Summary
Hiring algorithms are increasingly used to screen candidates at the top of the recruitment funnel. While they can help employers manage high volumes of applications, they may also filter out qualified candidates whose careers do not match narrow patterns.
Lucy Standing, psychologist, careers expert and founder of Brave Starts, argues that algorithmic hiring needs stronger HR governance. The risk is particularly acute for older workers, career returners, carers and generalists whose experience may be valuable but difficult for automated systems to recognise.
For HR leaders the key question is whether the organisation understands what its hiring algorithms are optimising for, who they are excluding and whether human judgement still has a meaningful role in the process.
You advertise a role. Within minutes, hundreds apply. For a well-known brand, that can tip into the thousands. The LinkedIn Easy Apply button and AI assisted applications have opened the floodgates and it seems there is no closing them.
What AI has done is collapse the time cost of applying. Where once a job seeker had to ration their energy and invest real effort into each application, that calculus no longer applies. Consequently, people can apply everywhere, for everything, whether they perceive a fit or not. We’ve lost the value of a candidate doing some of the self selection legwork.
On the recruiter side, the incentives are equally distorted. With clients mainly focused on ‘how quickly can you fill a role and how cheaply can you do it, recruiters are not incentivised to invest time and energy on the candidate experience. To compete and win, they have to be faster and cheaper. Equally, there is no consequences for unethical behaviour. Job posts can be left live indefinitely. Ghost jobs (roles advertised with no genuine intention to hire) are a quiet feature of the landscape. Recruiters who don't get clients to advertise to an audience of zero – so its in their interests for their jobs board to look enticing enough to submit their CV.
The result is a hiring ecosystem where neither side behaves honestly: not because people are dishonest, but because the system has made honesty irrational. Candidates bulk-apply; employers ghost. AI generates the applications; algorithms screen them out. Volume goes up. Signal goes down. In the race to the bottom of the barrel the right person never makes it to the end.
What happens when hiring algorithms screen for patterns rather than potential?
Most large employers rely on some form of algorithmic screening at the top of the hiring funnel. A candidate applies, their CV is parsed and scored, and within seconds they are either surfaced or silently buried. The logic seems defensible: filter for keywords, rank by match percentage, apply experience thresholds. What could possibly go wrong?
The problem is that these systems don't evaluate talent. They evaluate pattern-matching and genuine talent has an irritating habit of not fitting patterns.
Consider who gets removed before any human sees their name: the operations leader who spent eighteen months building their own business before it folded; the mum returning to work after a 15 year career break; the senior professional who took off 18 months to care for an ageing parent; the exceptional generalist who has never held the precise job title your algorithm is searching for.
These are not edge cases. Research from Harvard Business School and Accenture identified what they termed "hidden workers" – millions of qualified candidates rendered invisible to employers not because of a skills deficit but because automated systems cannot recognise them. Eighty-eight per cent of employers in the study acknowledged their screening tools were filtering out viable candidates. Most had no plan to address it.
Why older workers are vulnerable to algorithmic hiring bias
There is one group that algorithmic screening is particularly effective at removing, and one that most HR leaders would rather not discuss openly: older workers.
The bias rarely operates explicitly. Instead it works through proxies. Systems that weight for "digital nativity" disadvantage those who predate the tools but can master them quickly; CV parsing that struggles with careers spanning multiple decades; job titles that have changed, industries that have merged; skills continuously reinvented defaults to filtering out – not filtering in.
The legal exposure is real. In February 2026 a woman who had worked for PricewaterhouseCoopers for more than 40 years settled a case of age and disability discrimination against the firm for £150,000, supported by the Equality Commission for Northern Ireland. Among the allegations: a senior colleague questioned her knowledge of new technology and asked whether training was "something she was interested in at her age." She was told that given her length of service, she was simply working towards her pension. Following her formal grievance, she faced unjustified performance criticisms despite consistently strong reviews. Organisations that allow algorithmic systems to embed the same assumptions; that long tenure signals diminishing value, that unfamiliarity with recent tools signals incapability - are building identical liability into their recruitment infrastructure at scale, and with far less visibility than a human conversation.
In the UK workers aged 50 and over represent a third of the working-age population. The data on how this group is experiencing the job market is getting harder to ignore. The Brave Starts survey, which tracks the experiences of people over 45 navigating their future careers, has recorded a significant shift: in 2020 and 2021, 41% of respondents said they had experienced or feared discrimination because of their age. In the most recent responses (from 2025 and 2026), that figure has risen to 57%. The problem is not fading – it is getting worse.
