What really happens after you apply

Modern parsing systems can often detect columns and reconstruct reading order. Textkernel reported that a newer layout model increased well-rendered column resumes from 62% to 90% in an internal evaluation. That is evidence that column parsing is technically possible, not proof that every ATS handles every resume correctly. S1

Independent research points the same direction while confirming the difficulty: layout-aware parsing systems can normalize diverse resume formats, but layout heterogeneity remains a genuine technical challenge. S14

Parsing can still fail because of:

• A poor internal reading order
• Text boxes or tables constructed unpredictably
• Important content in headers or footers
• Icons used without text labels
• Scanned or image-only PDFs
• Complex sidebars
• An older or weaker parsing engine
• The specific PDF generator used to export the file

• Use selectable text.
• Keep contact information in the document body.
• Avoid relying on icons, charts, or graphics for essential information.
• Copy the finished PDF into a plain-text editor and verify reading order.
• Keep a conservative single-column version for government, legacy, or especially risk-sensitive applications.

AI-assisted recruiting tools are real and becoming more common. They can summarize resumes, evaluate evidence against employer-defined criteria, prioritize applications, detect possible fraud, and help recruiters decide what to review. S2 S5 Survey research confirms AI use in HR is rising broadly. S13

That does not mean a generative-AI model independently makes every rejection. Workday states that its AI is not designed to automatically reject candidates. Ashby describes a workflow where AI analyzes evidence against criteria while a recruiter reviews and takes the action. S2 S5

At the same time, automatic rejection absolutely exists. Greenhouse and Workable both document employer-configured rules that automatically reject or disqualify applicants based on application answers. S3 S4

Common knockout topics include:

• Work authorization or sponsorship
• Required location
• Schedule availability
• Salary constraints
• Required licenses or certifications
• Security clearance
• Willingness to travel
• Another employer-defined non-negotiable

Treat application questions as part of the evaluation. Read them carefully and answer accurately. An immediate rejection does not prove an AI model analyzed and disliked the resume.

Investigation traced the figure to a 2012 sales pitch by Preptel, a resume-optimization company that ceased operating the following year. No methodology, sample size, or peer-reviewed source was ever published. S26

The mechanism it describes is also backwards. Most applicant tracking systems rank and sort candidates rather than silently discarding them; the documented automatic rejections are employer-configured rules keyed to application answers, not resume content. S3 S4

Treat any advice built on this statistic with suspicion, including advice from tools that sell a fix for it. Spend the effort on fit, timing, and screening answers instead.

Truthful role terminology can make relevant experience easier to search, parse, and understand. That does not make keywords a secret password. Parsing a resume and ranking a candidate are different operations, and the systems doing the former are not usually the ones making the decision. S1 S14 S26

A resume may be read by several systems and people:

• A parser extracting structured fields
• A recruiter using exact search or filters
• An AI tool evaluating evidence against criteria
• A recruiter scanning quickly
• A hiring manager evaluating depth and outcomes

Keyword stuffing may satisfy a simplistic consumer scanner while making the resume less credible to the human readers who matter.

• Use the employer's terminology when it truthfully describes prior work.
• Expand uncommon acronyms at least once.
• Connect skills to real projects, scope, and results.
• Do not hide white text or repeat phrases unnaturally.
• Do not invent experience to imitate the listing.

The official documentation shows why a universal score is not structurally possible: the rules that actually disqualify candidates are configured per employer, per job. S3 S4 Review workflows likewise differ by team and volume. S5

Employers use different:

• Screening questions
• Filters
• Required criteria
• Recruiter workflows
• Integrations
• AI tools
• Applicant pools
• Human reviewers

Some tools score candidates. Some evaluate separate criteria. Some do neither. Consumer resume scanners can be useful diagnostics, but their percentage is generated by that product, not retrieved from the employer's ATS. S26

Ask whether a scanner's recommendation makes the resume clearer and more truthful. Do not chase a percentage as though it guarantees review or an interview.

An immediate rejection may result from:

• A knockout question S3 S4
• A location or authorization rule S3 S4
• A duplicate-application rule
• A closed or paused requisition
• A bulk status change
• A manually triggered rejection paired with an automated email S26

Review the screening answers and hard requirements before rebuilding the entire resume.

Ashby reports that every hire required more than 300 applications on average throughout 2025. Greenhouse reported that applications per job increased 102% between Q3 2022 and Q4 2024. S6 S7 Employers have responded by asking more of each applicant: application questions nearly doubled between 2021 and 2024. S10

• Prioritize fresh, credible, high-fit roles.
• Track which applications produce recruiter responses.
• Spend less time on weak or uncertain opportunities.
• Preserve enough time for interview preparation, networking, and follow-up.

