How recruiters can reduce bias in shortlisting
Most hiring bias does not happen in the interview room. It happens in the forty minutes somebody spends skimming a hundred CVs on a Tuesday afternoon, deciding in six seconds each who is worth a conversation. By the time you are interviewing, the pool is already shaped.
That makes shortlisting the highest-leverage stage in the whole process, and the one that usually gets the least structure. Below is what actually works to reduce bias in shortlisting, drawn from Acas guidance, GOV.UK’s employer actions and the UK research behind them.
Where bias actually enters shortlisting
It enters through proxies. Not through anyone deciding to exclude women, but through criteria that correlate with being a man and have nothing to do with whether someone can do the job.
Career gaps are the clearest example. GOV.UK’s guidance on reducing unconscious bias in CV screening notes that applicants showing periods out of work are consistently less successful than those without gaps, that the majority of UK returners are women, and that a study of STEM professionals found 53% of returners reported negative bias when applying for jobs because of a lack of recent experience.
Then there is continuous employment as a proxy for commitment, brand-name employers as a proxy for quality, and confident CV language as a proxy for competence. Each one filters out women at a slightly higher rate than men, and the effect compounds across a hundred applications.
The scale of what sits downstream is worth keeping in view. Of the 11,183 UK employers who filed gender pay gap figures for the 2025 to 2026 reporting year, 78% reported a median gap in favour of men, and women were a median 40.6% of the top pay quartile against 56.3% of the lowest. Shortlisting is not the whole cause of that. It is the earliest point you control.
Build the checklist before you read the CVs
Acas is unambiguous on this in its guidance on choosing who to interview. Turn every point in the job description and person specification into a checklist, assess each application against that checklist, and score it.
Two details matter more than people expect.
Write the checklist before applications arrive. Criteria written after you have seen the pool get bent, unconsciously, towards the candidates you already like. Agree them at the point you sign off the job advert.
Use a wide enough scoring range. Acas warns that a one to five scale often produces a pile of fours and fives that does not help you separate anyone. If everyone clusters at the top, the tie is broken by feel, which is where bias lives.
Separate essential criteria from desirable ones and hold the line on it. A long list of desirables that quietly functions as essential is one of the most common ways a shortlist narrows to people who look like the last person in the role. Our guide to writing job ads that attract women covers the same problem one step earlier.
Anonymise with care, and fix the CV format too
Removing names, addresses, photographs and dates of birth before screening is a reasonable default, and GOV.UK notes that anonymised CVs may improve women’s chances of being interviewed and hired. Most applicant tracking systems have a screening mode that does this.
It is not a complete answer, though, and GOV.UK says so directly: anonymisation is not a one-size-fits-all solution, and it can cut across targeted recruitment work aimed at building a more representative workforce. Check it does not conflict with what else you are doing.
The more interesting intervention is the format itself. GOV.UK’s guidance, drawing on Behavioural Insights Team research, recommends offering a CV template that lets candidates list experience in number of years rather than specific dates, so that a two-year gap for caring stops being the first thing a screener sees. The reported effect is an increase of around 15% in the chance of getting an interview or an offer. Pair years of experience with role-specific competencies and you get a better decision, not just a fairer one.
Two people, one written record, and a check on your AI
Acas advises that ideally two or more people should be involved in choosing who to interview, because a second reader reduces the risk of individual bias and unintended discrimination. In practice, the cheapest version of this is to have two people score the borderline pile independently before they discuss it. Discussing first collapses the second opinion into the first.
Record the reasons. Not for bureaucracy: a written score against named criteria is what lets you check your own process later, and it is what you would need if a rejected candidate ever raised a claim.
Then look hard at any automated screening you use. GOV.UK is explicit that AI CV screening tools learn from historical hiring data and may repeat the gender biases in it, and that automated tools often score employment gaps negatively, which disproportionately affects women. The legal position is unchanged by the technology: the employer is responsible for non-discriminatory recruitment under the Equality Act 2010, and you must review and adjust the outputs. The Department for Science, Innovation and Technology publishes guidance on responsible AI in recruitment worth reading before you switch anything on.
Measure the shortlist, and stay the right side of the law
Almost every employer measures who they hired. Very few measure who they shortlisted, which is where the loss actually happens.
Track the proportion of applicants who are shortlisted, interviewed and offered, broken down by sex, and where you can by sex combined with ethnicity or disability. Compare it against previous campaigns as a baseline. If women are 45% of applicants and 22% of the shortlist, you have found your problem, and no amount of interview training will fix it. Bear in mind that equality monitoring data is special category personal data under UK data protection law, so involve whoever owns data protection in your organisation before you start collecting it.
One boundary to hold. Under the Equality Act 2010 you can take positive action, including targeted outreach and, in the narrow circumstances of section 159, choosing a candidate from an under-represented group where candidates are as qualified as each other and it is a proportionate response rather than a blanket policy. What you cannot do is set quotas or select someone because of their sex regardless of merit. That is unlawful discrimination, and it also hands ammunition to everyone who wants to dismiss the work. Our practical guide to inclusive recruitment and how to build a balanced team when you are hiring go into both in more depth.
A checklist to reduce bias in shortlisting
If you do nothing else, do these five.
- Agree scored criteria from the person specification before applications open.
- Use a scoring range wide enough to separate strong candidates.
- Screen without names, dates of birth or photographs, and offer a years-of-experience CV format.
- Have two people score the borderline pile independently, and write down why.
- Report shortlist rates by sex, not just hire rates.
Frequently asked questions
What is the single most effective way to reduce bias in shortlisting?
Scoring every application against a written checklist that was agreed before applications were read. Acas recommends turning each point of the job description and person specification into criteria and scoring against them. It is unglamorous, it takes an extra half hour at the briefing stage, and it does more than any amount of unconscious bias training.
Does name-blind recruitment actually work?
It helps and it is not sufficient on its own. GOV.UK notes anonymised CVs may improve women’s chances at screening, while warning that anonymisation is not a one-size-fits-all solution and can conflict with targeted recruitment campaigns. Removing dates and reformatting experience by years of experience appears to do at least as much work as removing the name.
Can I use AI to shortlist candidates in the UK?
You can, but you remain responsible. GOV.UK warns that AI screening tools trained on historical hiring data may repeat existing gender bias and often penalise employment gaps. Under the Equality Act 2010 the employer is accountable for a discriminatory outcome regardless of which tool produced it, so you need to review outputs, test for disparate impact and follow the government’s responsible AI in recruitment guidance.
How many people should be involved in shortlisting?
Acas advises ideally two or more, because a second person reduces the risk of personal bias and unintended discrimination. Have them score independently before they compare notes, particularly on borderline applications.
Hiring women? Get RecruitHer early access, stronger signals and less inbox noise.
Sources: Acas, Recruitment: choosing who to interview. GOV.UK, Reduce unconscious bias in CV screening (Office for Equality and Opportunity). Department for Science, Innovation and Technology, Responsible AI in recruitment. Gender pay gap service, 2025 to 2026 reporting year, analysed by RecruitHer in August 2026. This is educational information, not legal advice.
Last reviewed: August 2026