The AI Arms Race in Hiring: How Job Seekers and Employers Navigate a Broken System

Jodi Beggs, a data scientist and active job seeker, recently experienced firsthand the intricate and often perplexing world of automated application screening. Her resume was "dinged" for what might seem like minor inconsistencies to a human eye: it was two pages long, she used her middle initial on one document but omitted it on another, and she spelled out "percent" instead of using the "%" symbol. Each of these seemingly trivial details cost her points when her application materials were run through an online system designed to assess their compatibility with a job description. This feedback, she noted, was typical of tools like Jobscan, which analyze applicant materials for keyword matching and formatting compliance. Her reflection encapsulates a growing sentiment among job seekers: "If my job is to please the robots, then even if I don’t like AI-generated content, maybe the machines do." This highlights the prevailing belief that success in today’s competitive job market often hinges on appeasing algorithms rather than directly impressing human recruiters.

The core premise driving the widespread use of such optimization tools is that hiring committees extensively utilize Artificial Intelligence (AI) within their Applicant Tracking Systems (ATS) to automatically rate and rank candidates. The common understanding is that only the top 10 to 20 percent of applicants ever receive human consideration, leaving the vast majority—the bottom 80 percent—to languish unseen. To bypass this digital gatekeeper and make it onto the coveted short list, candidates are increasingly advised to meticulously tailor their CVs and cover letters, packing them with precise keywords and adhering to specific formatting requirements that an ATS is presumed to favor.

However, this pervasive belief, while understandable given the opacity of the hiring process, isn’t always accurate. While automated ranking is indeed a feature in some organizations’ hiring processes, many others still manage their hiring personally, with human recruiters overseeing the entire journey from application to offer. Daniel Chait, CEO of Greenhouse, an ATS company, emphasizes the significant amount of "folklore" surrounding these tools, particularly from the job seeker’s perspective. He clarifies that no two ATSs are identical, and the underlying technology evolves rapidly. An AI tool embedded within an ATS might function one way today and entirely differently tomorrow. Crucially, whether an ATS even employs AI for candidate ranking depends entirely on the specific product, the features a company chooses to pay for, and how those features are actually utilized by the hiring team. Despite this nuanced reality, the powerful perception that AI is universally involved means that job seekers will continue their efforts to "game the system" using their own AI-powered tools.

This situation creates what Chait terms an "AI doom loop," a tragic scenario where both sides—job seekers and employers—are using AI to solve their respective problems, but in ways that exacerbate the overall issue. The result is a cycle where more AI use begets more AI use, ultimately benefiting no one. This mutual reliance on AI tools is contributing to a significant breakdown of trust in the hiring ecosystem. Candidates invest immense time and effort into their job searches, often receiving no feedback, leaving the application process feeling like a "black box." Simultaneously, employers report receiving hundreds of applications that appear nearly identical, leading them to increasingly rely on AI rating systems within their ATSs to find some quick differentiation, thereby fueling the very cycle they are trying to manage.

The efficacy of applicant optimization tools like Jobscan is also subject to scrutiny. One recruiter demonstrated this by intentionally submitting a resume with "atrocious Jobscan scores" yet still secured a dozen interviews and ultimately a job offer. This anecdote suggests that factors beyond mere keyword matching, such as the overall quality of experience, networking, or direct human referral, can still significantly influence hiring outcomes. Furthermore, the business model of these tools—charging $30 to $50 per month—implies that they are not exactly incentivized to quickly remove users from the job market.

This complex landscape is further complicated by broader economic conditions. A "low-fire, low-hire economy" means job postings are scarce, intensifying competition. The proliferation of "ghost jobs"—positions advertised but not actively being filled—and outright scams adds another layer of frustration, prompting some states to consider new legislation to combat these deceptive practices. The cumulative effect is a profound erosion of trust between job seekers and employers, making the job search an increasingly arduous and often disheartening endeavor.