We win when you apply to fewer jobs—because you found better ones to apply to.

There is no universal "first 50 resumes" cutoff.

However, Ashby describes chronological review as a traditional workflow and gives a real recruiter example where only 10% to 15% of a large applicant pool might otherwise be reviewed. S5 With more than 300 applications per hire on average, the gap between 'submitted' and 'reviewed' is where timing does its work. S6

When an employer starts reviewing immediately and stops after building a sufficient interview slate, timing affects whether a qualified application enters the seriously reviewed group.

Apply promptly to strong matches without sacrificing accuracy. Use alerts and prioritize fresh listings.

Job descriptions often mix:

1. Hard constraints — work authorization, required license, schedule, location, clearance
2. Core capabilities — the work the person must be able to perform
3. Preferences — ideal domain exposure, tools, or adjacent experience

The mistake is assuming every bullet is mandatory—or that none of them matter.

On the famous statistic: the widely repeated claim that men apply when they meet 60% of requirements while women wait for 100% has no research behind it. It traces to a speculative remark attributed to a Hewlett-Packard executive, not to data, and spread through business books and business press until it read as established fact. When the Behavioural Insights Team actually tested it with more than 10,000 job seekers, the real numbers were 52% for men and 56% for women — a four-point gap in the opposite direction from the folklore, present only among less-qualified participants. S20 Peer-reviewed work has examined the same claim. S21

Apply when the core match is credible, the hard constraints are satisfied, and any gaps can be explained honestly. Do not use a percentage — yours or a statistic's — as the decision rule.

Remote work removes geographic barriers for applicants, but it also expands the employer's candidate pool. Ashby's startup-hiring data shows materially higher inbound volume for remote jobs. S11 That sits inside a broader rise in applications per job. S7

Expect remote roles to require faster response, stronger differentiation, clear evidence of independent work, and confirmation that the employer can hire in your location.

A field experiment across more than 7,000 applications found tailored cover letters produced a 16.4% callback rate, generic letters 12.5%, and no letter at all 10.7%. The gap between tailored and generic is larger than the gap between generic and nothing. S25

A cover letter can be useful when it adds information the resume cannot efficiently communicate:

• A non-obvious career transition
• A specific connection to the company's problem
• Relevant context not visible in job titles
• Motivation that is genuinely specific
• Evidence of written communication

Generic praise usually adds little. As AI-generated tailoring becomes common, polished language may also become a weaker signal by itself — research on cover letters in the AI era found that AI assistance improved alignment and callbacks while simultaneously eroding the value of tailored language as a distinguishing signal. S16

Write one when requested or when it adds meaningful context. Do not spend substantial time producing a generic letter for every role — the evidence says it buys little. S25

More than 300 applications per hire does not mean every job receives exactly 300 applications, but it demonstrates the scale of the inbound-capacity problem. S6 Recruiter accounts describe reviewing a fraction of large pools under time pressure. S5

Some employers review every application. Others do not. The workflow depends on pool size, specialization, staffing, regulations, and hiring urgency. S26

Improve the probability of review through fit, freshness, clarity, and credible context. Do not interpret silence as proof that a recruiter carefully compared your resume against the entire pool.

The famous number comes largely from small eye-tracking studies. A 2018 Ladders study involving 30 recruiters reported an average initial screening time of 7.4 seconds. S15

That supports the broader conclusion that initial scans can be extremely fast. It does not establish a universal duration for every recruiter, role, industry, or hiring stage — and the 'six seconds' version in circulation is not even the number the study reported. S26

Resume review often has two modes:

1. A rapid relevance scan
2. A deeper review of candidates who appear promising

Make role, level, recent experience, scope, and strongest evidence easy to find quickly while preserving enough substance for deeper review.

Hiring is not a complete tournament in which every possible applicant is objectively ranked.

The strongest evidence here is unglamorous: hiring methods are not very predictive to begin with. A major 2022 re-analysis of personnel-selection research found that decades of published validity estimates had been systematically overstated, and that even the best-performing method — the structured interview — correlates with job performance at roughly r = .42, with most common methods lower. S23 An employer cannot select 'the best candidate' with instruments that imprecise, regardless of intent.

Outcomes are also affected by:

• Who was found and reviewed S8
• Timing
• Referrals and internal candidates S8
• Compensation
• Location
• Interview availability
• Team preferences
• Changing role needs
• Whether the employer can close the candidate

And because employers increasingly share the same screening infrastructure, the same applicants can be filtered out repeatedly across many jobs — a pattern researchers call algorithmic monoculture. S19

Stop treating each rejection as a verdict on your ranking against the field — the process does not produce that ranking. Focus on entering more processes where you are visible early and the fit is genuine, which is the part you can influence.

