Skip to content
Jobsearch.ing

Claim under review

75% of resumes are rejected by an ATS before a human ever sees them.

UnverifiedReviewed 2026-08-22

Short answer

We could not trace this figure to any study, dataset or survey. It is repeated across career-advice sites, resume services and news articles, but the citation trail leads to other articles rather than to a measurement, and the widely repeated origin story about a defunct resume-optimisation vendor is itself uncited. Separately, the number describes something no available source measures: employers do report using software to filter and rank applicants, but no primary source we found reports a share of resumes rejected automatically. Treat the 75% as unsourced rather than as either true or false.

Evidence status

Unverified

We could not trace the claim to any identifiable primary source. This is a statement about the evidence, not about the world: an unverified claim is not thereby false, it is unsupported, and we say so rather than guessing.

What we found

The claim appears in effectively identical wording across resume-writing services, career-coaching sites and syndicated advice columns. Where those pages cite anything, they cite each other, or an intermediate article that itself cites nothing.

A commonly repeated account attributes the figure to a 2012 marketing claim by a resume-optimisation vendor that closed the following year. We were unable to locate any primary document from that vendor stating the figure, and the pages asserting this origin do not cite one either. We therefore record it as an unverified account of an unverified number rather than as the answer.

The strongest genuine evidence about automated screening comes from Harvard Business School and Accenture's Hidden Workers study. It reports that more than 90% of surveyed employers use a recruitment management system to filter or rank candidates, with 94% for middle-skills and 92% for high-skills roles. That is a statement about employers configuring software, not about a proportion of resumes being discarded.

The same report does document genuine automatic exclusion, but of a specific and much narrower kind: just under half of the companies surveyed automatically screen out a resume showing an employment gap of more than six months, on that basis alone.

Earliest appearance we could trace

Not established. Our earliest located appearances are secondary career-advice articles that already treat the figure as common knowledge, which is characteristic of a claim that entered circulation without a citable original.

Where the claim appears

What each page offers as its source. A page citing another article rather than a study is how an unsourced figure becomes common knowledge.

  • Harvard Business School — Managing the Future of Work2021-09-04

    Employers report using recruitment management systems to filter or rank candidates before review.

    Cites: Its own survey. Included here as the source frequently mis-cited in support of the 75% figure, not as an instance of the claim.

    View source page ↗

What was actually measured

What was measured
Nothing, in the case of the claim itself. The adjacent evidence measures how employers configure hiring software and what proportion of employers apply particular automatic exclusions.
Population
The claim states no population. The nearest evidence covers employers in the United States, United Kingdom and Germany.
Sample
No sample exists for the 75% figure.
Time period
Not stated by the claim.

What the evidence supports

  • Software is genuinely involved before a human reviews most applications at larger employers.
  • Some resumes are excluded automatically on specific configured criteria, such as an employment gap of more than six months.

What it does not support

  • That 75%, or any particular percentage, of resumes are automatically rejected.
  • That software makes hiring decisions. Filtering and ranking order a list; a person still works through it.
  • That reformatting a resume will change an outcome by any measurable amount. No source we found measures that.

Limitations

  • Absence of evidence is not evidence of absence. Some employers may well auto-reject at high rates; nobody has published a figure we could verify.
  • Vendor platform data exists but is not comparable across employers, and vendors have a commercial interest in how these numbers are read.

Jobsearch.ing interpretation

Our reading, marked as ours. Take the evidence above and reject this if you read it differently.

The interesting question is not whether the number is right but why it is so durable. It is durable because it is useful: it explains an unexplained rejection without blaming the applicant or the employer, and it sells a remedy.

Our position is that the honest answer to "why did nobody reply" is usually volume, not formatting. That is a less satisfying answer and it does not come with a product attached.

Sources

The evidence packets behind this article. Each was read directly, at the table or section named in its details.

  1. Harvard Business School and Accenture

    Academic · Tier 1 · Sample

    8,000+ workers and 2,250+ executives, US, UK and Germany, surveyed 2020.

    Joseph B. Fuller, Manjari Raman, et al.

    Hidden Workers: Untapped Talent

    Supports

    Most large employers use software to filter or rank applicants before human review.

