AI job applications bury real candidates under synthetic noise
The job market has reached a breaking point: artificial intelligence now generates applications faster than any recruiter can read them. The promise was efficiency — a frictionless path from ‘I want a job’ to ‘you are hired.’ What materialized is a system where the ease of applying has made applying meaningless. The gap between what the technology claims to do and what it does is not a bug; it is the operating principle.
The problem is not that AI is too smart, but that it is too accommodating. It writes a cover letter for any role, tweaks a resume for any keyword, and submits applications with the emotional investment of a spam bot. The technology does exactly what it promises — and that is the deception. It promises to make the candidate more competitive, but in doing so, it makes every candidate identical. When everyone can generate a perfect application in seconds, perfection becomes the baseline, and the baseline becomes noise.
Andrew Stockwell, former head of people at the software-buying company Vendr, remembers the old routine with professional nostalgia. [1] He would post a listing, wait a few days, and review a few dozen applications — maybe a hundred if he was lucky. Recruiters would identify other potential candidates, adding a few more names to the mix. Stockwell read every resume, conducted every interview, and passed along only the strongest candidates to hiring managers. It was a process built on the assumption that the pool of applicants was finite, manageable, and mostly serious. That assumption is now dust.
Sometime in the last year, Stockwell says, everything changed. The process started the same way, but within a day or two, he was not looking at 100 candidates — he was flooded with hundreds, sometimes topping a thousand. A good chunk were ‘total bogus,’ Stockwell says, fake candidates generated by bots or automated systems. Many more appeared to have been written using AI, stretching the truth and making it difficult to distinguish one applicant from another. The sheer quantity was overwhelming, and Stockwell found himself paying his talent-acquisition professionals just to look through applications all day long.
This is the first layer of the deception: the AI pretends to be a job applicant, but it is actually a volume generator. The technology does not understand the role, the company, or the human being behind the resume. It understands patterns in job postings and produces text that matches those patterns. The result is a pile of documents that look like candidates but contain no signal. Recruiters face a paradox: they have more data than ever before, yet less information. The AI has not made the process more efficient; it has made it more opaque, hiding the few real candidates under a mountain of synthetic ones.
The industry response to this avalanche has been predictable. Recruiters are turning to AI to fight AI, deploying automated screening tools and chatbots to conduct first-round interviews. Ophir Samson, head of voice AI for Greenhouse, a recruiting platform, watched this unfold from the inside. [2] He joined the company earlier this year when it acquired his startup, which used AI to conduct job interviews. The pitch was straightforward: an AI interview can sort the unserious from the serious, offering recruiters context beyond a resume. But this solution contains its own deception. The AI claims to judge potential, but it is actually just another filter, another layer of abstraction between the human candidate and the human recruiter.
Samson recalls what recruiters told him a year ago: ‘We want to make it as easy as possible to apply for jobs.’ They promised candidates a ‘seamless experience,’ a one-click application process that would remove all pain points. What they got, Samson says, was 2,000 applicants in 24 hours for a single job. ‘That is a shitty experience for everyone.’ Good applicants cannot stand out, while bad candidates slip through the cracks. Recruiters face the tortuous task of judging thousands of candidates every single day. Instead of taking the time to search LinkedIn for dream employees, they spend hours in applicant tracking systems, tweaking filters to determine whose applications are worth a 30-second skim. [4]
Now, the narrative has flipped. Recruiters are telling Samson, ‘actually, we kind of want friction.’ The friction is good. They want to make it harder. This is a stunning reversal, a recognition that the ease they once championed was not a feature but a flaw. The desire to remove pain points made sense in a different era, when the highest number of qualified candidates meant a company could maximize its chances of finding the best fit. For years, recruiters posted on job boards and LinkedIn, signed up for the ‘Easy Apply’ feature, and promised a ‘one-click’ application process. Post-pandemic, those efforts went into overdrive as employers struggled to fill open roles.
Jane Curran, chief transformation officer at real estate giant JLL, remembers the job-hopping era with a wry laugh. [3] ‘Everyone was job hopping, because you literally could have three offers in an afternoon.’ Now, she says, ‘it is the polar opposite.’ Openings peaked at a record 12.3 million in March 2022, according to Bureau of Labor Statistics data, followed by two years of decline. Since mid-2024, openings have hovered at around 7 million, give or take 500,000. The market has contracted, but the barriers to entry have collapsed even further. AI, which entered the mainstream in late 2022 with the public release of ChatGPT, allows applicants to rewrite resumes and track new postings almost instantaneously.
The math is brutal. Companies like JobAssist, Sonara, and Ladder’s Apply4Me promise applicants that their technology will handle all the busy work, submitting ’10x as many applications with less effort than one manual application.’ [6] If you wanted to, you could apply for dozens of jobs a day. This formula has led to ballooning candidate pools and laborious hiring processes. LinkedIn tells WIRED that submissions per applicant on the platform are up 46 percent compared to February 2020. [5] Since the launch of ChatGPT, applications are up 22 percent. The explosive number of applications led to LinkedIn adding limits to cut back on automated and low-quality submissions. This month, the platform is rolling out a feature that informs seemingly underqualified applicants that they probably are not a good fit for a role, and suggests other jobs.
The deception here is layered. The AI application tools claim to help the candidate, but they actually devalue the candidate’s currency. A resume used to be a signal of intent, a carefully crafted document showing that a human had invested time and thought into a specific opportunity. Now, a resume is a commodity, generated in bulk and sprayed across the internet like digital graffiti. The tools claim to level the playing field, but they actually tilt it toward those who can game the system, not those who can do the work. The candidate who spends an hour crafting a personal cover letter is now competing against a bot that can produce a thousand variations in the same hour. The human effort is not just wasted; it is invisible.
