Nearly a quarter of Brits – 24% – have used artificial intelligence (AI) for mortgage advice, but confidence in its accuracy is low.
Barratt Homes found that the main reason people used AI was to understand mortgage jargon, cited by 16%, while 12% used the technology for guidance on how much they could borrow. Some 12% turned to AI for general home buying advice, 11% to understand interest rates and 10% to compare mortgage deals.
Despite the notable use of the technology, the research revealed only 7% of people were ‘very confident’ about how accurate AI is.
Women have less trust in AI than men, with 39% expressing low confidence in its accuracy with mortgage guidance, compared to 32% of men.
Different results from various tools
Terry Higgins, group managing director at TNHG Mortgage Services, said consumers should be cautious when using AI for any financial decisions, as different tools could make different suggestions with the same prompt.
Conversations you need to have with landlords before the Renters’ Rights Act
Sponsored by BM Solutions
Higgins said: “AI search tools can be helpful for making mortgage jargon and other information more accessible to first-time buyers and homeowners. However, users should be careful when using them, especially for financial decisions, as outputs aren’t always accurate and could be biased.
“A qualified mortgage adviser will look at your full financial picture and will match you with lenders and products that suit your circumstances. They can also explain the full costs involved and guide you through each step of the purchase, which is something AI simply can’t do.”
|
Model
|
Affordability stance |
Fixed-rate view |
Guidance style |
|
Copilot |
The most neutral, saying the purchase “lined up well with typical UK lending patterns” without leaning optimistic or cautious. |
Offered a balanced comparison of two‑ vs five‑year fixes, framing the choice around lifestyle, budgeting, and risk comfort. |
Practical, simple, consumer‑friendly. Focused on actionable steps like getting a decision in principle. |
|
ChatGPT |
The most cautious, describing the £230,000 target as “borderline but plausible,” depending on lender strictness and affordability checks. |
Leaned toward a five‑year fix, highlighting stability for first‑time buyers and the risk of future rate rises. |
Detailed, structured, broker‑like. Used step‑by‑step breakdowns and multiple caveats. |
|
Grok |
The most optimistic, describing the scenario as “promising and affordable,” mentions lenders offering five times and above income multiples. |
Favoured a two‑year fix, citing 2026 forecasts, expected rate drops, and competitive short‑term pricing. |
Analytical and data‑heavy, heavily referencing market conditions and interest rate movements. |
Higgins said calculating affordability was “one of the most complex parts of buying a house”, with “so many factors at play”.
“While AI can be a good starting point, it’s been known to make mistakes with basic calculations and even hallucinate or invent fake data and numbers,” he said, adding: “The same applies for things like interest rates – AI can provide a helpful breakdown, but it can’t be trusted to give you accurate, real-time information.”
Higgins said an experiences adviser would assess a client’s financial circumstances in detail and help people understand how changes in rates or circumstances could impact options.
Higgins said consumers could not always get an accurate picture of their affordability using AI, without sharing data they might not feel comfortable putting online, which an adviser could handle securely.
He added: “AI may also miss certain options like deposit‑boosting schemes or equity loans, meaning buyers could miss out on savings that a qualified adviser would highlight. Indeed, buying a new home opens a range of mortgages [that] aren’t always available to consumers going direct to the lender.”