However, not all organizations subscribe to heavy AI-driven screening. Many maintain a human-centric approach to vetting candidates, driven by their organizational philosophy and culture rather than solely by size or applicant volume. Kim Jones, Vice President of Human Resources at Toshiba, firmly states, "We have humans review every application." While she doesn’t mind applicants using AI to polish their materials, she clarifies that such efforts are "not really going to help with getting through the ATS" at Toshiba. Their primary culling factors are fundamental: meeting job requirements, alignment with salary expectations, and other practical considerations like rehire status. Jones has, however, observed unwanted AI usage during the interview stage, noting instances where "you hear the pause, maybe hear the typing, then they come up with a verbose answer," indicating candidates are using AI to formulate responses in real-time.

Another compelling case comes from Doist, a small, fully remote company that hires internationally, leading to a high volume of applications. Nadia Vatalidis, Head of People at Doist, and her team conducted an experiment to evaluate the effectiveness of ATS rankings. They took previously filled roles and fed the job descriptions along with all saved applicant materials into an ATS AI. Their goal was to see if the AI would shortlist the same candidates who were ultimately interviewed and hired. The results were telling: "In two instances that we tested, the person we hired wasn’t in the short list." Vatalidis noted some overlap in who received interviews, but crucially, their new hires—individuals who had been performing exceptionally well for months—did not make the AI’s cut. This suggests that AI algorithms, while efficient at matching keywords, may struggle to identify the nuanced qualities, potential, or cultural fit that human evaluators value and which ultimately lead to successful hires.

Job seekers, too, are beginning to explore innovative ways of leveraging AI beyond simply optimizing their resumes for imagined ATS algorithms. James Jacobsen, a design professional, embarked on a job search five months ago, dedicating the equivalent of a full-time role to searching and applying. Initially, he used large language models like Claude and ChatGPT to polish his application materials and align them with job descriptions. However, he found this traditional approach wasn’t "moving the needle" in terms of getting him noticed.

Jacobsen then pivoted, transforming his use of AI into a personalized, reverse ATS. Instead of focusing solely on application materials, he instructed Claude to streamline and track his entire job search. His AI assistant was tasked with combing job listings, analyzing descriptions, and meticulously logging them. He developed an elaborate scoring system based on factors such as the type of work, seniority level, and his varying salary requirements for in-person, hybrid, and remote roles. This allowed his AI assistant to highlight the top jobs he should prioritize. When he decided to reject a position, he added a note, ensuring that if the same role were reposted weeks later, his AI could immediately remind him why he wasn’t interested. This strategic shift enabled him to spend less time sifting through irrelevant postings and more time focusing on high-quality positions, though it did not immediately yield an offer. Jacobsen didn’t stop there; he also used Claude and ChatGPT to critique his portfolio, leading to a significant six-hour revamp. Two days later, he received a call from a prospective employer, although it did not ultimately result in an offer. His experience exemplifies a shift from passively adapting to AI to actively using it as a personal, proactive tool for managing the job search process.

In this challenging environment, Daniel Chait advises job candidates like Jacobsen that "doing more of the same isn’t the answer." He acknowledges that both sides are unhappy, stating, "This is the first time I can remember when both sides are unhappy. It’s just not working." For job seekers, his message is clear: "The answer is not just more spray-and-pray." Instead, Chait recommends a more thoughtful and human-centric approach. He suggests spending time researching companies, especially those that might not be the first names that come to mind, to find genuine cultural and mission alignment. Kim Jones of Toshiba reinforces the value of a often-neglected practice: submitting a cover letter. She notes that she "almost never sees a cover letter anymore" and that applicants who provide one immediately stand out. This simple act demonstrates genuine interest and provides an opportunity to convey personality and specific motivations beyond what a resume can. Furthermore, networking, an "underrated skill in today’s job market," remains a powerful way to make personal connections and uncover opportunities that might otherwise be missed by automated systems.

Ultimately, the current job market presents a complex interplay of human intention and technological influence. For job seekers navigating this intricate landscape, Chait offers a final, empathetic message: "It’s not you; it’s the system, and it stinks." This sentiment encapsulates the frustration and confusion experienced by many, highlighting the systemic issues that both job seekers and employers grapple with in the era of AI-driven hiring.