Key takeaways
- With around ten applicants per available rental property in August 2026, the pressure to automate shortlisting is real[1].
- The Equality Act bars discrimination in who is offered a home, on what terms and in how people seeking premises are treated[2].
- In England it is now illegal to discriminate against applicants who have children or receive benefits, with civil penalties of up to £7,000[3].
- Significant automated decisions need safeguards, including human intervention and a way to contest them, so keep a person genuinely in the decision[4].
Why automated shortlisting is tempting
Popular rental homes attract far more interest than one negotiator can handle. In August 2026 Propertymark letting branches averaged 121 new applicant registrations and around ten applicants for every available property[1]. Software that ranks applicants, flags 'high risk' profiles or decides who is invited to view promises to save hours and make choices look objective.
The problem is that a tool can be consistent and still unfair. If it learns from past decisions, it repeats past patterns. If it uses data that stands in for a protected characteristic, it can exclude groups without anyone intending it. And because letting agents acting for landlords can be held liable under the Renters' Rights Act enforcement regime[5], the agency carries the risk of the tool's choices.
What applicants think about machine decisions
People are wary of computers making decisions about them. In the government's public attitudes survey, 91% of UK adults said they were concerned about important decisions being made by computers without human input[6]. Few decisions feel more important to a renter than whether they get a home. An applicant who believes they were rejected by an algorithm, without explanation, is more likely to complain and less likely to accept the outcome.
- Concerned 91%
Source: Department for Science, Innovation and Technology (GOV.UK), 2024[6]
Show the data
| Item | Value |
|---|---|
| Concerned | 91% |
The legal frame in England
Three sets of rules overlap. Under section 33 of the Equality Act 2010, a person with the right to dispose of premises must not discriminate as to the terms on which they offer the premises, by not disposing of them, or in their treatment of people seeking premises[2]. That covers protected characteristics such as race, disability, sex, religion or belief, sexual orientation, and pregnancy and maternity, and it covers indirect discrimination, where a neutral-looking rule puts a group at a disadvantage without justification.
The Renters' Rights Act adds specific protection: it is illegal for landlords and agents to discriminate against prospective tenants in receipt of benefits or with children. Landlords can still assess affordability, but not on the basis that someone has children or receives benefits, and breaches can bring civil penalties of up to £7,000, with repeat penalties for continued breaches[3].
UK GDPR then governs the automation itself. Where a significant decision is based solely on automated processing, the controller must give the person information about it, let them make representations, let them obtain human intervention and let them contest the decision[4].
How bias gets into a tool
| Data the tool uses | What it can stand in for | Safer approach |
|---|---|---|
| Current postcode | Race or ethnicity in some areas | Do not use location as a scoring factor |
| Name or language | Race, nationality or religion | Exclude entirely from scoring |
| Employment gaps | Pregnancy, maternity, disability or caring | Assess current affordability instead |
| Benefit income marked as risk | Benefit receipt, now protected in England | Count benefits as income like wages |
| Household size | Children in the household | Use only for legal occupancy limits |
| Thin credit file | Age, recent arrival in the UK | Allow guarantors and alternative evidence |
| Past landlord decisions | Whatever biases those decisions held | Do not train on historical acceptances |
None of these needs to be intended. A tool that downgrades applicants with employment gaps may simply be picking up mothers returning from maternity leave or people recovering from illness. That is why testing outcomes matters as much as checking inputs.
Testing a tool for bias
List every input
Ask the supplier for a full list of data the tool uses and how each affects the score or decision.
Remove obvious proxies
Switch off location, name, nationality and any field that indicates children or benefits, unless there is a lawful, specific need.
Test with matched profiles
Create fictional applicants identical except for one factor, such as benefit income or a child in the household, and compare the results.
Compare real outcomes
Where you hold the data lawfully, compare acceptance rates for groups over a quarter and investigate large gaps.
Check affordability rules
Confirm that benefits, maintenance and pension income count in the same way as wages.
Document and repeat
Record the tests, the findings and the changes made, and repeat after any update to the tool.
Human review that actually counts
A person clicking approve on every automated result is not meaningful review. For the safeguards to work, the reviewer needs to see why a score was given, have the authority and time to overrule it, and actually do so when the evidence supports it. Track how often reviewers change the tool's recommendation; if the answer is never, the review is probably a formality.
- Show reviewers the reasons behind any refer or decline, not just the score.
- Give applicants a clear route to provide more information, such as a guarantor or proof of a new job.
