product · 13 June 2026
Asking an AI about your mortgage? What it can and cannot tell you
General AI assistants are helpful for mortgage concepts, but cannot see your real finances or give regulated advice. Here is what to expect.
By Moxim Team
More and more people are opening a chat window late at night and typing something like: "Can I afford a £400,000 house?" It is a completely reasonable thing to do. The question is what you actually get back.
The honest answer is: it depends enormously on what the assistant is working with. Here is a plain account of what AI assistants are good at, where they fall short, and how that picture is starting to change.
What a general AI assistant does well
A good AI assistant can be genuinely useful at the education layer of the mortgage journey. That is a meaningful thing.
Ask it to explain what a loan-to-value ratio is, how a fixed-rate period works, or what "stress-tested affordability" means in plain English, and a decent assistant will do that clearly and patiently at any hour. There is no such thing as a stupid question when you are talking to a machine.
It can walk you through the typical stages of a mortgage application, explain the difference between a decision in principle and a formal offer, or help you understand why lenders care so much about your last three months of bank statements. It can help you compare the general shape of repayment versus interest-only, or explain what a Help to Buy scheme was and why it no longer exists.
This kind of conceptual grounding matters. Many people arrive at a broker's office uncertain of the vocabulary, let alone the process. If a chat assistant helps someone walk in feeling more prepared, that is a real service.
What a general AI assistant cannot do
Here is where the honest reckoning starts.
A general AI assistant has no access to your actual finances. It does not know what lands in your account each month, what you spend on subscriptions or childcare, or whether you have a credit card balance that has been sitting there for two years. When you type "I earn around £55,000 and spend about £1,800 a month," you are asking it to work with a rough self-report. It cannot verify that, and self-reported numbers are notoriously imprecise, not because people lie, but because most of us do not have an accurate mental ledger.
A general AI assistant also does not know current lender criteria. Affordability criteria change. Lender appetite shifts with the market, with funding conditions, with regulatory guidance. A model trained on data up to a certain point will reflect the world as it was then, not necessarily as it is now. Any specific number it gives you ("based on your income you might borrow up to X") is a pattern-matched estimate, not a calculation grounded in a lender's current rulebook.
And nothing a general AI assistant tells you constitutes regulated advice. It cannot, and should not. Mortgage advice is a regulated activity in the United Kingdom. A confident-sounding figure from a chat window is not a quote, not an offer, and not an eligibility result. It is, at best, an educated guess dressed up in fluent prose. That is worth knowing before you start making plans around it.
The shift that is beginning to happen
This is where the picture starts to change, carefully and with caveats.
The limitation described above, that assistants are working from guesses rather than facts, is a structural one. It comes from the fact that a general chat assistant has no connection to your real data. But that connection is technically possible, with your knowledge and consent.
There is an emerging open standard called the Model Context Protocol (MCP) that allows AI assistants to connect to external data sources in a structured, declared way. Think of it as a common language that lets an assistant ask a service, with your permission, for specific, factual information, rather than inferring it from whatever you type into a box.
The practical implication is this: an assistant that can, with your active consent, read your real income and spending through Open Banking, or access a readiness credential you own, can ground its answer in checkable facts rather than your best recollection. The answer to "where do I stand?" becomes a reflection of your actual position, not a generalisation about people who roughly resemble you.
This is not yet standard. Most assistants most people use today are still working from the general model described above. But the direction is clear.
Consent and control are not optional extras
Any version of this future worth building has to put the person in control. Connecting an AI assistant to your financial data is only valuable if you understand what it is reading, you have chosen to allow it, and you can withdraw that permission at any time.
The consent layer is not a footnote. It is the point. The goal is not to give lenders or brokers or anyone else a richer picture of you without your knowledge. It is to give you a richer, more accurate picture of yourself: one you can choose to share, with a broker, with a lender, with whoever you trust to help you make a good decision.
How Moxim is built for this
Moxim is building an Open Banking-powered mortgage-readiness layer: an illustrative read of your affordability, drawn from your real transaction data, and a portable credential you own and can take to any broker.
From the start, we have built the site to be machine-readable by design. We publish an llms.txt file and an MCP descriptor so that AI assistants can discover and accurately explain what Moxim is and how it works. In time, and with your explicit consent, assistants will be able to work from your real readiness data, drawn from your actual bank and HMRC records, rather than a guess.
We showed our work at the FCA Mortgages TechSprint 2025.
We are not going to oversell where we are. The credential is in development. The consent infrastructure is being built carefully. The goal is a future where you do not have to choose between a chatbot that guesses and a process that feels opaque, but we are honest that we are not there yet.
What exists today is a clear, illustrative walkthrough of how that readiness picture works, so you can see the shape of it before connecting your own data when early access opens.
Everything Moxim shows today is illustrative and educational, never a quote, an offer, or a lending decision. Moxim is not yet PRA-authorised or FCA-regulated. If you would like to see where you stand, see the experience.