Real estate agents edition
Funda's Virtual Styler draws doors straight through rooms, the Dutch Data Protection Authority fines Uber 800 million euros, and 7.2 million people in the Netherlands now use AI.
Funda's AI styler makes mistakes: a wall to the neighbours and a staircase to nowhere
The Virtual Styler in the Gold package sometimes creates images that make a home look different than it really is. You need to explain that to the buyer, not Funda.
On 23 September, media reported that Funda's AI tool Virtual Styler sometimes creates unrealistic or misleading images. Examples given: a wall that seems to run into the neighbour's home, a staircase that leads nowhere, and a small room that looks big. The tool is part of the Gold package that Funda launched in mid-September. It's meant to help buyers picture a different layout, but according to the reports, AI sometimes gets this wrong. This is not a distant risk. If you use the Gold package for a client, these images end up on your listing. Funda's rules say AI images must be clearly labelled, but that doesn't fix the error: a buyer who sees an extra door thinks that door is really there. This discussion isn't new. Earlier this month, NVM and the TV programme Radar already said that home listings can be misleading. This tool adds to that. This week, check every AI-styled image of your own listings against the original photo, and remove anything that changes the floor plan. A styler like this doesn't look at the building plans. It looks at the photo and fills in what such rooms usually contain. That's where the error creeps in: the model doesn't know what a load-bearing wall is or where a staircase leads. It creates an image that looks correct. That explains the reported mistakes. A door in a wall is a believable detail to the model, even if the neighbours live there. And a small room looks bigger when the model adds furniture that wouldn't actually fit. So you can't check an image just by seeing if it looks nice. Put it next to the real photo and the floor plan, and ask for every difference: does this change what the buyer thinks they're seeing? The line is the same as always: showing atmosphere is fine, making the home look better than it is, is not. If an AI mistake changes the layout, it's no longer atmosphere, it's a different home. You're the one facing the buyer at the viewing, not the supplier of the tool. Also ask yourself what you're paying for. Gold is the most expensive package, and Funda is raising its prices too. Trade bodies are asking for clarity on what you get in return, and Funda points to this very AI feature as justification for the price rise. If the styler isn't reliable, you're paying more for a feature that gives you extra checking work. That's worth factoring into your advice to the seller.
Background for subscribersThis edition shows that AI tools in your profession can make mistakes that you have to catch: a styler that changes the floor plan, a system that rejects people without checks, and a valuation tool that takes over typing but not your judgement. In every case, you remain responsible for what the client sees or hears, not the supplier of the tool.
800 million euro fine because a system decided about people on its own
The Dutch Data Protection Authority fined Uber more than 800 million euros. The reason: Uber removed drivers from its system in a fully automated way, without any human checking it. It's one of the largest privacy fines ever for automated decision-making. For your profession, this isn't some distant issue. Letting agents and property managers work with tenant selection. Valuers work with model values that feed into a lending decision. Agencies use tools that pre-sort candidates or responses. Wherever a system pushes people in or out of a process, this ruling applies. This week, write down which step in your tenant or file selection happens without human review. The difference isn't the technology, it's who signs off. A model that ranks candidate tenants doesn't formally decide anything. But if nobody ever looks at the bottom of that list, the ranking is, in practice, the decision. The test is simple: can someone at your firm actually reject a system's suggestion, and can you show that this really happens? If you can't name a single case where a person overruled it, that human step is just a formality. Note: this doesn't mean you can't automate anything anymore. That wasn't the problem here. The problem was that people were dropped from a process without anyone checking it, and without any way to challenge it. For tenant selection, there's also the Good Landlordship Act: you must be able to explain the grounds on which you select. A model that sets its own ranking doesn't give you that explanation automatically.
Background for subscribersThis is a taster.
Subscribers read the whole Real estate agents edition: every story, each with its background in plain language.
Subscribe →Not ready for that? Read the free letter first →
AI and tech, made simple.
This is the Real estate agents edition. Subscribers get the background to every story, in plain language.
Subscribe
inklaretaal · Amsterdam · © 2026