A few years ago, Ken Schumacher was working for a technology company. Part of his job involved assessing potential hires: hopping on a Zoom call, giving an applicant an engineering test (kind of like a crossword puzzle with code instead of words), and going on “mute for an hour” as the applicant struggled through it.
Except many of the candidates weren’t struggling. The firm’s exercises were getting posted on sites such as Glassdoor. “All these savvy 23-year-olds would, of course, practice the problem three times, come to me, and crush it,” Schumacher told me. “Now the bigger problem is everyone’s using AI to write their resume.” They’re also using AI-powered chatbots and teleprompters to help them get to the next round. It’s become “really, really hard for anyone to figure out who’s real and who’s fake,” Schumacher said. (This turned out to be its own market opportunity—he now runs a start-up using AI to detect AI cheating by job candidates.)
The problem might be particularly acute in software engineering. But the same dystopian phenomenon is contorting the whole labor market. AI has Tinderized hiring. Workers are applying to hundreds of positions and never hearing back; companies are receiving thousands of resumes and struggling to respond. More than that, AI has Amazoned hiring. It has scrubbed the uniqueness out of applications, flooded the market with same-same-seeming offerings, increased the number of frauds, and replaced personal discretion with brute algorithmic assessment.
People had once hoped that Silicon Valley might not only smooth out the logistics of getting a job but also make the process fairer. Unbiased tools would replace alma-mater networks. Digital portals would accept applications from anyone, anywhere. Workers would get free access to templates, practice exams, and advice. “Technology in general tends to improve the efficiency of job matching,” Mitchell Hoffman, a labor economist at UC Santa Barbara, told me. But AI in particular seems to destroy it. Employers and employees are locked in an “arms race, where it’s AI-on-AI crime,” Kathleen Creel, a philosopher and computer scientist at Northeastern University, told me.
In just a few years, tools such as ChatGPT and Claude have commoditized the production of cover letters and resumes. Large shares of job applicants are using generative chatbots to polish their language and summarize their accomplishments—raising the average quality of these personal documents, at the cost of “compressing” and “homogenizing” the information they convey, as one Columbia Business School paper put it. In Silicon Valley, the phenomenon is sometimes called “signal collapse.” CVs used to be filled with advertent and inadvertent signals for hiring managers to parse: degrees and certifications and languages spoken, as well as formatting errors and unusual digressions and over-honest admissions. Now everyone looks better and everyone looks the same and everyone parrots important key words and everyone uses punchy action verbs. Hiring managers are straining to “distinguish underlying expertise,” and to separate precious signal from cacophonous chatbot noise.
Because AI has reduced the time job seekers spend writing applications, people are submitting hundreds of them on ZipRecruiter, LinkedIn, and other sites—often for positions that are not quite a fit, often never to hear anything back. Applicants aren’t getting feedback on their strengths and weaknesses. They’re not getting intel on what they might need to do to get hired, or the skills they might need to obtain the job they want. Signal collapse is running in both directions.
To sift through all of those applications, companies are turning, again, to AI. A recent survey by Resume Builder found that four in five companies are using AI to scan resumes, two in five are using chatbots to communicate with candidates, and one in five is giving AI interviews. AI screening tools in particular are creating an “algorithmic monoculture” in hiring, a new study of 4 million job applications has found. More candidates are being rejected by every firm that they apply to; in general, hiring decisions are more uniform than they would be if HR managers were not using algorithmic screens. A decade or three ago, firms would “go through the resumes and pick out schools they knew, or fancy schools, or schools that had good programs” in relevant fields, Creel told me. “That’s bad in its own way, from an egalitarian point of view. But at least different people had different arbitrary screening criteria.”
Moreover, screening algorithms are discriminating against Black and Asian candidates, the paper found. “There’s this filtering happening, and we don’t understand it, because these systems are opaque and custom-built for different institutions, and we don’t know the functionality,” Sarah Bana, a co-author of the paper and a digital fellow at the Stanford Digital Economy Lab, told me. Nevertheless, the algorithms’ tendency to advantage the advantaged and disadvantage the disadvantaged seems clear.
For candidates who make it past an AI screen, a test such as the one Ken Schumacher administered may await. Companies ask applicants to describe how they would respond to an unruly customer, or figure out an engineering fix to a software problem. In a lot of cases, this step involves AI writing the exam, proctoring the exam, and taking the exam. “I look at this stuff all day, every day, and even I have to pinch myself sometimes—the amount of fraud!” Schumacher said. “It’s just totally ridiculous.” The result is that AI is slowing HR down instead of speeding it up, because companies have to spend more time scrutinizing applications and performing background checks. A number of large firms are making face-to-face interactions a bigger part of the hiring process again. Google is ensuring that candidates have “at least one round” of in-person interviews, to “make sure the fundamentals are there,” Sundar Pichai, its CEO, has said. Cisco is increasing “verification steps and enhanced background checks that may involve an in-person component.” Companies are also extending probationary periods and hiring people on contract, with the intention of turning them into full-time employees after managers can vet their work.
Finally, some firms are resorting to old-fashioned methods: referrals, alumni networks, local job boards, headhunters. Schumacher said that firms were “retreating to pedigree.” They aren’t doing this because “the Harvard engineers are the best in the world,” he said. “But it’s the safer play in an era where it’s really hard to tell who is legit and who is kind of full of crap.”
Even with those techniques, hiring has simply gotten harder—so much so that the entire labor market might become a touch more sclerotic, and business dynamism might dim a bit. Companies are concerned that worker tenure will go down, as managers end up having to fire underqualified workers and ill-suited employees get frustrated and leave. Bana worries about businesses replicating their current workforces, and missing out on employees who might break them out of groupthink and expand their ambitions. Hoffman, the labor economist, raised the issue of AI reducing workers’ incentive to learn new skills and deepen their expertise.
What should a person looking for a job in this AI hellscape do? Nobody I spoke with was certain. But being yourself in a cover letter and dropping it off in person doesn’t seem like the worst idea.