tools to make photos neater and more appealing, while taking care not to change the characteristics of a property entirely.

“It has to be done correctly,” he says. “There must be appropriate disclaimers to let the consumers know that the photo has AI elements if there are too-drastic changes to the photo.”

Mr Luqman of 99.co agrees, noting that the platform is monitoring developments closely.

“At this stage, we believe that if a property image has been AI-generated or significantly altered, it should be labelled as such,” he says.

Renderspace’s Mr Yee argues that AI-generated renders are not inherently problematic, comparing them to artists’ impressions of new developments.

“When you buy a house, you don’t buy the previous home owner’s living condition,” he says. “You probably will renovate it and redo the furnishing.”

However, mislabelling remains a concern. “We always tell agents to state it in their description,” he adds.

For now, to agents like Ms Cabasug, the tactic remains risky because of its potential to backfire.

“Where it could be super relevant is for sales. For rentals, not so much,” she says. “But when people show up, they’re going to be disappointed, right?”