"Qualified" is not the same as "selected," and "not selected" is not the same as "unqualified." The imprecision of selection methods means a rejection carries much less information about your ability than it feels like it does. S23

A qualified candidate can lose because:

• Another candidate was already deep in process
• A referral or internal candidate received earlier attention S8
• The role changed or was paused
• The company preferred a narrower background
• Compensation did not align
• The candidate applied after the viable slate was already built
• The application was never meaningfully reviewed S9

Look for patterns across multiple applications. One rejection provides little information. Repeated failure at the same stage can reveal a useful problem with targeting, screening answers, resume communication, interview performance, or compensation.

A candidate may be rejected through:

• A knockout rule S3 S4
• Recruiter application review S5
• A bulk action
• An AI-assisted criteria workflow S5
• A role closure
• Another stage before hiring-manager review

Do not interpret a generic rejection as a detailed judgment from the entire organization. The hiring manager may never have seen the application.

Ashby found that roughly 40% of referred candidates reached an interview, compared with roughly 3% of inbound applicants. S8

A referral does not guarantee fit or advancement. It does provide an additional trust and attention signal — and there is independent reason employers value it. Peer-reviewed research across nine large firms found referred workers were 10-30% less likely to quit and performed better on rare high-impact outcomes, with the benefit driven mainly by better job fit rather than referred workers being uniformly stronger. S22

Use referrals where a genuine professional connection exists. Give the referrer a concise explanation of the fit rather than asking them to endorse experience they cannot evaluate.

Inbound applications continue to generate the largest share of hires in aggregate, reaching 52% of hires in Ashby's Q2 2025 data. S9

Both of these can be true:

• Cold applications generate many hires overall. S9
• Each individual inbound application faces low odds. S6

Keep applying directly to strong-fit jobs, but do not rely on inbound applications as the only channel.

Internal, referred, sourced, agency, and inbound candidates enter different workflows and convert at different rates — the referred-versus-inbound gap alone is roughly 40% against 3%. S8 S9 Some public roles already have internal candidates, referrals, or sourced prospects in progress.

Shared screening infrastructure can compound this: when many employers rely on similar tools, the same candidates may be repeatedly excluded across the market rather than getting independent looks. S19

Treat the posting as one channel rather than the whole game. Where a real connection exists, use it early — before the slate fills. Where it does not, prioritize freshness and fit, which are the levers actually available to an inbound applicant.

Two independent methods agree ghost jobs are a minority of postings. Greenhouse classifies 18-22% of jobs posted on its platform as ghost jobs in a given quarter, and 3 in 5 candidates believe they have encountered one. S24 A separate February 2026 analysis of 176,268 Indeed listings across 49 industries and every U.S. state put the figure at roughly 1 in 7 (about 14%), rising to 21% for senior-level postings and 51% in wholesale. S29

The two estimates differ because they measure different populations with different definitions. Report the range, not a single number. S30

An old or reposted listing may also be:

• A genuine slow-moving search
• An evergreen pipeline
• A reopened role
• A duplicate or syndicated listing
• An automatically refreshed posting
• A paused role left online
• A role with interviews already underway
• A role that ultimately closes without a hire

Use multiple signals: original posting date, employer career-site presence, recent company hiring activity, role specificity, and continuous reposting behavior.

Roughly a fifth of postings on one major ATS platform are classified as ghost jobs each quarter S24, and an independent analysis of 176,268 Indeed listings found about 1 in 7 live postings were ghost jobs, with 4% still active four months or longer. S29

A listing can remain visible while:

• Hiring is paused
• Finalists are already interviewing
• Budget approval changes
• The role is being redefined
• A syndication partner has not removed it
• The company is collecting future candidates

Employer surveys explain why. Hiring managers report keeping listings up because the company is 'always open to new people', in case an exceptional candidate applies, or to project growth they are not actually experiencing. S30

Favor verified, recent listings and treat freshness as important—but not perfect—evidence of employer intent.

Silence is usually not deliberation. Greenhouse found 61% of job seekers reported being ghosted after interviews, up nine points in eight months, and that underrepresented candidates experienced it at higher rates (66% versus 59%). S24 Separate survey work puts the share of job seekers ghosted by an employer at over half. S28

Silence may mean:

• Rejection
• Delay
• Internal disagreement
• A paused role
• An overloaded recruiter
• A process that was never formally closed

Follow up once when a legitimate contact exists, then continue the search. Do not put other opportunities on hold because an employer failed to communicate.