How hiring algorithms inherit bias from past recruitment decisions
When Amazon scrapped its AI recruitment tool in 2018, the lesson was stark: train an algorithm on historical hiring decisions and it replicates historical preferences, including every structural bias embedded in that data. Most organisations have no equivalent transparency into what their own systems have quietly learned to do.
Rules feel neutral. They are not. They encode the assumptions of whoever designed them and perpetuate the patterns of a workforce that, in most organisations, was not diverse to begin with. Algorithms that penalise employment gaps disproportionately affect women. Systems weighted towards specific institutions entrench socioeconomic advantage. Keyword matching that demands precise terminology disadvantages candidates from adjacent industries with equivalent skills.
Local government administrator Paul Shuttleworth has brought legal claims challenging the use of terms such as “digital native” and “social media native” in civil service recruitment. Most of his claims have been struck out or found to be outside the tribunal’s jurisdiction but some age discrimination and victimisation complaints against the Cabinet Office are proceeding. None of this needs to be intentional to cause harm at scale, which is precisely what makes it more dangerous. Automated systems do not create immunity from discrimination claims, and regulators in the UK and EU are increasingly focused on algorithmic accountability in employment.
What HR should ask before using hiring algorithms
What is this tool actually measuring?
Who is being filtered out before a human sees them?
Do our screening criteria predict performance or simply reflect past assumptions?
Can candidates re-enter the process if the system gets it wrong?
Are we auditing rejection data by age, gender and other relevant characteristics?
Who in HR owns accountability for algorithmic hiring decisions?
What responsible algorithmic hiring looks like
None of this is an argument against technology in recruiting. The question is not whether to use these tools, it is whether you are governing them with the same scrutiny you would apply to any other consequential people decision.
Audit your rejection data, not just your hires. Most organisations invest significant effort analysing who they bring in. Very few examine who their screening process removes, and why including by age. If you don't have visibility into your rejected pool's demographic composition, that in itself is the finding. You can’t address a problem you can’t see. Ask the question. Remove the blinkers.
Separate screening from assessment. Algorithms are useful for administrative tasks: confirming minimum requirements, managing high-volume communication, scheduling at scale. The moment your system ranks candidates on qualities it cannot actually measure, you have handed consequential human judgement to a tool that was never built to make it.
Revisit your criteria with performance data, not assumptions. The requirements in your screening logic were written by someone, at some point, based on some version of what the role needed. When did you last test them against actual performance outcomes? Which qualifications predict success, and which are proxies for familiarity?
Build human re-entry points into the funnel. The most resilient processes treat algorithmic tools as a first pass, not a final word. Structured mechanisms for recruiters to surface candidates who scored below threshold but warrant a second look preserve human judgement where it matters most.
Finally, consider if you need to advertise on a wide platform. If the best candidates are those really keen on your organisation, your own job site and job page will help reduce the eyes on the jobs to those most keen.
The strategic question for HR leaders
The organisations that will hire best over the next decade are not those with the most sophisticated screening technology. They are those whose people leaders ask the harder question: what is our algorithm actually optimising for and is that still what we want?
Speed and volume alone are poor criteria for talent strategy. They are logistics! When logistics start shaping who gets a chance fixing it belongs in the CHRO's remit (not buried in a vendor contract which is likely in any case to not absolve you from responsibility).
The machine is making decisions about your talent pipeline every day. Are you paying enough attention to know what decisions it’s making?
FAQs
What is algorithmic hiring?
Algorithmic hiring is the use of automated systems to screen, rank or filter candidates during recruitment. These tools may parse CVs, match keywords, assess experience or prioritise candidates before a recruiter reviews them.
Why can hiring algorithms create bias?
Hiring algorithms can create bias when they are trained on past recruitment decisions or built around narrow assumptions about what a good candidate looks like. This can disadvantage people with career gaps, older workers, carers, women, career changers and candidates from adjacent industries.
How can HR leaders reduce bias in automated recruitment?
HR leaders can reduce bias by auditing rejection data, testing screening criteria against performance outcomes and creating human re-entry points for candidates who may have been wrongly screened out. They should also ask vendors to explain how tools make recommendations and what data is used.
Should organisations stop using AI in recruitment?
The issue is not whether organisations should use AI in recruitment but whether HR leaders understand how these tools work, what they are optimising for and who they may be excluding.
About the author
