Discrimination in Hiring: What It Is and How to Prevent It (2026 Guide)

TL;DR

  • Most discrimination in hiring is designed in, not intended. Unnecessary requirements, inconsistent interviews and inaccessible steps filter people out before anyone makes a conscious choice.
  • In the UK, the Equality Act 2010 protects job applicants from the first job advert to the job offer. Direct discrimination is almost never lawful; indirect discrimination is lawful only if you can show it’s a proportionate means of achieving a legitimate aim.
  • Structure is the best defence. The same questions, the same scoring rubric and recorded reasons for every candidate make decisions fairer and far easier to defend at an employment tribunal.
  • Reasonable adjustments and accessibility aren’t optional extras. Offer them up front, in every step.
  • AI can reduce or amplify bias. Use tools that are explainable and audited, and keep a human accountable for the final decision.

Discrimination in hiring happens when a candidate is treated less favourably, or put at a disadvantage, because of a protected characteristic such as age, race, sex or disability. It rarely looks like a manager saying “we don’t hire people like you”. It looks like a “recent graduate” in a job advert, an interview held in a building with no step-free access, or an online application form that doesn’t work with a screen reader.

A fair hiring system protects candidates and your organisation. Beyond compliance, a consistent, transparent process reduces disputes, builds trust with applicants and leads to better hires. This guide explains what the law says (with the UK’s Equality Act 2010 as the main reference, plus Australian and US pointers), where discrimination hides in everyday recruitment, and the practical steps that prevent it without slowing your team down.

What “discrimination in hiring” means in 2026

In Great Britain, the Equality Act 2010 makes it unlawful for an employer to discriminate against job applicants as well as employees. It covers how you advertise, how you decide who to interview or offer a job to, and the terms you offer. Applicants don’t need to have worked for you to bring a claim.

The Act protects people based on the following protected characteristics:

  • age
  • disability
  • gender reassignment
  • marriage and civil partnership
  • pregnancy and maternity
  • race (including colour, nationality and ethnic or national origin)
  • religion or belief
  • sex
  • sexual orientation

Northern Ireland has separate legislation with broadly similar protections. (This section summarises the UK position for orientation only. It isn’t legal advice.)

The six types of discrimination in recruitment

TypeWhat it meansHiring example
Direct discriminationTreating a candidate less favourably because of a protected characteristicRejecting an applicant because she mentioned her civil partnership, or because a manager wants “someone younger”
Indirect discriminationApplying a rule to everyone that puts people who share a particular protected characteristic at a particular disadvantage, without justificationRequiring 10 years’ experience for an entry-level role, which screens out younger candidates
Discrimination arising from disabilityA disabled candidate is treated unfavourably because of something connected to their disabilityMarking a candidate down for limited eye contact linked to autism
Failure to make reasonable adjustmentsNot making reasonable changes so a disabled candidate isn’t at a substantial disadvantageRefusing extra time on a timed test for a candidate with dyslexia
HarassmentUnwanted conduct related to a protected characteristic that violates dignity or creates a hostile environmentJokes about a candidate’s accent or religion during an interview
VictimisationTreating someone badly because they’ve complained about discrimination, or supported someone who hasBlacklisting a candidate who raised a discrimination concern about a previous application

Victimisation protection applies to anyone who makes a complaint in good faith. It’s lost only if someone knowingly gives false information.

Can discrimination in recruitment ever be justified?

Sometimes, but the bar is high and it depends on the type.