Tessa White, a former HR executive who spent two decades in corporate America before leaving in 2018, has built a following of 800,000 on TikTok by sharing advice and commentary on the job market. [8] She believes the current situation is unworkable, and she traces it directly to the recruiters’ quest for speed and ease. ‘Every time we seem to strive for efficiency, we seem to give up quality,’ she says. White is not a Luddite; she uses technology daily. But she sees the AI-driven application process as a fundamental betrayal of what hiring should be. ‘We’re currently in a place where employers are complaining that they can’t find good people, and people are complaining that they can’t find jobs.’ Both sides are telling the truth, and both sides are being deceived by the same illusion of efficiency.
The illusion is that more applications mean more options. In reality, more applications mean more noise, and noise is the enemy of signal. The AI claims to be a magnifying glass, helping recruiters see candidates more clearly. Instead, it is a fog machine, obscuring the very people it was supposed to reveal. White argues that even the high-tech applicant tracking systems are ‘an antiquated way to look at people and skill sets.’ AI simply does not have the capabilities, she says, to accurately judge someone’s potential. It can parse keywords, match job titles, and flag gaps in employment history, but it cannot see the spark of curiosity, the resilience in the face of failure, or the quiet confidence that comes from years of hard-won experience.
This is the core of the deception: the AI claims to understand human potential, but it only understands data points. The gap between the claim and the reality is not a technical limitation; it is a philosophical one. The technology reduces a person to a pattern, and then it judges that pattern against other patterns. The result is a hiring process that is simultaneously hyper-rational and deeply irrational, one that produces reams of data but very little wisdom. Recruiters are drowning in information, starved for insight, and increasingly reliant on the very tools that created the problem.
The response has been a retreat to the human. The deluge of applications has increased recruiters’ reliance on referrals and internal hires. While these human-to-human connections have always been important, today, some see them as the only ways to circumvent the glut of applications. Stockwell believes this benefits people whose backgrounds align with existing employees’, which is not conducive to a diverse workforce. ‘The whole thing is a big mess,’ he says. The mess is not just in the volume of applications; it is in the way the system has inverted itself. The tools that were supposed to democratize access to jobs have made it harder for outsiders to break in, because the only reliable signal left is a personal connection.
Stockwell is now dealing with the mess from the other side of the application portal. In February, he left Vendr and is now on the job hunt. He treats the search like a job, ‘pounding the pavement’ and attending in-person networking events. He knows the system from the inside, which means he knows that his carefully crafted resume will likely be swallowed by the same algorithmic maw he once managed. He knows that the AI tools that generate applications are not his allies but his adversaries, flooding the market with synthetic competition. He knows that the only way to stand out is to be present, physically, in a room, where a handshake and a conversation still carry more weight than a thousand keyword-optimized documents.
The irony is that the technology that was supposed to make the job market more efficient has made it more tribal. The ease of applying has created a barrier of noise, and the only way through that barrier is a personal connection. This is not a bug; it is a feature of a system that has outsourced its judgment to machines. The machines are not evil; they are just literal. They do exactly what they are told, and what they are told is to find patterns in text. The patterns are not people, and the people are lost in the patterns.
Curran expects some companies to turn to knock-out questions that cut down the application pool via strict criteria. Others might add skills testing earlier in the process. These are attempts to reintroduce friction, to make the application process more like a gauntlet and less like a drive-through window. The logic is sound: if applying is easy, then applying is meaningless. By adding hurdles, recruiters hope to filter out the bots and the halfhearted, leaving only the genuinely interested and qualified. But this approach has its own risks. The friction might also filter out the overqualified, the underconfident, and the non-traditional candidates who do not fit the mold.
White is not convinced these changes fix the underlying problem. The frictionless resume drop has rendered the first round of the application process essentially worthless. ‘Employers are having such a hard time hiring people with the current antiquated process,’ White says. ‘Even AI and applicant tracking systems, which sound so high-tech, are an antiquated way to look at people and skill sets.’ The technology is new, but the thinking is old. It is still based on the assumption that a resume is a reliable proxy for a person, which was always a shaky assumption and is now, in the age of AI-generated text, almost completely untenable.
The gap between what AI claims to do and what it actually does is not a secret; it is the business model. The companies that sell AI application tools claim to empower candidates. The companies that sell AI screening tools claim to empower recruiters. Both are selling the same thing: the illusion of control. The candidate believes they are controlling their job search by submitting hundreds of applications. The recruiter believes they are controlling the hiring process by filtering thousands of resumes. Neither is actually in control. They are both passengers on a runaway train, and the tracks are made of synthetic text.
The technical problem is solvable. You can build better filters, more sophisticated algorithms, and more nuanced screening tools. You can add friction, knock-out questions, and skills tests. You can even, as some companies are doing, require candidates to submit video introductions or complete work samples. But the social problem is not solvable with technology. The social problem is that we have outsourced a fundamentally human judgment to machines, and the machines are not up to the task. They can sort, filter, and rank, but they cannot discern. They cannot look at a person and see potential, because they do not see people at all. They see data.
The insight that emerges from this mess is that the problem is technically solved but socially unresolved. The technology works exactly as designed. It generates applications, filters candidates, and schedules interviews. The failure is not in the code; it is in the values encoded within it. We have built a system that optimizes for volume, speed, and cost, and we are surprised that it produces a poor experience for everyone involved. The AI does not deceive us by lying; it deceives us by giving us exactly what we asked for. We asked for efficiency, and we got it. We asked for ease, and we got it. We asked for a system that treats people like products, and we got it.
Sources
1. Vendr
2. Greenhouse
3. JLL
4. LinkedIn
5. WIRED
6. JobAssist
8. TikTok