Do ATS Systems Automatically Reject Resumes? The 75% Stat Is a Myth

Mostly, no. Applicant tracking systems rarely reject a resume on their own. In a 2025 study that interviewed 25 US recruiters across more than ten platforms, including Workday, iCIMS, Greenhouse, and Lever, 23 of the 25 said their systems never automatically reject a resume over formatting, design, or missing keywords. Automated rejections do happen, but they come from eligibility questions the recruiter configures, like work authorization, not from software judging your bullet points. And the famous claim that 75% of resumes are rejected before a human ever sees them has no study behind it at all.

Where the 75% statistic actually came from

Researchers who went looking for the origin of the 75% figure traced it to a 2012 sales pitch by Preptel, a resume-optimization startup that folded in 2013 without ever publishing a methodology. No data set, no replication, no paper. The number survived anyway because it sells: it makes the hiring process feel like a rigged machine that only a paid tool can unlock. When recruiters in the 2025 survey were asked where they had encountered the claim, they pointed to social media (68%), career coaches and resume services (20%), and unverified media coverage (12%).

One honest caveat before going further. The survey behind these numbers is small: 25 recruiters, interviewed qualitatively, by a resume-tool vendor (Enhancv). A percentage like 92% means literally 23 people. Treat the figures as directional, not definitive. That said, the finding runs against the vendor's own commercial interest in ATS fear, it was corroborated by independent outlets, and a targeted search for any evidence supporting the 75% claim came up empty. The direction is clear even if the precision is not.

What an ATS actually does with your resume

An applicant tracking system is closer to a filing cabinet with search than to a gatekeeper. When you apply, a parser runs a staged pipeline: it segments your resume into standard sections (work experience, education, skills), categorizes the contents into typed fields (recognizing that "January 2018 to April 2022" is a date range and "Software Engineer" is a job title), and indexes the result so recruiters can search and filter candidates. The resume is stored, structured, and surfaced. A human does the rejecting.

Match scores against the job description are real in some systems, but they are not a universal gate. Workday can rank candidates by a descending match percentage. Greenhouse instead uses a human-graded scorecard, not an algorithmic numeric rank, and Lever uses semantic relevance with no candidate-facing score. In the same 25-recruiter survey, about 44% of recruiters had some form of AI fit score available, and of those, more than half ignored or disabled it; only around 8% (two people, on Bullhorn and BambooHR) ran systems configured to auto-reject on a low match or missing skills. For most recruiters the score is a rough sorting aid before manual review, advisory rather than authoritative.

The rejections that are automatic: knockout questions

There is one place where the rejection genuinely is instant, and it has nothing to do with how your resume is written. Recruiters can configure knockout questions on the application form itself: are you authorized to work in this country, do you hold the required license, can you work from this location. Answer outside the requirement and the system filters you out before review. In the survey, every recruiter whose platform supported knockout questions used them, and most named them as the dominant automated filter. If you have ever received a rejection minutes after applying, this is the likely mechanism: a form answer, not a keyword scan.

The real risk: parse failure, not robot rejection

The way software actually hurts a resume is quieter than rejection. If the parser cannot extract your data cleanly, the structured profile a recruiter searches is incomplete: a missing phone number, a scrambled job history, an empty skills field. Nobody rejected you. You just became harder to find and harder to evaluate. The checkable evidence on what breaks parsing is mostly small-scale testing rather than vendor disclosure, but it converges on a consistent set of risks:

• In one test of 8 ATS platforms, a two-column layout parsed incorrectly in 7 of the 8. Naive parsers read top to bottom across the full page width and interleave the columns.
• In the same 8-system test, content in tables was skipped by 5 of 8 systems, and sidebar text boxes were dropped entirely by 4 of 8.
• Contact details placed in the document header element (rather than the body) were missed by 6 of the 8 systems, producing candidate profiles with no name or phone number. Some enterprise parsers skip the header and footer regions of a Word file altogether.
• Text embedded in graphics is ignored, and decorative bullet glyphs (checkmarks, arrows, diamonds) can be replaced with garbage characters or cause bullets to merge into one paragraph.

Two qualifiers keep this honest. First, these are small tests, not industry-wide audits; modern layout-aware parsers handle native word-processor columns better every year, so treat layout complexity as graduated risk (text boxes worst, then tables, then columns), not as an automatic failure. Second, the old PDF-versus-Word panic is largely outdated: current commercial parsers accept PDF, DOCX, and a long list of other formats. The format matters far less than whether the content inside it reads as one clean column of real text.

What this means for how you write a resume

Stop optimizing against an imaginary rejection robot, because the tricks aimed at it backfire on the humans who actually read the output. The white-text keyword hack is the clearest example: parsers extract text without regard to color, so the hidden block often appears in plain view on your candidate profile, recruiters uncover it with a simple select-all, and a 2026 iCIMS study found resumes flagged for manipulation were 67% less likely to advance. You are not fooling the machine. You are handing a human a reason to doubt you.

What the evidence supports instead is unglamorous: make your resume easy to parse and honestly relevant. Use a single column with standard section headings. Keep your name and contact details in the document body, not the header. Write dates as plain date ranges. Use the actual vocabulary of the job description where it truthfully describes your experience, because relevance affects where you surface in a recruiter's search and sort, even though it almost never triggers a rejection. The ATS is not deciding your fate. It is building a record of you for the person who will. Make sure that record arrives complete.

Sources

• Enhancv, 25-recruiter study on ATS auto-rejection (2025) ↗
• ITBrief coverage: study debunks the 75% ATS rejection myth ↗
• QuickCV, parse test across 8 ATS platforms ↗
• Prosple on the white-font resume hack and match scores ↗