  • Indirect discrimination. You can justify indirect discrimination if the requirement is a proportionate means of achieving a legitimate aim. “Must hold a valid HGV licence” for a lorry driver is easy to justify. “Must be available for any shift at no notice” for a part-time retail role, which disproportionately affects carers (most of whom are women), is much harder.
  • Direct discrimination. You generally can’t justify direct discrimination. The one exception is age: direct age discrimination can be lawful if it’s a proportionate means of achieving a legitimate aim, but cost-saving alone won’t do it. Turning down candidates over a certain age because they “won’t stay long” won’t pass the test.
  • Occupational requirements. In limited circumstances, a role can require a particular protected characteristic where it’s genuinely necessary for the job – for example, a women’s refuge counsellor role that must be filled by a woman. The requirement must be objectively justified, and you need to recheck it each time you recruit for the role.
  • Positive action. You can encourage applications from groups that are under-represented or disadvantaged – for example, a job advert stating that you particularly welcome applications from people with a particular characteristic. In a genuine tie-break between two candidates of equal merit, you may prefer the candidate from the under-represented group. Positive discrimination (hiring someone because of a protected characteristic, such as a quota) is generally unlawful.
  • Disability. Non-disabled people can’t claim disability discrimination, so you can run schemes for disabled applicants. Requiring a particular disability is riskier and needs an occupational requirement or positive action basis.

Australia and the US

Australia: the Fair Work Act’s general protections cover prospective employees, so refusing to employ someone for a discriminatory reason can be “adverse action”. Federal laws cover sex, race, disability and age discrimination, and every state and territory has its own anti-discrimination law.

United States: Title VII of the Civil Rights Act (race, colour, religion, sex and national origin), the Age Discrimination in Employment Act (40 and over) and the Americans with Disabilities Act are the main federal protections. The Equal Employment Opportunity Commission (EEOC) enforces them, and many states and cities add their own rules, such as salary history bans. US law uses “reasonable accommodation” where UK law says “reasonable adjustments”. 

Wherever you operate, align your recruitment processes with local law and take advice on edge cases. Nothing below is legal advice; it’s a practical blueprint for removing discrimination in recruitment from day-to-day hiring.

Where discrimination appears in the hiring process

The most common risks come from design, not intent. Small choices create barriers that filter people out before anyone speaks to them.

Job descriptions and job adverts

Ambiguous criteria, inflated degree requirements and gendered language narrow the candidate pool unnecessarily. “Culture fit” language can mask bias against people from different backgrounds. If a requirement isn’t tied to the job’s essential tasks, leave it out.

In the UK, the Equality and Human Rights Commission can take action over a discriminatory job advert even if no individual has complained.

Wording to avoidWhy it’s riskySafer alternative
“Recent graduate”, “young, energetic team”, “digital native”Signals a preference for a particular age group – potential direct or indirect age discrimination“Entry-level role – full training provided”
“Waitress”, “salesgirl”, “handyman”Gendered job titles risk sex discrimination“Server”, “sales assistant”, “maintenance operative”
“Native English speaker”Ties the requirement to national origin – a race discrimination risk“Fluent written and spoken English” (only if the role needs it)
“10+ years’ experience”Can disadvantage younger candidates where length of service isn’t a real requirementDescribe the skills and outcomes the role needs
“Must be physically fit”Can exclude disabled candidates without considering adjustmentsName the actual task (e.g. “lifting boxes up to 10kg”)

Sourcing and word-of-mouth recruitment

Relying on a single network reproduces sameness. Word-of-mouth recruitment inside homogeneous teams can favour people of a particular race, school or background. Use several channels to widen reach so certain groups aren’t excluded before the selection process starts.

Application forms and early screens

Requests for photos, dates of birth or personal details unrelated to the role create risk. So does rejecting qualified candidates because of CV gaps or educational background rather than capability. Check that your application form works on mobile and with assistive technology, and that candidates can ask for a different format.

Interviews and decisions

Unstructured conversations invite unconscious bias, confirmation bias, affinity bias and age bias. Questions about religious beliefs, plans for children, marital status or health are never appropriate. Harassment during interviews can be verbal or physical, and both create a hostile environment. Keep notes on objective criteria, not “gut feel”. For a list of interview questions to steer clear of, see our examples of unfair hiring practices.

Pay and job offers

In the UK, you generally can’t ask about a candidate’s health or disability before making a job offer, except in narrow cases such as checking whether they need reasonable adjustments for the selection process or can carry out a task that’s intrinsic to the role. Make any health questionnaire part of a conditional offer to the successful applicant.

