If you've read more than one article about job hunting, you've run into some version of this claim: "75% of resumes are rejected by ATS software before a human ever sees them." It's repeated constantly, cited by career centers, resume tools, and major outlets alike.
It appears to be fabricated. Researchers tracing the claim have found it originates from a single 2013 sales pitch by a since-defunct startup, with no underlying study, dataset, or methodology ever published. It then spread through a citation chain — one outlet cited it, the next cited that outlet, and so on — until it looked authoritative simply from repetition. A search of academic literature turns up no research supporting the number.
That doesn't mean ATS software is irrelevant to whether you get an interview. It means the actual mechanism is different from "a robot silently deletes your application," and it's worth understanding the real version, because it changes what's actually worth optimizing for.
What an ATS is actually doing
The clearest way to think about an applicant tracking system: it's a search engine for recruiters, not a rejection machine. A recruiter opens a role, sets some criteria, and the system's job is to make several hundred applicants searchable and sortable instead of an unstructured pile of PDFs.
To do that, it has to first turn your resume into structured data. That process generally looks like:
Extraction. Text gets pulled out of whatever file you submitted — for scanned or image-based PDFs, this uses optical character recognition (OCR); for text-based files, it's more straightforward.
Parsing. The system uses natural language processing to identify what kind of information each piece of text is — this job title, that company name, this date range, that skill — a task often called named entity recognition. This is the step that formatting can quietly break: text inside tables, multi-column layouts, headers/footers, or embedded graphics often doesn't extract cleanly, so information that's obviously a job title to a human reader can come out garbled or misattributed to the system.
Indexing. Rather than re-scanning every resume every time a recruiter searches, most systems build something like a search index — a map of which words appear in which candidate's profile, similar to the index at the back of a book. That's what makes instant keyword search across thousands of resumes possible.
Where the real gatekeeping happens
The actual filtering that resembles "rejection" mostly comes from what recruiting systems call knockout questions — a small number of yes/no or qualifying questions set explicitly by the employer: work authorization status, willingness to relocate, a hard minimum years-of-experience cutoff. Those are configured by a person, for a specific role, and they're usually visible to you as questions in the application form itself, not hidden resume-scanning logic.
Recent industry data backs up how rare true automatic content-based rejection actually is: one 2025 study found that roughly 92% of ATS deployments don't configure any rule that auto-rejects a candidate based on resume content at all. The system ranks and surfaces; a person still makes the call on who moves forward.
So what's actually going wrong when a resume "doesn't work"
If outright rejection-by-robot is mostly a myth, the more accurate failure mode is quieter: a resume that's technically fine but ranks low in the search, so it never surfaces near the top of what a recruiter looks at. Data from resume-scoring tools bears this out — one analysis of first-time resume submissions found a median score of just 48 out of 100 against a specific job posting, with the average resume missing roughly half of the keywords that posting was actually searching for.
That's not a formatting failure or a bot glitch. It's a mismatch between the language on the resume and the language the search is running on — which is exactly the gap tailoring closes, and exactly why the next post in this series is about keywords specifically.
The practical takeaway
Clean, standard formatting (avoid dense tables, multi-column layouts, text boxes, and headers/footers for anything important) still matters, because it's what lets the parsing step do its job correctly. But formatting paranoia is the smaller issue. The bigger lever is making sure your resume's language actually overlaps with what a specific posting is searching for — which is a matching problem, not a trick-the-robot problem. Callback is built around that distinction: it reads the actual job description and rewrites your resume to match its language, rather than trying to game a system that mostly isn't rejecting anyone in the first place.
Sources: ResumeAdapter — ATS Statistics 2026: The "75% Rejection" Stat Is Fake, Jobscan — 8 Things You Need to Know About Applicant Tracking Systems