    Does not

    It never measures a share of resumes rejected — only how employers configure their systems.

    Evidence details
    Supports (full)
    More than 90% of surveyed employers use a recruitment management system to filter or rank candidates before a person reviews them, and just under half apply an automatic exclusion to resumes showing an employment gap of more than six months.
    Does not support (full)
    The report does not measure what share of resumes or applications are rejected, automatically or otherwise. It reports how employers configure their systems and what employers believe about the result, not an observed rejection rate. Nor does it support any claim that software makes a final hiring decision.
    Measure
    Rate
    Counts
    Employers using a recruitment management system to filter or rank divided by Employers surveyed
    Unit
    Employer
    Population
    Employers and executives with hiring responsibility, alongside workers classified as 'hidden' — those seeking work or more hours who are screened out of consideration
    Geography
    United States, United Kingdom, Germany
    Period
    Survey fielded 2020; report published September 2021
    Sample
    2,250 observations
    Method
    Cross-national survey of more than 8,000 hidden workers and more than 2,250 executives, combined with a labour-market model estimating the size of the hidden-worker population from US Current Population Survey categories.
    Found at
    Filtering and ranking figures in the section on hiring systems; employment-gap exclusion in the section on negative selection criteria
    Quoted
    more than 90% of employers in our survey use their RMS to initially filter or rank potential middle-skills (94%) and high-skills (92%) candidates
    Version read
    HBS white paper PDF, published 3 September 2021
    Notes
    The report says RMS or ATS rather than ATS alone. A recruitment management system is the broader platform; conflating the two is one route by which the report gets misquoted.
    Record
    Published Sep 4, 2021 · Accessed Aug 22, 2026 · Archived copy
  2. CareerPlug

    Industry report · Tier 2 · Complete population

    10M+ applications · 60,000+ US small businesses · 2023

    2024 Recruiting Metrics Report: What We Learned About Hiring in 2023

    Supports

    Employers invited an average of 2% of applicants to interview, ranging from 1.2% in retail to 4.9% in education and child care.

    Does not

    Any individual job seeker's chance of an interview, or any population beyond US small business.

    Evidence details
    Supports (full)
    Across more than 10 million applications to over 60,000 small businesses in 2023, employers invited an average of 2% of applicants to interview, ranging from 1.2% in retail to 4.9% in education and child care.
    Does not support (full)
    Does not describe any individual job seeker's chance of an interview: the denominator is every application an employer received, not applications from comparable candidates. The report describes no screening, de-duplication, or quality filtering applied to that denominator. Says nothing about large employers, professional or knowledge roles, or markets outside the United States.
    Measure
    Rate
    Counts
    Applicants invited to interview divided by All applications received for a posting
    Unit
    Applications received by an employer
    Window
    Calendar year 2023
    Population
    Applications received by small businesses using CareerPlug's applicant tracking system
    Geography
    Not stated in the report. The vendor's customer base is United States small business.
    Period
    Calendar year 2023
    Method
    Complete records of hiring activity passing through one applicant tracking system. No sampling, weighting, or exclusion criteria are disclosed beyond the volume analysed.
    Found at
    Page 8, hiring-funnel summary; per-industry benchmark pages
    Quoted
    employers invited an average of just 2% of applicants to interview for their open roles
    Notes
    Published by an applicant-tracking vendor, and the report also serves as marketing for that product. The page 8 summary and the per-industry pages disagree on the retail figure; the per-industry page is used here.
    Record
    Published Apr 1, 2024 · Accessed Aug 21, 2026

Products sold on this claim

Software in this category is commonly marketed on the strength of this statistic. That is not an accusation against any particular product — it is context for anyone deciding whether to pay for one.

Cite this

If you are citing the underlying figure, cite the original source — not this page. This review is a secondary analysis, and citing it in place of the primary source would repeat the mistake it documents.

Cite this review for the analysis: the source trail, the evidence status, and what the evidence does and does not support.

First published
2026-08-22
Last updated
2026-08-22
Evidence last checked
2026-08-22
Reviewed by
Maid Dizdarevic

How evidence statuses are defined, and what each one commits us to, is set out in the methodology.

Related