Anchor offers to a pay band and objective factors, not a candidate’s previous salary, which can carry forward past pay gaps. Document your rationale. People doing equal work must receive equal pay regardless of sex, and inconsistent offers to other protected groups create their own risk.

AI discrimination in hiring: risks and realities in 2026

Technology can help or harm, depending on design and oversight. AI bias in hiring typically comes from three sources:

  • Training data that mirrors past patterns, such as over-valuing certain universities or penalising names associated with particular ethnic groups.
  • Model design and proxies, such as postcode standing in for race or socio-economic status.
  • Poor governance: unclear consent, vague purpose and no audit trail.

Using a tool doesn’t move the legal responsibility. If an algorithm screens out a group at a particular disadvantage, the employer still has to justify it under the Equality Act, just as it would for a human decision. The fix isn’t “no tech”; it’s good tech with guardrails:

  • Use AI systems that publish their methods, allow review and support challenges.
  • Ask for bias testing across protected characteristics, and regular re-audits.
  • Keep sensitive data out of the model and strip irrelevant attributes.
  • Keep human judgement in the final decision. Treat AI as decision support, not the decision maker.

Sapia.ai is built on these principles: structured, mobile-first chat interviews, explainable scoring aligned to your rubric, and visible pass-through by stage so you can spot disparities quickly, while hiring managers remain accountable.

How to prevent discrimination in the hiring process

Preventing problems up front is far simpler than dealing with a complaint later. Here’s a practical sequence any recruitment team can run.

1) Write inclusive job descriptions tied to the work

  • Describe outcomes, not personalities.
  • Keep requirements short and essential; remove degree requirements that don’t affect someone’s ability to do the job.
  • Use gender-neutral language, and test every requirement against the question “could I justify this if challenged?”
  • Tell candidates how to ask for reasonable adjustments, and state the pay range and selection steps.

2) Broaden sourcing and remove barriers

  • Mix channels: community groups, targeted job boards, returner programmes and networks that reach under-represented groups.
  • Use positive action lawfully: say you welcome applications from groups that are under-represented in the role, but keep the role open to everyone.
  • Don’t exclude people through scheduling alone (school runs, religious festivals, shift workers who can’t attend daytime interviews).

3) Standardise early screens

  • Anonymise early reviews so names, ages and schools aren’t visible. Our guide to blind hiring covers how to set it up.
  • Replace ad-hoc phone screens with a short, structured first step. Structured interviews, where everyone answers the same questions scored against behavioural anchors, are one of the most reliable ways to reduce bias.
  • Add a small, job-relevant task (a draft customer message or a prioritisation exercise) so evidence counts for more than similarity.

Sapia.ai supports this with a mobile, structured first interview scored against your criteria, which then books live steps automatically. It’s one way to reduce hiring bias using AI interview software without losing the human experience.

4) Train hiring managers and panels

  • Run short, job-specific sessions on unconscious bias, affinity bias, confirmation bias and age bias. Our unconscious bias training guide has a starting point.
  • Use diverse interview panels to bring different perspectives into the room.
  • Share a one-page rubric for each role, so everyone scores the same behaviours in the same way.

Training raises awareness, but it can’t remove bias from people. Good practice is to build processes that limit the room for bias, rather than relying on training alone.

5) Make every step accessible and offer reasonable adjustments

  • Ask every candidate, at every stage, whether they need adjustments – and make it easy to say yes.
  • Common adjustments include extra time, a different format (text instead of video, or a phone call instead of an online test), screen-reader-compatible forms, rest breaks, and step-free or quiet interview venues.
  • Tell candidates who they’ll meet and how long each step takes.
  • Keep notes factual and tied to the criteria. Avoid shorthand like “not a fit”.
  • Keep medical information separate from hiring decisions.

6) Control pay and offers

  • Anchor every job offer to the role’s band and documented evidence of skills; don’t use prior pay as the benchmark.
  • Check pay parity regularly, for new hires and against existing employees in similar roles.
  • Document any exceptions explicitly.

