Somewhere between 70 and 90 percent of large employers now route job applications through an applicant tracking system before a human recruiter ever opens a resume. For job seekers, this has created a strange and often frustrating gap: a candidate can be well qualified on paper and still never hear back, not because a hiring manager rejected them, but because software filtered them out before a person saw the file.
The mechanics of how these systems work, what they look for, what they ignore, and where they break down, are rarely explained clearly to the people whose careers depend on getting past them.
Inside the Applicant Tracking System
An applicant tracking system is, at its core, a database and workflow tool built to manage the flood of applications a company receives for open roles. It stores resumes, cover letters, and candidate contact information, tracks where each applicant sits in the hiring pipeline, and gives recruiters a single dashboard to manage dozens or hundreds of job postings at once. The artificial intelligence layer that has drawn so much attention in recent years sits on top of this older infrastructure, adding automated parsing, scoring, and ranking features to a system originally built mostly for record-keeping.
- Resume parsing engines: convert an uploaded file into structured data, pulling out fields like job titles, employers, dates, and education, though formatting errors in the original document often scramble this process.
- Keyword matching modules: scan parsed text for terms pulled from the job description, flagging resumes that contain or lack specific skills, certifications, or job titles.
- Scoring and ranking algorithms: assign each applicant a numeric score based on how closely their resume matches the posted requirements, then sort the applicant pool by that score for recruiter review.
- Knockout question filters: automatically reject candidates who answer screening questions in ways that disqualify them, such as lacking a required license or visa sponsorship eligibility.
Not every company uses every feature. Smaller employers often rely on a system mainly for organization and basic keyword search, while larger corporations with thousands of monthly applicants lean more heavily on automated scoring to narrow the pool before a recruiter spends any time reading. Market consolidation has also shaped how these tools behave.
A handful of large vendors now supply applicant tracking software to a sizable share of major employers, which means the parsing quirks and scoring habits of just a few platforms affect a disproportionate number of job seekers nationwide. When one of these dominant vendors updates its parsing logic or rolls out a new scoring model, the practical effect ripples across thousands of companies simultaneously, often without any public announcement reaching the candidates whose applications are being evaluated differently as a result.
How Keyword Matching Shapes Resumes
Keyword matching remains the single most consequential mechanism in resume screening, and it has reshaped how job seekers write their application materials. Systems typically compare the text of a resume against the text of a job posting, looking for exact or near-exact matches on skills, job titles, software names, and certifications.
A candidate who has done the work described in a posting but used different terminology, calling a role a “team lead” instead of “manager,” for example, can score lower than a less experienced candidate who happened to mirror the posting’s exact language. This has given rise to an entire cottage industry of resume optimization advice built around matching language precisely, sometimes called tailoring a resume to a job description. Career coaches commonly recommend:
- Mirroring exact phrases: using the same wording for skills and job titles found in the posting rather than close synonyms.
- Avoiding tables and columns: since many parsing engines struggle to read multi-column layouts or text embedded in graphics, scrambling the field order.
- Including both acronyms and full terms: writing “search engine optimization (SEO)” rather than only one form, since a system may search for either.
- Matching required years of experience: stating years explicitly next to a skill when the posting specifies a minimum, since some systems filter on that exact number.
Critics of this system argue it rewards a kind of formatting literacy over real capability, effectively adding a hidden skill, knowing how to write for software rather than for a person, as a prerequisite for getting hired at all. Career services offices at universities and nonprofit workforce programs have increasingly built keyword-matching instruction directly into their resume workshops, treating it as a basic literacy skill alongside interview preparation.
This has produced an uneven playing field of its own: job seekers with access to career coaching, paid resume services, or simply a social network that includes recruiters tend to learn these norms faster than those navigating the process alone, adding a layer of inequality to a hiring step that was originally intended to make screening more consistent and less dependent on personal connections.
Ranking Algorithms and Candidate Scores
Beyond simple keyword counts, more advanced platforms use machine learning models trained on historical hiring data to predict which applicants are likely to be interviewed or hired, then rank new applicants against that pattern. These models can incorporate dozens of weighted factors simultaneously, from education level and years of experience to the specific combination of skills listed, the employment gaps present, and even the geographic distance between a candidate’s address and the job location.
