How each ATS actually reads a resume.

What each system takes

Every value here is sourced on the system's own page. A dash means the vendor does not publish that number — not that there is no limit.

The number you keep seeing was made up

You have probably read that 75% of resumes are rejected by an applicant tracking system before a human sees them. It appears in news articles, university careers pages and the marketing of nearly every resume tool — including tools that charge you to solve it.

It traces back to a 2012 sales pitch by a company called Preptel, which sold resume optimisation services and went out of business the following year. No study, sample or methodology was ever published. The HR consultant who went looking for the evidence found, in her words, no conclusive statistical evidence of the premise — merely conjecture.

What these systems genuinely do is parse, store, rank and filter. Applications do get screened out automatically, but by knockout questions and search filters that a hiring team configured — work authorisation, minimum years, employment gaps — not by software forming an opinion about your resume. A resume that parses badly still reaches a person; it just arrives as a worse record of you, harder to search and missing the fields recruiters filter on.

That distinction is the reason this site shows you extracted text instead of a score. Formatting is worth fixing because it decides what the record says — not because a robot is about to reject you.

Workday, Lever and iCIMS are missing on purpose

Those three are among the most searched names in this category, so leaving them out costs us traffic. We looked. Workday and iCIMS keep their documentation behind a customer login, and Lever's help centre renders only through JavaScript, so nothing about what any of them accepts could be traced to a primary source.

There are confident file-type lists and size limits for all three circulating online. Every one we followed led back to a resume tool trying to sell something, never to the vendor. Publishing those numbers would have been easy, and it would have made these pages exactly the kind of thing they were written to replace. When we can source it, we will add it.