There is no measured ATS rejection rate, and the figure most often quoted for one, that an applicant tracking system rejects 75% of resumes before a human ever sees them, has never been shown to come from a study. The trail leads to a resume optimization vendor called Preptel around 2012, and to a 2014 Fortune column that said the number "probably" came from there. No method, sample or dataset has been published since. What does exist is better evidence about where applications actually die: on knockout questions, on filters an employer chose, and in a queue no person reaches. This post traces the number, then lays out that evidence.
The claim
It comes in a few shapes. "75% of resumes are rejected by ATS." "Three quarters of applications are never seen by a human." "Robots throw out most resumes." The number drifts too: UnchartedCareer, one of two career blogs that have traced it, reports seeing it as 70%, 75% and 88% depending on who is repeating it. The other, JobCannon, traces it to "a 2012 Preptel sales pitch" and says "no methodology, study or sample size was ever published".
The claim usually arrives attached to a product: a resume scanner, a match score, a template sold as "ATS-proof". It is worth asking what the number would have to mean to be true. Someone would need access to many employers' systems, count the applications that were removed by software with no human involved, and divide by all applications. No ATS vendor, researcher or government agency has published that count.
Where it came from
Here is the trail, using only documents we could read ourselves.
| Date | Document | What it says |
|---|---|---|
| 29 Feb 2012 | CIO, on Preptel's launch | Preptel claims "a 90 percent accuracy rating" at finding the keywords an ATS ranks on. No 75% figure. |
| 23 Apr 2012 | RecruitingDaily, on Preptel | Qualitative claims only, such as "Computers have a very specific way that they look at resumes". No 75% figure. |
| 22 Aug 2014 | Fortune column | "That 75% rejection rate your friend cited probably came from a study by a job search services firm called Preptel." |
| Since then | Career blogs and resume tools | The number, usually without a link. |
A few details in that trail matter.
Preptel sold the cure. The CIO piece describes a service that analyzes a resume against a posting and recommends changes so it ranks higher in an ATS. Fortune identifies Preptel's founder, Jon Ciampi, as "an alumnus of ATS maker SumTotal Systems." A firm that sells resume optimization had an obvious reason to describe the problem as large.
Fortune hedged. The column was answering a reader whose friend had cited the figure. It did not link the study, name its sample or describe its method. It said only that the number "probably" came from Preptel. The same column then passes on Preptel's own advice on fonts and on uploading Word files instead of PDFs.
The contemporary coverage does not contain it. We read both 2012 articles about Preptel we could find. The CIO article carries Preptel's numbers for its own product: the 90 percent accuracy claim, a test of roughly 30 resumes by an analyst firm, and an analysis of 13,000 jobs. It contains no figure for how many resumes an ATS rejects. The RecruitingDaily piece contains no statistics at all.
So the most charitable reading is that a vendor said it somewhere we cannot find, without a published method, and a magazine repeated it with a "probably". That is not a statistic. It is a quote with no source document.
What the better evidence says
Three sources are worth more than the 75% figure. None of them gives you a rejection rate either, and each has limits you should know.
Employers admit their filters exclude qualified people
The strongest is Harvard Business School and Accenture's Hidden Workers: Untapped Talent report from September 2021, built on surveys of more than 8,000 hidden workers and more than 2,250 executives in the US, UK and Germany. Its key finding is about belief, not measurement: 88% of employers "believed that qualified high-skills candidates were vetted out of the process because they did not match the exact criteria established by the job description." The report adds: "That number rose to 94% in the case of middle-skills workers."
The report also describes how that happens. Employers set requirements such as "a bachelor's degree, a professional license, or certification" to filter the pool, then rank the remaining applicants on attributes such as "a minimum period of experience or the presence of key words on their resume or in their application". And it names a specific filter: "Almost half the companies surveyed weeded out resumes that present such a 'work gap.' If an applicant's work history has a gap of more than six months, the resume is automatically screened out by their RMS or ATS".
That is a real mechanism with a real source. It is also very different from "75% of resumes are rejected". The employer chose the criteria, and the software applied them.
Recruiters describe knockout questions, not resume scanning
The second source is smaller and comes from a vendor. Enhancv, which sells resume software, interviewed 25 US recruiters and talent acquisition professionals in September and October 2025. 23 of the 25 (92%) said rejections were "manual or triggered only by eligibility filters". 2 said their ATS automatically rejected applicants for other reasons. The write-up reports that 84% of recruiters rely on knockout questions to screen for work authorization, certifications or location, and that 44% said their ATS includes an AI or "fit" score, with only 8% using such scores definitively.
Twenty-five interviews is not a market measurement, and Enhancv says so itself. Read it as a description of practice that happens to match the vendor documentation below.