7) Govern your AI hiring tools

  • Ask vendors for model documentation, bias audit reports and explainability.
  • Require periodic re-testing, and log changes and outcomes.
  • Define who can access candidate data, for what purpose and for how long.
  • Give candidates a clear route to request human review.

8) Give feedback and close the loop

  • Send every candidate an outcome, not just the successful candidate.
  • Where you can, offer constructive feedback tied to the criteria. Candidates who understand a decision are less likely to assume it was discriminatory.
  • Give candidates a clear, safe way to raise a concern.

How to detect and measure discrimination

You can’t fix what you can’t see. Build light-touch measurement into everyday work. For a deeper walkthrough, see bias in hiring: 5 ways to detect and remove it.

Metrics that matter

  • Representation and pass-through by stage: applied → screened → interviewed → offered → hired, broken down by lawfully collected, self-reported data (for example race, sex, age group and disability).
  • Time to first interview and time to offer for each group.
  • Offer rates, pay parity and early retention (30 and 90 days).
  • Candidate pulse: “Was the process clear and fair?”

If a group starts strong at application but drops sharply at interview, the problem is that step, not the pipeline. That’s what indirect discrimination looks like in data: a neutral-looking step that puts one group at a particular disadvantage. Change one variable (questions, panel make-up, scoring guidance), re-measure and document what you did.

Reviewing decisions

  • Keep short decision logs tied to criteria. “Scored 2/5 on customer recovery” is useful; “not our type” isn’t.
  • Hold a regular 20-minute review to scan outliers.
  • Make it easy for candidates and employees to raise concerns early and safely.

How to avoid discrimination in hiring: practical examples

Here’s how to turn policy into behaviour:

  • Scenario – interview: A panel member prefers candidates who mirror their own university path. Fix: hide education on the score sheet; focus on work samples and consistent behavioural questions.
  • Scenario – sourcing: Roles are filled mainly through internal referrals and the team lacks diversity. Fix: redesign referrals to prompt breadth, widen channels and track representation by source.
  • Scenario – AI screening: A tool downgrades female candidates who attended women’s colleges because of skewed training data. Fix: vendor audit, re-sample historical data, block education as a proxy, add human review.
  • Scenario – job advert: A graduate scheme advert asks for “recent graduates, 2024 or 2025”. This excludes people who graduated earlier or later in life. Fix: describe the role as entry-level and open the scheme to anyone with a degree, whenever they graduated.
  • Scenario – interview venues: All final interviews take place on the second floor of an office without a lift. Fix: check accessibility when booking interview venues, and ask every shortlisted candidate if they need adjustments.
  • Scenario – language requirement: A warehouse role asks for “excellent written English” although the job involves no writing. Fix: remove it, or replace it with the level the job actually needs (e.g. “able to follow safety instructions in English”).

Each fix gives every candidate a fair chance without slowing the team down.

What happens if a candidate brings a discrimination claim?

Most guidance stops at prevention. It’s worth knowing what a claim actually involves, because it shows why good records matter. (UK process summarised for orientation only. )

  1. Early conciliation. Before making a discrimination claim to an employment tribunal, the candidate must usually notify Acas, which offers free conciliation to try to settle the dispute.
  2. Time limit. Claims generally have to start within three months less one day of the act complained about, with the clock paused during early conciliation.
  3. Burden of proof. If the candidate shows facts from which the tribunal could conclude discrimination happened, the burden shifts to the employer to show it didn’t. This is where scoring sheets, decision logs and consistent questions become your evidence.
  4. Compensation. There’s no cap on compensation for discrimination. Awards can cover financial loss (such as lost earnings from a job they should have been offered) and injury to feelings.

Recruitment discrimination claims are rarely won or lost on intent. They turn on whether you can show the decision was based on job-related criteria applied consistently to every candidate.

Documentation and compliance: getting the basics right

Good documentation is your best evidence of fair, job-related decisions.