- Tell applicants in plain words that automated tools are used and that a person makes the final decision.
- Record the reviewer's reason when they agree with or overrule the tool.
Treat every applicant the same way
Consistency is the best protection. Ask every applicant the same questions, in the same order, and apply the same written affordability policy. Right to rent checks must be carried out on all adults aged 18 or over who will live in the property[7], so they must never be applied selectively to people who seem foreign. Decide in advance how you choose between several applicants who all pass, for example in the order complete applications arrived, and record that rule.
Avoid soft filters that slip into adverts and automated replies, such as 'ideal for professionals', 'no DSS' or 'suits a couple'. They signal exclusion even when no formal rule exists, and generated text often repeats them from older adverts.
Where automation is reasonable
None of this means avoiding technology in lettings. Much of the application process is administrative and benefits from automation, as long as the tools move every applicant along the same path and leave the choice to a person.
- Acknowledging every enquiry and application with the same information and next steps.
- Offering viewing slots in the order enquiries arrive, using the same rules for everyone.
- Collecting documents and chasing referees and employers on a fixed schedule.
- Checking that an application is complete before a person reviews it.
- Producing a summary of each complete application in a standard format for the decision-maker.
The line to hold is simple: tools may organise, remind and summarise; people decide who is offered the home, and can explain why.
Questions for a referencing or scoring supplier
- Which inputs does the tool use, and can we switch any of them off?
- Has the tool been tested for bias, by whom, and can we see the results?
- How are benefits, maintenance and irregular income treated in affordability?
- Does the tool make a decision, or only a recommendation for a person?
- What explanation can we give an applicant who asks why they were declined?
- Who is the controller for applicant data, and what does the processing agreement say?
Worked example: fixing a scoring rule
Example: a fictional agency's referencing tool marks applicants whose income includes universal credit as 'refer', which staff have been treating as a soft decline. A matched-profile test shows two fictional applicants with identical total income and rent history: the one paid partly through universal credit is referred, the one paid entirely through wages passes.
Example, continued: the agency asks the supplier to treat benefit income as income in the affordability calculation, retrains staff that a refer outcome needs a person's review with reasons, and reviews the last three months of referred applicants. It contacts two applicants whose applications were wrongly declined to apologise and invite them to apply for current vacancies.
Records to keep
- The written affordability and selection policy, with version dates.
- The impact assessment for the tool and your bias tests.
- For each property: who applied, in what order, the outcome and the reason.
- Reviewer notes for every refer or decline.
- Any applicant challenge and how it was resolved.
For the wider data protection checks, see data protection checks before using AI tools, and for the specific rules on children and benefits, our guide to discrimination rules in lettings.
Frequently asked questions
Can letting agents use AI to choose tenants?
They can use tools to support decisions, but the agency remains responsible for fairness. Significant decisions based solely on automated processing need safeguards, so the safest approach is a recommendation reviewed by a person who can overrule it.
Is it legal to reject tenants on benefits in England?
No. Landlords and agents must not discriminate against prospective tenants because they receive benefits or have children. Affordability can still be assessed, but benefit income must be treated fairly.
How can a neutral rule be discriminatory?
If a rule that applies to everyone puts a group with a protected characteristic at a particular disadvantage and cannot be justified, it can be indirect discrimination. Scoring on postcode or employment gaps are common examples.
What should we tell an applicant who is declined?
Give the main reason in plain words, explain how they can provide more information or ask for a review by a person, and keep a record of what you told them.
References
Propertymark (2026). Housing Insight Report, August 2026 (opens in a new tab) ↑
legislation.gov.uk (The National Archives) (2010). Equality Act 2010, section 33: Disposals, etc. (opens in a new tab) ↑
Ministry of Housing, Communities and Local Government (GOV.UK) (2025). Guide to the Renters' Rights Act (opens in a new tab) ↑
legislation.gov.uk (The National Archives) (2026). UK GDPR, Article 22C: Safeguards for automated decision-making (opens in a new tab) ↑
Ministry of Housing, Communities and Local Government (GOV.UK) (2026). Enforcement measures for landlords: Renters' Rights Act 2025 (opens in a new tab) ↑
Department for Science, Innovation and Technology (GOV.UK) (2024). Public attitudes to data and AI: Tracker survey (Wave 4) report (opens in a new tab) ↑
GOV.UK (2026). Checking your tenant's right to rent: Who you have to check (opens in a new tab) ↑