You have probably seen the statistic: applicant tracking systems discard three quarters of resumes before a recruiter ever opens them. The short answer for 2026 is that no verifiable study supports that number. It traces back to a single resume-optimisation company that went out of business in 2013, and the consultant who went looking for its origin found no supporting research. What the evidence does show is narrower and more useful: employers' software genuinely does rank and filter candidates, knockout questions genuinely do reject people instantly, and bad formatting can make a resume parse badly. But parsing failures and automatic rejection are different problems, and this article keeps them separate — with a source attached to every claim. For the formatting side, see our ATS-friendly resume format guide.
Where the 75% number comes from
The figure is usually traced to Preptel, a resume-optimisation company. Christine Assaf, an HR consultant who researched its origin, described the source as "seemingly devoid of any study, survey, or context", and reported that she "didn't find any conclusive statistical evidence" behind the number; Preptel went out of business in August 2013 (HiringThing). CNBC repeated the figure in 2019 with credit to Preptel, which is how a defunct company's marketing number hardened into career-advice folklore (HiringThing).
The number still spreads from sources you might expect to know better. Utah Valley University's 2026 handout on ATS-friendly resumes states that "70-80% of resumés never make it through to a hiring manager" without citing a source for the range (UVU Career & Internship Center). A career office repeating an unsourced figure is a useful reminder: ask where a statistic came from before you let it scare you.
What recruiters actually report
The most recent named study we found comes from Enhancv, published in November 2025: structured interviews with 25 U.S. recruiters in September and October 2025, covering more than ten ATS platforms (Enhancv). In that sample, 92% (23 of 25) said their ATS does not auto-reject resumes for formatting, content, or design. Two recruiters (8%, on Bullhorn and BambooHR) had AI or content auto-rejection configured. 44% had an AI "fit" score, but only 8% used it definitively — 36% used it only as a guide.
Read that with its stated limits. The sample is small and non-random, the authors themselves call the findings "themes, not statistics", the page reports knockout-question usage as 84% in one place and 100% in another, and Enhancv sells resume tools (Enhancv). It is evidence about a handful of real recruiting setups, not a census of the industry — which is still more than the 75% figure can say.
What the ATS vendors say
Greenhouse, among the most widely used ATS platforms, states plainly that "AI never automatically advances or rejects candidates" and that its platform "prohibits AI from making automated dispositions" (Greenhouse, AI/ML Security & Privacy). Vendor documentation is a product promise rather than an independent audit, but it is the closest thing to an official position — and we found no major vendor documenting auto-rejection of resumes over formatting.
What is actually real
None of this means software plays no role in who gets seen. The Harvard Business School and Accenture report "Hidden Workers: Untapped Talent" (September 2021) surveyed 8,720 hidden workers and 2,275 executives across the US, UK and Germany; among employers using a recruitment management system, most relied on it "to make a first cut or rank" middle-skills (94%) and high-skills (92%) applicants (HBS Working Knowledge, report PDF). Ranking and filtering is real. The cartoon version — a robot binning three quarters of resumes unread — is not what the source describes.
HR Dive's summary of the same report adds the mechanism: 88% of employers said qualified high-skills candidates were sifted out because they didn't exactly match the criteria in the posting (94% for middle-skills), and the report blames over-restrictive "negative filters" — excluding anyone without a degree, for instance — rather than formatting (HR Dive). Your application's fate hangs far more on literal criteria matching than on layout choices.
The rejections that really are automatic come from knockout questions — the yes/no screens on work authorisation, certifications or location. In the Enhancv sample, 84% of recruiters used them (Enhancv), and Oracle's Taleo documentation describes "disqualification questions … used in requisitions to screen candidates" as a standard feature (Oracle Taleo docs). Misanswer one — or answer one honestly in a way the employer configured as disqualifying — and the rejection is instant by configuration. That is the one place "automatic" is documented.
Mobley v. Workday: what the lawsuit does and doesn't say
The litigation often cited alongside the viral number deserves precision. In Mobley v. Workday (N.D. Cal., No. 23-cv-00770-RFL), the court conditionally certified an age-discrimination collective on 16 May 2025, finding a plausible "unified policy: the use of Workday's AI recommendation system to score, sort, rank, or screen applicants" (Proskauer's summary of the order). Workday stated that "1.1 billion applications were rejected" using its tools in the period. Note what the source does not say: whether people or automation made those rejections. And conditional certification is a procedural step, not a ruling on the merits.
Parse failures are not rejections
Here is the distinction this whole topic turns on: a resume that fails to parse correctly is not the same as a resume that gets rejected. Greenhouse documents a list of unsuccessful-parse causes — graphics, tables, headers and footers, columned layouts, contact details in a text box (Greenhouse) — but that page describes what stops the file from being read correctly, not an automatic rejection. Most frightening ATS advice online conflates "this can break parsing" with "this gets you auto-rejected". The sources support the first claim and — knockout questions aside — not the second.
| Claim | What the source says | Source type |
|---|---|---|
| The myth that "75% of resumes are rejected by ATS before a human sees them" | Traced to Preptel, which folded in 2013; the consultant who searched for the origin found no supporting study | Practitioner trace (myth) |
| ATS auto-rejects resumes for formatting | 92% of 25 recruiters interviewed in the Enhancv 2025 study said theirs does not; 8% had it configured | Study (small sample) |
| AI makes the decision | Greenhouse: "AI never automatically advances or rejects candidates" | Official vendor doc |
| Employer software makes a first cut | HBS/Accenture 2021: 94% (middle-skills) and 92% (high-skills) of surveyed RMS-using employers rely on it to rank or cut | Study |
| Knockout questions reject automatically | Taleo documents disqualification questions; 84% of the Enhancv sample's recruiters use knockouts | Official doc + study |
| Workday auto-rejected 1.1 billion applications | The figure appears in Mobley v. Workday; the source does not say whether people or automation made the rejections | Law-firm summary |
What to do instead
If robot rejection is mostly folklore, where should the effort actually go? Where the documented mechanisms are:
- Mirror the job description's exact keywords — literal criteria matching is the filter the evidence actually supports (how to find resume keywords)
- Answer knockout questions (work authorisation, certifications, location) truthfully and carefully — they are the one documented instant-rejection path
- Keep the layout parseable: single column, contact details in the body, no images standing in for text — the ATS-friendly format guide covers the full checklist
- Apply even when you don't meet every criterion — the HBS/Accenture data shows exact-match filtering discarding qualified people, so a strong-but-imperfect match is worth submitting
How to check your own resume
You can test both halves of the distinction yourself in a couple of minutes. Run your resume through the free resume checker to see how a parser reads it — whether your contact details, job titles and dates survive extraction. Then paste in the target job description and use the matcher to see which of its keywords your resume is missing. A resume that extracts cleanly and mirrors the posting's terms is doing everything the documented evidence says matters; the 75% number was never the thing to fear.
Sources
- HiringThing — Applicant tracking systems aren't excluding job applicants, people are
- Enhancv — Does ATS reject resumes? (2025 recruiter study)
- Greenhouse — AI/ML Security & Privacy
- Greenhouse — Unsuccessful resume parse
- HBS Working Knowledge — How to tap the talent automated HR platforms miss
- Hidden Workers: Untapped Talent — report PDF
- HR Dive — ATS may exclude whole segments of candidates
- Proskauer — Mobley v. Workday reaches conditional certification
- Oracle Taleo — disqualification questions
- UVU Career & Internship Center — Creating ATS-friendly resumes (PDF)