  • Keep clean records of job adverts, criteria, shortlists, scores, adjustments offered and job offers, in line with your data retention policy.
  • Keep medical and genetic information out of the decision stream.
  • Make sure managers know how to escalate concerns to HR.
  • Review your process at least once a year as good practice, and whenever you change a tool or a step, to ensure recruitment processes still treat every candidate consistently.

Where Sapia.ai fits

Most discrimination in recruitment comes from inconsistency at the front of the funnel: different questions, different reviewers and CV signals that stand in for ability. That’s the part Sapia.ai is designed to address.

What it does: Sapia.ai’s Chat Interview gives every applicant the same structured, text-based interview, scored against consistent criteria, with an explainable shortlist for hiring managers. Customer results include:

  • LNER removed CV uploads and video interviews from its process. It reported “a consistent 30% ethnic minority representation through all hiring stages”, a 97% completion rate and a 57% decrease in time to hire. (LNER case study)
  • Costa Coffee describes its Sapia.ai interview as “untimed, mobile-first, and blind to gender, age, ethnicity, or disability”, with “compatibility with screen readers and voice-to-text tools for accessibility” and diversity dashboards in Discover Insights to monitor representation. It reports an 89% completion rate and 8.8/10 candidate satisfaction from 180k interviews. (Costa Coffee case study)
  • Holland & Barrett: “All candidates receive personalised feedback with coaching tips”, and the retailer reports a 9/10 candidate satisfaction rating from 470k interviews. (Holland & Barrett case study)

These results show representation maintained and accessibility built in. They don’t show bias eliminated, and no tool can promise that.

What it doesn’t do: Sapia.ai won’t rewrite your job adverts, fix a referral culture, check your pay bands or replace legal advice. It doesn’t make the hiring decision – your managers do – and building a diverse workforce still depends on what happens after the shortlist.

If you want to see how a structured, interview-first process would work on your highest-volume role, book a Sapia.ai demo.

FAQs

What is discrimination in hiring?

Discrimination in hiring is treating a job applicant less favourably, or putting them at a disadvantage, because of a protected characteristic such as age, disability, race, religion, sex or sexual orientation. It can happen at any stage, from the job advert to the job offer, and it can be unintentional – for example, a requirement that screens out one group without a good reason.

What is the difference between direct and indirect discrimination?

Direct discrimination is treating someone less favourably because of a protected characteristic, such as rejecting a candidate because of their religion. Indirect discrimination is applying the same rule to everyone when it puts a group at a disadvantage, such as an unnecessary experience requirement that excludes younger candidates. Under the Equality Act 2010, indirect discrimination can be justified as a proportionate means of achieving a legitimate aim; direct discrimination generally can’t, except for age.

What are the nine protected characteristics under the Equality Act?

The nine protected characteristics are age, disability, gender reassignment, marriage and civil partnership, pregnancy and maternity, race, religion or belief, sex, and sexual orientation. Job applicants are protected on all of them, as well as employees.

What reasonable adjustments should employers offer during recruitment?

Common reasonable adjustments include extra time on assessments, information or tests in a different format, screen-reader-compatible application forms, step-free interview venues, rest breaks and the option to bring a support worker. The simplest approach is to ask every candidate at every stage whether they need any adjustments, and to act on what they tell you.

Is positive action the same as positive discrimination?

No. Positive action is lawful in the UK: you can encourage applications from under-represented groups and, where two candidates are of equal merit, prefer the one from the under-represented group as a tie-break. Positive discrimination – choosing someone because of a protected characteristic regardless of merit, such as a quota – is generally unlawful.

Can a job applicant take an employer to an employment tribunal?

Yes. Under the Equality Act 2010, job applicants can bring a discrimination claim even though they never worked for the employer. They usually have to contact Acas for early conciliation first, and claims must generally be started within three months less one day of the discriminatory act. Compensation is uncapped and can include financial loss and injury to feelings.

About Author

Laura Belfield
Head of Marketing

Get started with Sapia.ai today

Hire brilliant with the talent intelligence platform powered by ethical AI
Speak To Our Sales Team