I recently wrote about the purgatorial job market of 2026, where potential employees are largely evaluated by automated systems, participate in chatbot interviews, and even then often receive no feedback on their applications. Several job seekers described the experience to me as a “disheartening,” “dehumanizing,” “dystopian” underworld.
After this story was published, I received several emails from startup founders telling me that their new technology product would help fix job search failures. I spoke to a few of them, all of whom have identified a real “pain point” on the employer side: companies receive far too many CVs to evaluate and yet have difficulty finding truly qualified and suitable employees. Additionally, some job seekers use AI to misrepresent their skills and sometimes even their identities.
These startups claimed that their artificial intelligence-powered products – be it an app that offers a unique set of cognitive games and tests or a new way to process data and evaluate promising and unique candidates – would streamline and validate what machine learning has complicated and obscured. Their products included gathering a wealth of information about potential employees by mining data already available on the open internet or by monitoring a potential employee’s every keystroke and movement in a proprietary app.
I remain skeptical that adding another layer of artificial intelligence and surveillance to an insane process is the best solution. As employers continue to use AI to streamline the process of identifying good candidates, applicants have begun using similar tools to trick the systems that evaluate them. This is where the vicious circle begins: employers buy or build new tools that job seekers repeatedly use. It becomes a futile “spy versus spy” showdown rather than a useful way to achieve the ostensible goal of providing jobs to qualified people.
According to a 2026 report from Manpower Group, one of the largest staffing organizations in the world, “72 percent of employers said they are having difficulty finding the skilled workers they need.” Ifedapo Adeleye, the faculty director of the master’s in human resources management program at Georgetown University, told me that while AI makes some tasks more efficient for human resources professionals, the key results—finding the right people and reducing the time it takes to hire those people—don’t happen any faster and the entire process isn’t any more precise. There is also no solid evidence that companies retain the employees they hire. Gallup polls suggest that 52 percent of American workers are looking or actively looking for a new job, the highest percentage since this survey began in 2014.
This is a problem that appears to be solved only by a return to some analog processes and better regulation at the federal level that forces employers to make their current evaluation methods more transparent and fair.
Table of Contents
The failed promise of automated hiring
Employers have been using machine learning to evaluate potential employees for more than a decade, long before most of us thought about artificial intelligence, said Ifeoma Ajunwa, a law professor at Emory University and author of “The Quantified Worker: Law and Technology in the Modern Workplace.”
Ajunwa became aware of the ubiquity of automated hiring systems in the 2010s while interviewing formerly incarcerated people. Respondents had paid off their debts to society and were taking courses to acquire professional skills. But the automated hiring systems seemed to immediately reject them, either because of employment gaps or because they honestly answered a question about their criminal history.
A wave of startups in the mid-2010s claimed that using algorithms in hiring would make the process less biased and more efficient. In 2011, over 75 percent of job seekers searched for jobs online, and companies were inundated with too many applications to process, so they needed a way to start reviewing them. (Since ChatGPT became available in 2022, application volumes have become even more unmanageable; according to The Economist, “paid services like LazyApply and aiApply allow candidates to submit applications while they sleep, perfectly tailoring resumes and cover letters.”) But machine learning often reinforced existing biases and, for many HR professionals, introduced new complications like fake applicants and more fraud in skills assessments.
In 2018, Reuters broke the news that Amazon had developed a recruiting tool that discriminated against female applicants because its models were “trained to screen applicants by observing patterns in resumes submitted to the company over a ten-year period” and “most came from men, a reflection of male dominance in the tech industry.” Amazon no longer uses this tool, but scientific analysis and individual audits across many different industries have uncovered racial and gender bias.
A darkly comic example from Ajunwa’s book is what she calls the “story of Jareds.” One company had an attorney review its automated hiring system before implementing it. The lawyer asked the system, “What two factors would your system consider to be the most important in selecting a candidate?” The system responded, “(1) the candidate’s name is Jared and (2) the candidate played high school lacrosse.” Of course, the bot only knows the “suitable” CVs that it has been trained on.
This situation doesn’t work for hiring managers, who tend to get into their field because they actually like the people, and certainly don’t want to hire an army of Jareds. And there is still no solid evidence that automating recruiting has resulted in a better process or better outcomes.
Grace for applicants
The most important remedy for job candidates may be to standardize automated hiring laws at the federal level so that there is significantly more transparency about the use of AI and your personal data.
New York City laws could be a model for such legislation, my newsroom colleague Steve Lohr pointed out in 2023. “City law requires companies that use AI software in hiring to inform candidates that an automated system is being used. Additionally, companies must have the technology audited for bias annually by independent auditors,” he wrote.
Ajunwa also believes the Fair Credit Reporting Act should apply to all AI companies that collect data on applicants and then give them a score. Applicants should be able to access that score and “review it for inaccuracies and also request a correction from whoever has that file,” she said. That’s the argument behind a class action lawsuit filed in January against Eightfold AI, a company that uses a dataset containing the profiles of more than 1 billion people. The plaintiffs want transparency about how they were evaluated and what information was used for the evaluation.
None of this will change the fact that we are in a legitimately difficult market for job seekers. There are more applicants for open positions than there were a few years ago, and many employers don’t seem to be looking for diamonds in the rough; HR experts tell me that companies are increasing their expectations and demands because they are in a buyer’s market. Forcing employers to provide some transparency and even feedback to potential employees would help correct the power imbalance and make the job search process at least a little more humane.
One solution for employers is to resort to some of the imperfect analog methods of recruiting. In an essay in The Atlantic about the hideous job market, Annie Lowrey notes that “some companies are resorting to old-fashioned methods: referrals, alumni networks, local job boards, headhunters.” I’d like to hear what happens if a company experiments and requires all potential employees to submit a resume. (If hiring managers do this and want to tell me about the experience, email me here.)
Jessica Grose
Opinion writer
Even if you’re not looking for a job, machine learning will evaluate you at some point. When I spoke to Ifeoma Ajunwa, author of The Quantified Worker: Law and Technology in the Modern Workplace, for this article, she predicted that it will only be a matter of time before everyone looking for a job will be affected as various segments of the job-seeking population are systematically excluded by AI systems.
https://www.nytimes.com/2026/07/18/opinion/job-market-ai-employees.html
