Nine in 10 estate agents reckon AI software – known as automated valuation models (AVMs) – are routinely undervaluing properties.
A poll of 250 estate agents revealed UK homeowners could potentially be losing out when selling their property – as 87 per cent feel AI-powered valuation tools fail to reflect their true value, including in Northern, lower-income, and rural areas.
But 78 per cent rely heavily on AI or automated valuation models (AVMs) when putting a price on a home – and as many as 23 per cent claim to ‘always’ use them.
Despite this, 73 per cent admitted they don’t even fully trust the tools they’re using to make these calculations.
And 15 per cent only adjust the final calculations by between £0 and £1,000 – which may mean some sellers are losing a lot of the potential value on their homes.
A spokesperson from estate agency platform Alto, which commissioned the research, said: “AI is encroaching more and more on so many areas of our lives, but housing valuations isn’t one that usually springs to the front of people’s minds.
“Agents are telling us AI is undervaluing sellers’ homes – and we should be listening.
“These tools can save time and provide a starting point, but they’re no substitute for local knowledge and real-world experience.
“The risk is that homes are priced based on flawed data, and sellers lose out.
“We trust agents – and we believe they should be empowered, not replaced.”
Almost eight in 10 agents (77 per cent) think this technology, at least sometimes, fails to account for key factors that influence a property’s true value.
One area they felt was especially lacking was the algorithms’ abilities to detect added value to a house, in areas such as recent renovations, unique features, or the condition of the property.
Other important elements often overlooked during an evaluation were noise levels (29 per cent), crime rates (26 per cent), and economic conditions (24 per cent).
Especially when considering elements homeowners have invested in themselves, these oversights can be costly – with the most ignored additions being extensions (26 per cent), loft conversions (26 per cent), and smart home technology (25 per cent).
The least accurate valuations were also found to occur in properties in rapidly changing markets (27 per cent), rural locations (23 per cent), and transitional neighbourhoods (21 per cent).
Agents have their own criticisms of these helpers, with the top complaints centring around their lack of accuracy on evaluations generally (23 per cent), but especially when it comes to more unique properties (32 per cent).
However, they also bemoaned the lack of personal touch (21 per cent) and potential for outdated information (20 per cent), limited data sources (18 per cent), and algorithmic bias (17 per cent).
Almost four in 10 (37 per cent) also believe themselves and their fellow agents are prone to overusing this technology.
But two thirds continue to use the AI-driven systems for the prime reason that they save time, according to OnePoll.com data.
Estate agents in specific UK regions that are most likely to report AI under-valuations include the North East (22 per cent), East of England (22 per cent), and Yorkshire and the Humber (21 per cent).
This has led to 11 per cent of agents in the North East saying, on average, they have to adjust AI-generated listing prices by £20,000 or more – underscoring the risk that sellers may unknowingly lose out when marketing their home.
Alto’s spokesperson added: “This just goes to show there are inconsistencies with this technology – across the country and when it comes to specific features.
“We need to ensure that we’re not losing our human touch and that people can trust in their estate agents during a time that can already be incredibly stressful.
“AI has its place, but it should be used to support professionals, not override them.
“Selling your home can be an emotional time and will only be more so if people feel they’re not getting what they truly deserve.”