The weighting of these factors is rarely disclosed to applicants or even to the recruiters using the software, since vendors generally treat their scoring methodology as proprietary. This opacity has become a point of regulatory concern in several jurisdictions, where lawmakers have begun requiring companies to audit their automated hiring tools for discriminatory patterns before deploying them at scale. A widely reported case involving a major technology company several years ago found that an internal hiring model had learned to downgrade resumes containing the word “women’s,” as in “women’s chess club,” because its training data reflected a historically male-dominated applicant pool, illustrating how a model can replicate patterns in its training data without anyone intending that outcome.
Vendors have since built in various safeguards, including blinding models to names, photos, and other demographic proxies during the initial scoring pass, though critics note that proxies like zip code, school name, or even certain hobbies can still correlate with protected characteristics even after direct identifiers are removed.
A separate concern researchers have raised involves training data drift over time. A model trained on a company’s hiring outcomes from several years ago reflects the preferences, biases, and labor market conditions of that earlier period, and if the model is not regularly retrained on fresh outcomes, it can keep applying outdated assumptions about what a successful hire looks like long after the job itself, the required skills, or the available labor pool have shifted. Some vendors now publish periodic audit summaries describing how often their models are retrained and what fairness metrics are checked, though independent verification of these claims remains limited across the industry.
Common Filters That Reject Qualified Applicants
Several categories of automated rejection happen with enough frequency that recruiting researchers have given them specific names, and they often catch candidates who would otherwise be strong fits for a role.
- Employment gap filters: flag or automatically deprioritize resumes with unexplained gaps longer than a set threshold, penalizing caregivers, people recovering from illness, or those who took time for further study without listing it as formal education.
- Overqualification flags: some systems deprioritize candidates whose experience or prior salary level is judged too high for the posted role, on the assumption they will leave quickly or demand more pay.
- Format rejection: resumes submitted as image files, unusual file types, or heavily designed templates with graphics and icons often fail to parse correctly, resulting in blank or garbled fields that score poorly regardless of content.
- Title mismatch rejection: a candidate whose most recent job title does not closely match the posted role’s title can be scored lower even if the actual duties overlap heavily with the position.
- Location and commute filters: automatic deprioritization or rejection of candidates outside a defined radius of the job site, even for roles that could reasonably be performed remotely or with relocation.
Recruiters themselves are often aware that qualified people fall through these cracks, and many manually review a sample of lower-scored applications specifically to catch strong candidates the algorithm may have missed, though this safety net is inconsistent across companies and depends heavily on recruiter workload.
Seasonal hiring surges make this problem worse. Retail, logistics, and hospitality employers staffing up for a holiday season or a sudden demand spike often receive application volumes far above their normal baseline in a short window, and the pressure to fill roles quickly pushes many of these companies toward stricter automated filtering rather than looser review, since there simply is not enough recruiter time to manually check every rejected file during a surge period.
Industry surveys of recruiting teams have found that a sizable share privately acknowledge their filters reject candidates who would likely perform the job well, viewing the trade-off as an acceptable cost of managing volume rather than a flaw to be fixed immediately.
Strategies for Getting Past the Filters
Career advisors and recruiting researchers generally converge on a similar set of recommendations for applicants trying to navigate these systems, even though no single approach works universally across every vendor’s software.
- Submit as a standard text-based document: a clean, single-column Word or PDF file with conventional section headers parses far more reliably than a visually designed resume.
- Match the posting’s exact terminology: pulling specific skill names, tools, and certifications directly from the job description into the resume body.
- Quantify achievements with numbers: many scoring systems weight resumes containing measurable outcomes, such as percentages or dollar figures, more favorably than purely descriptive bullet points.
- Apply directly through the company site when possible: some evidence suggests direct applications are scored differently or receive more recruiter attention than those submitted through aggregator job boards.
- Use a referral when available: an internal employee referral frequently routes a resume to a recruiter’s direct review queue, bypassing some automated scoring steps entirely.
None of these tactics guarantee an interview, since the underlying job market conditions, the volume of competing applicants, and the specific hiring manager’s preferences all still matter. But they reduce the odds of a qualified candidate being filtered out purely due to a formatting or phrasing mismatch rather than a real lack of fit.