The vendors document where automatic rejection lives
The third source is the software itself. Greenhouse's application rules are "automatic actions applied to job post questions", triggered "when an applicant's response to a question fulfills the rule's conditions". Auto-Reject is one of three rule types, and you have to create a custom question before you can create a rule. In a 2025 post for candidates, Greenhouse gives the example of answering "no" to "are you legally allowed to work in this country?" and says the rules "are set and controlled directly by recruiters".
Ashby works the same way. Its auto-reject rules "automatically archive candidates based on their answers to your global application questions", send a rejection email the employer picks, and are available only on its Enterprise plan.
Neither describes software that reads your resume and throws it out for its formatting.
What actually rejects candidates
Put the evidence together and the rejections sort into five kinds.
| Mechanism | Automatic? | What triggers it | What you can do |
|---|---|---|---|
| Knockout question | Yes | Your answer to a yes or no question: work authorization, sponsorship, location, a license, minimum years | Read each one slowly and answer truthfully |
| Employer filter | Yes, if configured | Criteria the employer set, such as a degree, a credential or a work gap | Know the hard requirements and skip roles you plainly miss |
| Search and ranking | No, but it decides order | Whether your wording matches what the recruiter searches for | Use the posting's words for work you have done |
| A person reading | No | Fit, as a recruiter judges it | Make your strongest evidence easy to find |
| Volume | No | The queue is long and the role fills | Apply while the posting is live |
The volume row deserves a number, because it is the honest version of a rejection rate. Ashby's 2026 Talent Trends report says "The average recruiter today is processing 291 applications per hire, compared to roughly 100 in early 2021." By our arithmetic, that means roughly 290 of every 291 applicants are not hired. Greenhouse's benchmarks put applications per job at 244 in 2025, up from 116 in 2022, across more than 6,000 companies. Nearly everyone is rejected from nearly every job. That has always been true of hiring, and it says nothing about whether a person or a program made the call.
Parsing, the thing most myth versions blame, is missing from the table on purpose. In the recruiter flow Greenhouse documents, a resume that fails to parse means someone enters the details by hand, and the file stays attached. That is a cost, and worth avoiding with a clean file, which our post on an ATS friendly resume format covers. It is not a rejection.
Why the myth keeps selling
It cannot be checked from the candidate's side. A rejection email does not say who decided. In Huntr's 2025 annual job search report, built on a December 2025 survey of 1,049 job seekers, 64% (675 people) reported "they've been rejected by artificial intelligence before any person reviewed their qualifications". None of them could see inside the employer's system. That is a measure of what people believe, and it is a reasonable belief when nobody tells you anything, but it is not evidence that software made the decision.
It sells products. If most resumes are thrown out by a machine, the fix is a tool that beats the machine. That is the business Preptel was in, and it is the business behind most "ATS score" products today. Our post on resume ATS scores looks at what those scores measure.
It is kinder than the alternative. "A robot rejected me" is easier to hear than "a recruiter read it and chose someone else", or "200 other people applied first". Enhancv's recruiters, asked where the myth comes from, mostly pointed at job seekers repeating it from LinkedIn or TikTok (68%) and at coaches and resume blogs recycling old advice (20%).
Round numbers travel, even in good reports. The Hidden Workers report itself cites Jobscan for "99% of Fortune 500 companies use an ATS". Jobscan's own 2026 count is 97.4%. The same report contains a genuine 75%: "In the U.S., the usage was most prevalent, with 75% of employers using these technologies". That is a share of employers using recruiting software, not a share of resumes rejected, and the two are easy to blur. The 88% that sometimes stands in for the myth is also the Hidden Workers figure, which measures what employers believe about their own filters, not how many resumes software removes.
What to do instead of worrying about it
The useful conclusion is not "the ATS is harmless". Filters an employer configures do exclude qualified people, and an application nobody reaches is in practice the same as a rejection. The useful conclusion is that the risks are specific:
- Answer the screening questions with care. They are the part of the application most likely to carry an automatic rule.
- Use the posting's own words for experience you really have, because recruiters search for terms taken from their own posting. Our guide to what an ATS is walks through search and ranking in each major system.
- Send a file that parses, and check it by copying the text out.
- Skip roles where you miss a hard requirement, such as a license or a required degree. That is where filters do reject people.
This is also why ApplyEze does not sell an "ATS score": no ATS exposes one, so any score you see was computed by the tool showing it to you. What it does instead is write a resume for each posting, and a cover letter when the form asks. ApplyEze brings the posting's named skills and tools onto your skills line, even when your own record does not list them. There is no step to remove one before a live run, so use a dry run to see what it writes, and be ready to discuss anything on the page. A certification the posting asks for is not added unless you hold it. Your employers, job titles, dates, schools, degrees and certifications come from your own record, and code rejects a rewrite that adds a figure, employer, school or credential your record does not hold.