Jobseekers are increasingly finding themselves being rejected by machines before they have even had the chance to speak to a human, with ethnic minority applicants being disproportionately effected.
New research from non-profit People Like Us and Censuswide suggests automated recruitment software is rejecting UK candidates within minutes of them applying — and, in some cases, before the application deadline has even closed.
The findings have prompted the launch of Reject the Rejections, a campaign calling for greater transparency and oversight of the artificial intelligence and automated systems now being used to sift through job applications.
The research, based on a survey of 2,000 UK jobseekers, found that 80% had been rejected for a job in the past two years.
Of those, almost three quarters (74%) suspected at least one rejection had been generated automatically, without a human ever reading their application.
Yet candidates are rarely told when automation is involved. Just 15% said an employer had informed them that an automated system had been used to assess their application.
Ethnic minority candidates face greater hurdles
The research suggests the impact is particularly pronounced for ethnic minority jobseekers.
Ethnically diverse candidates said they applied for around a third more jobs than their white counterparts, but were also more likely to receive a rejection within an hour — 41% compared with 32%.
They were also more likely to alter their CVs in an attempt to get past automated screening.
Overall, 60% of jobseekers said they had hidden or changed something on their CV because they feared an automated system would reject them. That figure rose to 70% among ethnic minority candidates.
Some were removing details directly connected to their identity. Ethnic minority candidates were nearly three times as likely to remove their country of origin from their CV (14% compared with 5%) and more than twice as likely to remove religious or cultural markers (14% compared with 6%).
One in 20 candidates said they had even changed their name.
People Like Us has launched a powerful new short film starring Ebenezer Gyau (Black Mirror), built from real rejection lines received by job seekers and a free template so anyone rejected by a machine can demand answers.
People Like Us says this problem predates AI, pointing to decades of UK research into name bias in recruitment. However, it argues that automated systems can introduce greater speed and scale into the process — while removing the human being who could potentially be asked to explain a rejection.
The findings also follow analysis by UCL of more than 100,000 live graduate applications, which found that ethnic minority candidates were well represented among applicants for graduate jobs but were disproportionately rejected during initial screening and online testing, despite having similar qualifications.
Jobseekers are learning to game the machines
The research paints a picture of an increasingly bizarre recruitment arms race, with applicants and employers both turning to AI to deal with the other side’s use of it.
More than two thirds of jobseekers (66%) said they had tried to “game” recruitment software.
Around 27% had used AI tools such as ChatGPT to rewrite their CV in an attempt to get past automated screening, while 13% said they had hidden keywords in their CVs using white text.
The researchers point to the sheer volume of applications as one reason for the growing reliance on automated systems. The average employer belonging to the Institute of Student Employers receives around 140 applications for every graduate vacancy.
That can create a cycle in which AI-assisted applications increase the number of applications employers receive, encouraging employers to rely even more heavily on automated screening.
And some candidates are deciding not to enter the race at all.
Around 31% of jobseekers said they had chosen not to apply for a job they believed they were qualified for because they suspected AI would screen them out. Among ethnic minority candidates, that rose to 39%.
Rejected at machine speed
For some applicants, the wait for a rejection is now measured in minutes rather than days or weeks.
The research found that 34% of rejected candidates had received their rejection within an hour, while 9% had been rejected within five minutes.
Almost half (44%) said they had been rejected before the application deadline had even passed.
Overall, 64% of jobseekers suspected they had been rejected without a person ever seeing their application.
And 81% agreed that applying for jobs now feels like “trying to pass a test for a machine” rather than being seen as a person.
The psychological impact is also highlighted in the research. Some 38% said rapid automated rejections had knocked their confidence, while 34% said they had been put off applying for other jobs as a result.
Eight per cent said they had sought mental health support because of job rejections, rising to one in 10 among ethnic minority candidates.
Meanwhile, jobseekers said they received no response at all to an average of 35% of their applications.
How much do employers know about their own AI?
The research also raises questions about how much oversight employers have over the systems they are using.
Among employers using automated screening, 28% said they did not monitor whether outcomes differed according to ethnicity or other protected characteristics.
Of those that did monitor for differences, 82% said they had found them, with 36% describing the differences as meaningful.
The research also found that 30% of decision-makers either did not know, or were unsure, how their own screening tool scored candidates.
And only around 35% of hiring decision-makers surveyed could identify each of the rights candidates already have under UK data protection law, according to People Like Us.
Candidates want a human being to have the final say
The campaign also highlights a lack of awareness among jobseekers about their existing rights.
Three quarters (75%) of those surveyed did not know they may be able to request a human review of an automated rejection under UK data protection law.
Meanwhile, 42% said they had wanted to challenge a rejection but could find no way of doing so.
Of those who did request a human review, just 28% said they received one.
People Like Us is now calling for employers to disclose when automated or AI-assisted screening is being used in recruitment, including before candidates apply.
It also wants automated recruitment systems to undergo independent audits for potential bias.
The organisation is urging the Government to introduce mandatory disclosure through the forthcoming Equality (Race and Disability) Bill, arguing that greater transparency around recruitment should form part of wider efforts to tackle inequality in the workplace.
The campaign points to New York, where independent bias audits of certain automated employment decision tools have been required since 2023.
There appears to be considerable appetite among jobseekers for greater transparency, with 82% saying disclosure of automated screening in job adverts should be required by law.
AI may be making it easier for employers to process thousands of applications at once. But the research suggests that, for some jobseekers, the result is a recruitment process where getting through the door increasingly means figuring out how to impress a machine first.
Sheeraz Gulsher, co-founder of People Like Us said: “Employers aren’t the villains here, most resort to automations in good faith to cope with a surging volume of applicants, and four in ten did it believing they were reducing bias.
“However, good faith isn’t governance when 82% of the employers who checked found outcomes vary by ethnicity, ‘we didn’t know’ stops being a defence. The Equality (Race and Disability) Bill gives the Government the vehicle to make transparency the rule to enable fairer access to jobs for all.
“Automation has a clear role in managing surging application numbers – but checks and balances must keep recruitment equitable for all.”
Tom Heys, pay transparency and AI specialist at Lewis Silkin added: “If your hiring software is rejecting ethnic minority talent simply because of their name or background, that could be discrimination. And arguing ‘the algo did it’ won’t save you in a tribunal.
“The law doesn’t care whether the bias is human or digital. Employers who outsource decision-making to machines don’t get to outsource accountability along with it.”