The AI-generated resume crisis is breaking hiring

74% of US job seekers now use AI in their job search, rising to 78% across the UK, Ireland, and Germany, according to Greenhouse's 2025 AI in Hiring Report. Nearly half use it specifically to get past automated filters.

92% of recruiting leaders say AI-generated resumes are now commonplace, only about a third are highly confident a resume reflects a candidate's true abilities, and 64% say they've hired someone who misrepresented their skills, per Criteria Corp and Lighthouse Research & Advisory's 2026 Talent Acquisition Trends Study.

67% of US HR leaders say reviewing AI-generated applications has slowed hiring, while 84% report heavier workloads, in Robert Half's November 2025 survey.

Candidates are using AI to tailor applications faster than recruiters can read them, while employers are using AI to screen the resulting flood. Somewhere in that loop, the candidate who can actually do the job is getting lost.

The application arms race

Anshul Sharma is job hunting for senior data scientist roles and now applies to roughly twice as many jobs as he used to. “I have seen multiple rejections within minutes and hours of [submitting my] application which would suggest that no one really looked at my resume. That is what led me to use AI to help [create] a better resume which showcases all [my] skills for a particular role,” he says. “I am tailoring my existing resume so that information is real but it aligns with the job description.”

Sharma would rather not use AI for something this personal. “But with the amount of applications reaching the recruiter and AI screening tools rejecting applications for a few keywords, I think I don't have much choice.”

Erin Celise Smilkstein, a Senior Manager on a SaaS implementation team, hadn't applied for a job since 2018. "It's very disheartening to read a job description and think, 'I am absolutely perfect for this' and get a rejection 15 minutes after hitting submit when you spent two hours trying to make the resume sound perfect," she says. After tracing the no-reply email to the company's ATS provider, "that's when I really started using AI to... tailor my resume and cover letter to identically match... the keywords in the description," she says.

Smilkstein tailors fewer applications than most job seekers, but spends far longer on each one. "I have not applied to hundreds of jobs like I hear many people do, but I have applied to dozens," she says. "I am attempting to tailor every single application and attached resume to the job description, which means I am spending hours on every application."

When she applied for another role, Smilkstein was not immediately rejected. "I did get rejected two weeks later, which was even more disappointing, but I felt like that was a small win because at least I didn't get immediately filtered by the ATS," she says. "It doesn't feel like a showcase of talent anymore. It feels like trying to game or beat the system."

Recruiters feel the flood

Talent Acquisition Manager Margaret Buj, who also works as a principal recruiter and interview coach, says a single US-based customer success or product role at Mixmax can attract hundreds of applicants very quickly. "There are several things happening at once: a tougher job market, more layoffs, remote roles attracting candidates globally, easier application tools, and yes, AI making it faster for candidates to tailor or generate resumes," she says. "I suspect AI is contributing because it has lowered the effort required to apply."

Applicant volume has risen significantly at talent solutions firm Releady, says president and co-founder Quyen Pham, with auto-apply agents sending hundreds of tailored applications to individual postings, "with no human effort behind it and sometimes, no human behind it at all." The flood is particularly pronounced for fully remote roles. "The bottleneck ends up making it more difficult for real and qualified applicants to get through the funnel," Pham says.

Grady Gardner, GM and CRO of Braintrust AIR and Marketplace, oversees an AI recruiting platform that conducts tens of thousands of structured candidate interviews every month. "The rise in AI-generated applications has led to a need for automated screening in response," he says. "The quantity of spam or bulk-generated applications is higher than ever."

AI screening can’t filter out all the fakes

Founding member Neha Jain of WriteCV, an AI resume tool, works the resume-scoring side. "I see both ends of [the hiring process]: candidates using AI to write, and tools using AI to screen," she says. Screening catches "structure, formatting a parser [the software reading the resume] can read, whether a keyword is present," she says. What it struggles to establish is whether the experience behind those words is real. A bullet that reads "led migration to microservices, cut latency 40%" can score similarly to one that simply lists "microservices, latency, scalability." "One did the work, one listed the words. The tool often can't tell," Jain says.

Candidates can optimize directly for that weakness. Jain says obvious tactics such as hiding keywords in white text or copying language from a job description are now common. The subtler version is what she calls "AI padding": rewriting bullets to hit the terms in a posting without adding concrete evidence of the work. "It reads clean, passes the filter, and tells a recruiter nothing," she says.

