Applying Into the Void: What the AI Arms Race in Hiring Means for UK Jobseekers

Ask anyone who has looked for work in Britain over the past couple of years and you will hear a version of the same story. Hundreds of applications sent. Dozens of rejections received, several arriving so quickly that no human could plausibly have read the CV. A creeping suspicion that the whole process is being run by software talking to other software.

That suspicion is not baseless, but it is also not quite accurate, and the gap between what candidates assume is happening and what actually happens after you click "apply" has become a problem in its own right. It changes how people write their CVs, how long their job search lasts, and in financial terms, how much of a cash buffer they need before redundancy or resignation becomes survivable.

This matters for household budgets as much as careers. A job hunt that used to take ten applications and six weeks can now stretch across several months, and that shift has consequences for savings, pension contributions, mortgage affordability assessments and the simple question of how long you can go without a payslip.

Both sides have armed themselves

The machinery underneath modern hiring is not new. Applicant tracking systems have been sorting, tagging and filtering CVs since long before generative AI arrived, and most large employers have used some form of automated sifting for years. What has changed is the arrival of cheap, fluent language models on the candidate's side of the table.

Sheeraz Gulsher, chief executive of the non-profit People Like Us, has described the current situation as an "AI war", with jobseekers using AI to strengthen applications while employers drown in the resulting volume. His organisation is running a campaign called Reject the Rejections, which presses employers to be more transparent and accountable about automated recruitment systems and to audit them independently.

The escalation dynamic is familiar to anyone who has watched an arms race. Candidates suspect a bot is reading their CV, so they use a bot to write it. Employers then receive triple the applications, all oddly similar in structure and phrasing, so they lean harder on automation to cope. Industry analysis of how AI is rewriting the recruitment playbook has made the same observation: the tooling on each side is now partly a response to the tooling on the other.

The irony is that automated screening may be less widespread than candidates assume. In a survey of 425 recruiters conducted by CV-Library, roughly 90 said they used AI at the CV screening stage. Alec Stanwell, the company's marketing director, calls this a "perception gap", and argues it is damaging because candidates are writing for an imagined algorithm rather than a human reader. The result, he says, is that CVs increasingly look identical, which then makes automated filtering more attractive to overwhelmed recruiters.

One London agency, Splendid Communications, reported that applications for its three internship places tripled in a single year, with submissions that "all look the same". When the firm tried using AI detection tools to work out how much content was machine-generated, the tools produced such inconsistent results that staff abandoned them and sifted manually.

What the rules actually say

Most jobseekers do not realise they have any recourse at all. A Censuswide survey of 2,000 UK workers who had applied for at least one job in the previous 24 months found that 75% were unaware of their right to challenge an automated rejection.

Under UK data protection law, people subject to certain solely automated decisions with significant effects can request human intervention. The Information Commissioner's Office has published guidance on this, and its position on recruitment technology is that efficiency gains "must not come at the expense of people's rights". The regulator expects employers to be upfront about when automated decision-making is in use, how it works, and how it could affect an application, and to take active steps to identify and mitigate bias at every stage of building and deploying these tools.

The government has taken a similar line, noting that existing equality and data protection law already applies to AI systems and that it will act if further protections prove necessary. That is a holding position rather than a new regulatory framework, and it leaves a fair amount of ambiguity about what "solely automated" means when a human nominally signs off a decision that software has effectively already made.

In practice, exercising the right is harder than it sounds. Several jobseekers interviewed about their experiences said they would hesitate to ask for a human review for fear of damaging their chances with that employer. One business consultant in Lewisham described receiving a rejection on a Saturday morning, a day after applications closed, advising her to work on her skills and reapply in six months. Her objection was not the outcome but the absence of any genuine assessment behind it. An HR professional from Surrey, made redundant and tracking more than 100 applications, noted that many rejections landed within 24 to 36 hours, against a personal track record of landing a role within ten applications.

The evidence on bias is real but incomplete

The fairness question deserves care, because the available evidence is suggestive rather than conclusive.