A newer wrinkle involves generative writing tools that candidates now use to rewrite entire resumes in a matter of seconds, often producing text that matches a job posting’s language almost word for word. Some recruiters report that this has made keyword matching alone less useful as a signal, since a large share of applicants now arrive with near-identical phrasing, pushing some employers to weight structured assessments and verified work samples more heavily than resume text in their overall scoring models. The practical effect for job seekers is that resume optimization has become close to table stakes rather than a competitive edge, useful for avoiding an automatic rejection but less reliable as a way to stand out once a human reviewer is looking at a shortlist.
Employer Incentives for Automating Screening
From the employer side, the appeal of automated screening is straightforward: a single posting for a popular role at a well-known company can draw hundreds or thousands of applications within days, far more than any recruiting team could manually review in a reasonable timeframe. Automated scoring lets a small recruiting staff manage a large applicant volume without proportionally expanding headcount, which has become increasingly attractive as companies face pressure to keep hiring costs down even as application volumes rise, partly driven by candidates using their own automation tools to apply to dozens of postings quickly.
This has created something of an arms race: as job seekers adopt tools to mass-apply and tailor resumes automatically, employers lean further into automated filtering to manage the resulting volume, and each side’s tools push the other toward more automation rather than less. Some large employers have responded by adding structured assessments, skills tests, or short video interviews earlier in the pipeline, arguing that behavioral and skills-based evidence filters more fairly than resume text alone, though these additional steps add friction and time for applicants and can themselves be gamed or automated in turn.
Vendors in this space continue to add features, including resume-to-job matching scores visible to recruiters, automated interview scheduling, and increasingly, generative tools that draft outreach messages to top-ranked candidates, extending automation further into the hiring process beyond the initial filter. Some platforms now also offer predictive attrition scoring, estimating how long a candidate is likely to stay in a role based on patterns in their employment history, a feature that adds yet another automated judgment layer recruiters can weigh alongside the original screening score.
Final Thoughts
Automated resume screening has become a near-universal gatekeeper in modern hiring, shaping how millions of job seekers write and format their applications whether they realize it or not. The systems reward precise language matching and clean formatting over narrative storytelling, which creates real friction for candidates whose experience does not map neatly onto a posting’s exact terms.
Employers adopted these tools largely to manage overwhelming application volume rather than to replace human judgment entirely, and most still involve recruiters at some stage of the process. Learning how the filtering layer operates will not guarantee an interview, but it removes one source of invisible rejection from an already competitive job search.
Frequently Asked Questions
Do all companies use the same applicant tracking software?
No. Dozens of vendors compete in this market, ranging from platforms built for small businesses to enterprise systems designed for companies with tens of thousands of employees, and each vendor’s parsing and scoring logic differs, meaning a resume optimized for one system is not guaranteed to perform identically in another.
Can a human recruiter see a resume the system scored poorly?
In most configurations, yes, the resume still exists in the database and a recruiter can search or browse it manually, but a low automated score typically means it sits lower in the default sort order, reducing the odds it gets opened during a quick initial review of a large applicant pool.
Is it true that some resumes get rejected without any person ever seeing them?
It depends on the employer’s configuration. Some companies set knockout questions or score thresholds that trigger an automatic rejection email without recruiter review, especially for high-volume entry-level postings, while other employers require a person to review every applicant regardless of score before any rejection is sent.
Does submitting a resume as a PDF hurt my chances compared to Word format?
Most modern systems parse both formats reasonably well, though a PDF created from a heavily designed template with columns, text boxes, or graphics can still parse worse than a plain Word document. A simple, single-column PDF with standard fonts and headers is typically treated the same as its Word equivalent by current parsing engines.
Are these systems regulated by law?
Regulation is emerging but uneven. Some jurisdictions now require bias audits of automated hiring tools or mandate disclosure to candidates that automated assessment is being used, while many other regions have no specific rules governing these systems beyond general employment discrimination law, leaving enforcement inconsistent across different markets.
Should I list every keyword from a job posting even if it feels repetitive?
Moderation matters here. Including relevant terms naturally within real descriptions of past work tends to score well and reads credibly to a human reviewer, while stuffing a resume with disconnected keyword lists can trigger spam-like flags in some systems and looks obviously artificial to any recruiter who does open the file.