"We have algorithms to detect AI-generated text to almost 99% accuracy," says Thejus Sunny, a software engineer who has built ATS systems handling high-volume application screening and founded Rejectless. Detecting that a candidate used AI isn't the hard part anymore. "The biggest technical challenge here isn't AI-generated text," he says. "It's how AI transforms an average resume with mediocre experience into a well-suited resume within seconds by adding things that don't exist."

One redacted resume flagged by Sunny's internal screening system was classified by his team as auto-generated by an LLM agent. All seven bullets under the candidate's most recent role aligned with responsibilities in the job description, nearly every bullet carried a quantified metric, and two sections gave conflicting figures for what appeared to be one piece of work. Yet "this resume would almost certainly have passed a keyword-based ATS filter," Sunny says. "In fact it would score better than an honest resume, because it mirrors the job description nearly perfectly." His conclusion: "Fabrication optimizes for the filter; honesty doesn't."

Qualified candidates can get filtered out too

An AI-generated-text flag cannot distinguish between a candidate who fabricated experience and one who used AI to rewrite legitimate experience more clearly, Sunny says. “They would still get flagged if we evaluate based on AI-generated text, and some good candidates get filtered out here.”

“Some strong candidates have badly written resumes,” Buj says. “They may have the right experience, but they undersell it, bury the relevant work, or use internal job titles that do not translate externally. Those candidates can easily be missed unless a recruiter looks carefully.”

Jain says the strongest candidates can also be penalized for writing too plainly. An engineer might write “Built X, shipped it, cut load time” and leave out the buzzword a parser expects. “The gap isn't competence, it's vocabulary, and keyword-first scoring punishes exactly the people who describe real work without dressing it up,” she says.

Even an unusual education format or several years of relevant in-market experience without a formal degree can cause a screening tool to miss a qualified candidate, says Sarah Doughty, VP of Talent Operations at TalentLab. “Most AI recruitment tools would filter out Mark Zuckerberg because he did not graduate,” she says, “and yet many tech companies would clearly be very excited to have him consider joining their leadership team.”

A 2026 Stanford-led study of algorithmic hiring analyzed 4 million applications from 3.4 million job seekers to 156 employers using algorithms from one hiring vendor. Among candidates who applied to four positions, 10% were rejected from all 4, significantly more than researchers expected if each employer were evaluating applicants independently. The pattern did not appear in the comparison data. One vendor’s screening system, in other words, could influence a candidate’s chances across multiple employers.

Screening becomes authentication, and double the work

65% of US hiring managers say AI-enhanced resumes have made candidate skills harder to verify, according to Robert Half's November 2025 survey of more than 2,000 hiring managers. To validate candidates, 42% are spending more time reviewing applications and 38% have increased the number of interviews per candidate.

At Huntress, Manager of Recruiting Erin Bortz leads the corporate and sales recruiting teams. “The biggest notable difference, despite adding tools to help, is the amount of time spent on application review,” she says. “Between AI-enhanced resumes, fake candidates, and bot applications, recruiters have become detectives deciphering what's true and what's fake.”

A 2026 FAccT study of 22 U.S. recruiting professionals found only marginal efficiency gains from GenAI. Many participants still felt pressure to adopt it, partly to counter applicants’ own use of AI and partly to increase productivity.

At Releady, Pham’s HR team reads a resume “like a story,” checking whether a candidate’s career holds together as a coherent sequence. “Does the career hold together as a linear narrative, or do skills just materialize with nowhere to have come from?” she says. “An AI-assisted resume often reads clean on the surface and falls apart the moment you trace the actual sequence.”

The recruiter who conducts Releady’s initial video screen also appears during the client interview, a process Pham says has caught impostors. Interview questions now focus on experience a candidate has to explain in detail. “We stopped asking things an AI can answer and moved entirely to lived experience and judgment,” she says: “describe the environment, the decision you made and why, what failed, how you fixed it.”

At Birchbury, founder Matthew Tran caught one candidate who listed three well-known brands as past clients by asking for the live project file behind the portfolio. “A real designer's file shows version history: layers renamed, sections rebuilt,” he says. The candidate's file had “one clean layer with no history,” assembled the night before. An email to one of the listed clients confirmed the company had never heard of the candidate.

“A polished portfolio link means less too, since tools generate clean mockups that never get built,” Tran says. He now sends finalists a broken layout from Birchbury’s site and asks them to fix it and explain their reasoning. “A chatbot can describe a fix. It can't rebuild the file.”

“The work has shifted from reading applications to authenticating people,” Pham says. “The job of a recruiter is a different job today than two years ago.” Resumes, cover letters and individual interviews have also lost some of their value as standalone evidence. “None of them are proof of anything on their own now.”