Does an ATS really reject 75% of resumes?

Updated 2026-09-19 · 3 min read

You have read it a hundred times: robots throw away three out of four CVs before a human sees them. The number is made up. What replaces it is less scary and far more useful.

Where the 75% came from

The claim traces back to an unsourced sales pitch from 2012, repeated ever since by people selling a fix for it. A search of academic literature for an ATS rejection rate turns up nothing supporting the figure. When recruiters were asked where the number came from, 68% pointed at social media posts and another 20% at career coaches and CV-writing services — in other words, at people whose product is your fear of it.

What an ATS really does

An applicant tracking system is a database with a workflow on top. It collects applications, parses them into fields, lets recruiters search and sort, and tracks who is in which stage. Workday, Greenhouse, Lever and Ashby are the ones you meet most often.

In a 2026 study of 25 US recruiters across technology, healthcare, finance, publishing and retail, 92% said their system does not automatically reject anyone on formatting, design, missing keywords or a low AI match score. The 8% who did use auto-rejection applied it only to roles with highly specific hard requirements, such as a licence you either hold or don't.

So what feels like a robot rejection is usually a person: a recruiter with 300 applications, ten minutes, and a search box. That distinction matters, because it changes what you should optimise for. You are not trying to trick a parser. You are trying to be findable and obviously relevant to a tired human who is skimming.

Six rules that still matter

Parsing failures are real, even though blanket rejection is not. A CV that parses badly lands in the database with your job titles in the wrong fields, and then you are invisible in the search that the recruiter does run.

1. One column, no text boxes. Two-column layouts get read across the middle, which mixes your employer with your dates. This is the single most common parsing failure.
2. Real text, not an image. A beautifully designed CV exported as a picture contains no searchable words at all.
3. Standard headings. "Work experience", "Education", "Skills". Not "My journey".
4. Their words for your work. If the vacancy says "customer success" and your CV says "client care", a search for the first will not find you. Mirror their terms where they are honestly true of you.
5. Dates in a consistent format, month and year, on every role. Gaps look worse when the parser cannot line them up.
6. Send the format they ask for. PDF unless the posting says otherwise. Name the file with your own name and the role.

Three things not to waste time on

Keyword stuffing. A block of terms at the bottom of the page does not raise a score that mostly is not being used; it reads as desperate to the human who does open the file.

White text tricks. Hiding keywords in white on white is visible in every parsed view, and it reads as an attempt to cheat. It gets you removed, not ranked.

"ATS-score" tools that promise 98%. They score you against a model of a system that, as the study above shows, mostly is not scoring you at all. Spend that hour tailoring the top third of your CV instead.

In the Netherlands

Dutch employers often run AFAS, Recruitee, Homerun or Carerix rather than the American systems, and agencies run their own. The rules above do not change: the failure mode is a parse that scrambles your roles, and the decision is made by a person. One difference worth knowing: Dutch vacancies list requirements as a wish list far more often than tenders do, so a 60% match is usually worth applying for, while a public tender with knock-out criteria is checked literally.

← All knowledge articles