Before the Keys: The Quiet AI Reading Your UK Rental Application

You find a flat on a Tuesday, view it on a Thursday, and by Friday the letting agent wants your payslips, bank statements, employer’s details and a guarantor’s email address. Somewhere in that pile, a piece of software may already be forming an opinion about you. Not a human with a mug of tea and a highlighter — a referencing tool that checks documents, scores affordability, cross-references addresses and flags anything that looks unusual.

This is not science fiction, and it is not necessarily bad. Referencing has always involved judgement. The difference is that some of that judgement now happens through models, dashboards and risk scores. The tenant never sees the workings. The landlord often sees only a green, amber or red badge.

What the software is actually doing

There is no single ‘tenant score’ used across the UK. Letting agents and landlords buy different products, and those products do different things. Some verify identity documents. Some read payslips and bank statements, often through open banking, to estimate whether rent will be affordable. Some contact previous landlords and employers. Some check credit files, address history and public records. A few try to predict ‘risk’ using patterns in the data they hold.

The useful way to think about it is as a second reader. It is not usually making the final decision. It is sorting, flagging and summarising — telling a human that a document looks altered, that income does not match the bank statements, or that the applicant’s address history has a gap. That can speed things up. It can also introduce errors that are hard to spot, especially when the output is a simple score with no explanation.

The data trail you leave behind

When you apply, you are usually handing over more than a form. You may be asked to connect your bank account so a tool can scan transactions. You may upload PDFs that contain your employer’s name, your salary, your spending habits and your family arrangements. You may give permission for a credit check. You may provide a guarantor’s details, which means their data enters the process too.

All of this sits under UK data protection law. The letting agent or landlord is normally the data controller; the referencing company is often a processor, though it depends on the contract. You should be told what is collected, why, who it is shared with and how long it is kept. That information belongs in a privacy notice, not in a passing comment at the viewing.

Where bias can sneak in

Risk models learn from the past. If a letting agent’s historical decisions favoured applicants with certain postcodes, job types or family setups, a model trained on that data can repeat the pattern — even if nobody intended it. Protected characteristics under the Equality Act 2010, such as disability, race, sex or religion, can be wrapped up in proxies: the type of employment, a gap in work history, a benefit payment, an address that has been associated with arrears.

That does not mean every AI referencing tool is discriminatory. It means the risk is real enough that humans need to stay in the loop. A score should be a prompt for a conversation, not a substitute for one. If an applicant has a stable income from a pension, a disability benefit or a combination of part-time work and savings, a crude affordability model may miss the point. A human reviewer should be able to see that.

Your rights, in plain English

Under the UK GDPR, you can ask for a copy of your personal data through a subject access request. You can ask for inaccurate data to be corrected. You can ask how a decision was made, though the answer may be limited by trade secrets and the complexity of the model. You are entitled to meaningful information about the logic involved in solely automated decisions that have a legal or similarly significant effect on you. If such a decision is made, you can usually ask for human intervention, express your point of view and contest it.

In practice, many referencing systems are ‘AI-assisted’ rather than fully automated. That matters. A human signing off the decision changes the legal picture. But it should be a real review, not a rubber stamp.

There is also the question of fees. In England, most tenant fees are banned under the Tenant Fees Act 2019, including referencing fees charged to tenants. If you are asked to pay a separate ‘AI screening’ or ‘risk assessment’ charge, ask why it is being passed to you. Rules differ across the UK, so check the position in Scotland, Wales and Northern Ireland.

Questions worth asking before you pay a holding deposit

You do not need to be confrontational. A short email is enough.

• Is any automated or AI-assisted tool used in referencing or tenant selection?
• Which company provides it, and what data does it receive?
• Will a human review the file before a decision is made?
• Can I see what information was used and how any score was reached?
• How long will my data and my guarantor’s data be kept?
• What is the process if I think something is wrong?

Landlords and agents should be able to answer these questions without treating them as an accusation. If they cannot, that is useful information too.

If you are the one doing the letting

Using software to triage applications is reasonable. Letting it decide is not. Before adopting a tool, ask for a data protection impact assessment, a clear explanation of what the model measures, and evidence that it has been tested for bias. Keep a human decision-maker accountable. Record why an application was declined. Do not retain documents longer than you need them. And remember that a tenant who feels processed like a parcel is less likely to trust you when something goes wrong.

The bot may never blink. But the process still belongs to people — and in the UK, it still has to follow the rules.