Researchers at Stanford analysed four million job applications processed by a single AI hiring tool over the four years to December 2022. They found that some candidates were rejected from multiple roles more often than would be expected had each employer assessed them independently, and that black and Asian applicants were recommended at lower rates. The researchers themselves flagged the limits: one tool, one dataset, and confounding factors including education and immigration status that could partly explain the pattern.

Litigation is testing similar questions elsewhere. The HR software company Workday is defending claims in the United States alleging that its tools discriminated against applicants. Workday denies the allegations and says its products support rather than replace human judgement, and that they do not make hiring decisions. The case has not concluded, so no finding should be read into it.

What is harder to dismiss is the behavioural evidence from candidates. The same research that found widespread ignorance of appeal rights also found that 70% of ethnically diverse jobseekers had altered or hidden details on their CVs, including name, religion or country of origin, out of fear that software would screen them out. Among white candidates the figure was 57%. Whether or not any given system is biased, a substantial share of the workforce is now behaving as though it is, which distorts the information employers receive.

Not every system works the same way, which is why blanket statements about "AI hiring" tend to mislead. Barb Hyman, founder of Sapia AI, whose product grades candidates on text-based chat interviews rather than CVs, argues that the architecture matters: her company says it holds no demographic data and no record of whether a candidate was previously hired. Her broader point about trust is the useful one. Telling candidates what is being measured, how they will be scored, and giving them feedback does more for process integrity than trying to police them.

The financial arithmetic of a longer search

For readers of this site, the practical consequence is less about technology ethics and more about cashflow. If automated sifting lengthens the average search, the implicit cost falls on the candidate.

Trade coverage of the UK labour market has tracked claims that AI adoption is weighing on hiring in parts of the economy, particularly in entry-level and administrative roles. The causal picture is genuinely contested, since hiring volumes also respond to employer national insurance costs, interest rates and general economic caution, and it is difficult to separate the effect of automation from the effect of a soft market. The honest position is that nobody can yet attribute a precise share of weak UK hiring to AI.

What is more certain is the direction of employer incentives. The business case for automating HR functions rests on cost per hire, time to fill and administrative savings, and those numbers are compelling enough that adoption is unlikely to reverse. Candidates should plan around a process that is more automated next year than it is today.

That suggests a few things worth thinking through before you need them. A statutory redundancy payment is calculated from age, length of service and weekly pay, and is capped, so for many people it covers weeks rather than months. If your personal assumption has been that a new role arrives inside two months, it is worth stress-testing the household budget against four or six instead, and checking what fixed commitments could be paused rather than cancelled. Pension contributions, notice periods, holiday pay and any outstanding salary sacrifice arrangements all affect the real size of the buffer you are working with.

There is also a tax dimension that catches people out. Redundancy payments are generally tax-free up to £30,000, but pay in lieu of notice and accrued holiday pay are treated as earnings. A gap in employment part way through a tax year can also leave you having overpaid PAYE, which is worth checking rather than assuming HMRC will reconcile it promptly.

None of this is advice about what you should do, and individual circumstances vary enormously. The general principle is simply that the length of a job search is now a variable worth modelling rather than guessing.

What to change about how you apply

The tempting response to an automated system is more automation: mass applications, AI-polished cover letters, keyword stuffing. The evidence from recruiters suggests this is self-defeating, because it is precisely the flood of indistinguishable applications that pushes employers towards heavier filtering.

A narrower, better-targeted approach has a stronger logic to it. Fewer applications written with genuine specificity, direct approaches to hiring managers, and use of networks all sidestep the top of the funnel entirely. Keeping a record of what you applied for and when gives you the basis to request a human review, should you want to, and the right exists whether or not employers advertise it.

The uncomfortable truth is that nobody outside the company can see what happens after you click "apply", and that opacity is the core grievance. Until auditing and disclosure become normal practice rather than campaign demands, the most useful thing a jobseeker can control is the size of the financial runway behind them.