Do 75% of Resumes Get Rejected by ATS? The Claim, Traced
"75% of resumes are rejected by ATS before a human sees them" is probably the most repeated number in resume advice. It is also one that nobody has ever shown the working for. This page follows the claim back as far as the public record goes, says exactly where the trail stops, and then sets out what is actually documented about automated screening.
Where the 75% figure comes from
Every serious attempt to source the number ends at the same company. In an August 2014 advice column, Fortune told a reader that the 75% rejection rate "probably came from" a study by Preptel, a job-search services firm founded by Jon Ciampi, formerly of ATS maker SumTotal Systems. The column links to no study.
In 2020, HR consultant Christine Assaf traced the claim link by link in Your Job Application Was Rejected by a Human, Not a Computer. Her chain ran from a resume-writing site to a 2014 Forbes article that gave no source, and from there to CareerMinds and CNBC, which both attributed the figure to Preptel. Preptel, she found, had closed in August 2013. HiringThing reached the same dead end and notes that a 2019 CNBC piece still ran the figure as its headline claim.
So the honest status is: a number attributed to a defunct vendor, with no report, no sample size and no definition of "rejected" ever made public.
What Preptel itself was saying at the time
Preptel's own pages survive in the Internet Archive, and they are worth reading because they show what kind of claim this was. The company's homepage in January 2011 told visitors: "Less than 1% of resumes make it through a company's resume screen." The next line sold its product, Resumeter, as the way to "beat the machine."
A March 2012 Preptel blog post stated: "Over 80% of all hiring companies use computers to screen and rank résumés." It cited no source for that either. When CIO profiled Preptel in February 2012, the only outcome figure in the piece was the company's own: 50% of its users whose resumes had a "very strong" relevancy ranking landed interviews between September and December 2011. There is no 75% figure in that article.
Read together, these are sales claims from a firm whose business depended on job seekers believing software was the enemy. That does not make every one of them false. It does mean none of them is evidence.
| Source | Year | What it says | Cites a study? |
|---|---|---|---|
| Preptel homepage | 2011 | Less than 1% of resumes get through a resume screen | No |
| Preptel blog | 2012 | Over 80% of hiring companies screen and rank by computer | No |
| CIO profile of Preptel | 2012 | Preptel's own interview rate for high-relevancy users; no 75% | Company data only |
| Fortune advice column | 2014 | 75% 'probably came from' a Preptel study | No |
| Christine Assaf (HRtact) | 2020 | Traces the chain to Preptel, closed August 2013 | Finds none |
What is actually documented about automated screening
The best public evidence is the Hidden Workers report from Harvard Business School's Project on Managing the Future of Work and Accenture. It surveyed 2,275 executives in the US, UK and Germany, plus 8,720 workers, in 2020. Its findings are about filtering by criteria, not a blanket rejection rate:
- More than 90% of employers using a recruiting management system relied on it to "initially filter or rank" candidates: 94% for middle-skills roles and 92% for high-skills roles (HBS/Accenture report).
- 88% of employers agreed that qualified high-skills candidates "are vetted out of the process because they do not match the exact criteria established by the job description." The figure was 94% for middle-skills candidates (same report).
- Only 20% of the hidden workers surveyed said they had ever moved from applicant to candidate (same report).
Joseph Fuller's summary for HBS Working Knowledge puts the takeaway plainly: most employers acknowledged their systems rejected potentially productive applicants. That is a serious finding. It is also a perception survey of employers, not a count of resumes, and it is about criteria such as a missing credential or an employment gap rather than a parser failing to read a file.
How the big platforms say they decide
Vendor documentation shows where an automatic "no" can and cannot happen.
Greenhouse lets employers auto-reject applicants based on their answer to a Yes/No, single-select or multi-select application question, with the example of a role that requires a Class A commercial driver's license. Its AI ranking feature works differently. The Talent Matching FAQ says it "does not auto-reject or auto-advance any candidate, including candidates in lower match categories."
Workday's HiredScore assigns candidate grades from A to D based on how well a resume matches the job's basic and preferred qualifications. Workday's documentation says the grade appears in the recruiter's Spotlight view and Pending Actions dashboard; it is a signal for a person reviewing applications.
These are the systems large employers actually run. In the ATS Resume AI analysis of all 503 S&P 500 careers sites and 9,260 live job postings, June 2026, Workday was the most common identifiable system at 87 companies (17.3%), and Greenhouse appeared at 23 (full breakdown).
What to do instead of fearing a number
The useful version of the myth is the one the HBS data supports: exact criteria matter. Run this before each application.
- Answer knockout questions truthfully and precisely. On Greenhouse, an auto-reject rule fires on the answer you pick, so read each question for what it literally asks.
- Mirror the posting's required terms where they are true. A recruiter's keyword search and a match grade both start from the posting's wording. Our guide to ATS resume keywords covers how to pull them out.
- Check that your file parses. Save it as plain text and read it back; if sections collide, fix the layout. See what a resume parser does.
- State the criteria a filter can check. Credentials, licences and years of experience should be written out in full, not implied.
- Apply where you meet the basics. A Grade B in HiredScore's scheme means meeting all basic qualifications, which is a realistic target for every application you send.
If you want a faster read on step 2 and 3, the free resume checker needs no account.
Sources
- Fortune, "How to keep your resume from getting lost in cyberspace" (2014)
- Christine Assaf, HRtact, "Your Job Application Was Rejected by a Human, Not a Computer" (2020)
- HiringThing, "Applicant Tracking Systems Aren't Excluding Job Applicants—People Are"
- Preptel homepage, January 2011 (Internet Archive)
- Preptel blog, "Technology: foe or friend?" (2012, Internet Archive)
- CIO, profile of Preptel (2012)
- Harvard Business School and Accenture, "Hidden Workers: Untapped Talent" (2021, Internet Archive copy)
- HBS Working Knowledge, Joseph Fuller (2021)
- Greenhouse Support, "Auto-reject"
- Greenhouse Support, "Talent Matching FAQ"
- Workday HiredScore documentation, "Reference: Candidate Grades"
Frequently asked questions
- Who first published the 75% ATS rejection statistic?
- No original publication has been found. Fortune (2014) and later investigators attribute it to Preptel, a resume-software firm that closed in August 2013, but no Preptel report, sample or method has surfaced.
- Is there a real ATS rejection rate I can quote?
- Not a credible overall one. The closest documented figure is from the 2021 Harvard Business School and Accenture survey, where 88% of employers said qualified high-skills candidates are screened out for not matching exact job-description criteria. That is an employer opinion, not a count of rejected resumes.
- Can an ATS reject me automatically at all?
- Yes, when an employer configures it. Greenhouse, for example, can auto-reject based on an applicant's answer to a yes/no or multiple-choice application question. Its AI match scores, by contrast, do not auto-reject anyone, according to its own documentation.
- Why does the 75% claim keep spreading?
- It is easy to remember and it sells resume services. Christine Assaf's 2020 trace found it repeated across career sites and news articles, each citing another secondary source rather than data.