Somewhere on LinkedIn or TikTok, you’ve probably seen the claim: 75% of resumes get auto-rejected by an ATS before a human ever sees them. It’s stated with total confidence, usually right before someone sells you a $49-a-month tool to fix it.

Here’s what almost nobody mentions when they repeat that number: it traces back to a 2012 marketing claim from a small resume-optimization company that had gone out of business by the following year. No methodology was ever published. No study backs it. A career consultant who actually investigated the origin found the number was never measured at all, just asserted.

That single fact tells you something important about how ATS information spreads online: the scariest, most shareable claim wins, regardless of whether it’s true, and once it’s repeated by enough career coaches and TikTok creators, it hardens into “everybody knows this.” This piece is an attempt to actually sort out what’s real.

So what does an ATS actually do, mechanically?

Strip away the mythology and an Applicant Tracking System does four fairly mundane things in sequence, and understanding this sequence matters more than memorizing any single formatting rule, because it explains why some advice is real and some is folklore.

First, extraction. The system pulls the raw text out of whatever file you submitted, a PDF, a Word document, sometimes a plain text paste. This step can genuinely fail: a scanned image with no underlying text layer, a password-locked file, or a PDF generated by flattening a design tool’s output into an image can all leave the ATS with nothing to read. This is the one area where “the format broke things” is completely true, not mythical at all.

Second, segmentation. The parser tries to sort the extracted text into recognizable blocks: contact info, work experience, education, skills. It does this largely by recognizing common header words. A section labeled “Work Experience” gets identified correctly almost every time. A section creatively labeled “My Journey” or “What I’ve Built” can confuse a parser that’s looking for conventional terms, and the content underneath it may end up misfiled or skipped.

Third, parsing. Within each segmented block, the system tries to extract structured fields: job title, company name, start and end dates, bullet content. Inconsistent date formats, text inside tables used purely for visual layout, or content sitting in floating text boxes can all scramble this step, because the system reads in a particular logical order and these elements break that order.

Fourth, ranking, not rejecting. This is the step where most of the mythology lives. The system calculates a relevance score by comparing your parsed, structured data against the job description, then hands recruiters a sorted list. A low score doesn’t delete your resume. It places you lower in the list a recruiter is scrolling through. Whether anyone ever scrolls that far down is a question about recruiter time and applicant volume, not about the software silently executing you.

Where did the 75% myth actually come from, and why won’t it die?

Worth dwelling on this, because the persistence of a debunked statistic is itself revealing about how career advice circulates online. According to research tracing the claim’s origin, it began as an unsupported figure in a 2012 sales pitch from a company called Preptel, which had ceased operating by mid-2013. The number had no published survey behind it, no sample size, no methodology. It simply got stated as fact in marketing copy, and from there, career coaches, resume tools, and social media accounts repeated it because it was a useful, alarming hook, regardless of whether anyone checked where it came from.

A 2025 survey of recruiters found something telling about how this myth actually spreads now: more than two-thirds first encountered the “75% rejection” claim from anxious job seekers on social media, not from any industry source. A smaller share traced it to career coaches recycling outdated advice. In other words, the myth isn’t being manufactured by recruiters or by the software vendors. It’s circulating almost entirely among job seekers themselves, which makes it especially hard to correct, since the people repeating it have no contact with the recruiters who could tell them it isn’t how the system works.

What does the actual data show instead? A study based on structured interviews with 25 U.S. recruiters across industries and company sizes found that only 2 of the 25, eight percent, had configured their ATS to automatically reject applications based on content or match score. The other 92% either reviewed every resume manually or used narrow knockout questions (minimum years of experience, required certification, legal work authorization) rather than broad content-based filtering. Multiple recruiters interviewed across separate research efforts independently echoed the same point: the vast majority of resumes that enter an ATS get at least a human glance.

If it’s not auto-rejection, why do so many applications go unanswered?

This is the part that the “ATS rejects you” narrative gets backwards, and it’s worth sitting with because the real explanation is less dramatic but more useful. The dominant cause, by a wide margin in the data, isn’t a parsing failure or a missing keyword. It’s volume. A typical posting now draws upward of 250 applicants, and a genuinely in-demand role can pull in several hundred to a couple thousand within days. No recruiter, regardless of what software sits in front of them, is reading two thousand resumes closely. They run keyword searches, skim a ranked shortlist, and move forward with what they find there. If your resume never surfaces near the top of that ranked list, you’re not rejected in any formal sense. You’re simply outside the slice of applicants a time-constrained human actually reaches.

One analysis of roughly a thousand rejected resumes attempted to break down the actual causes behind the rejections, and the picture that emerges is far less mysterious than “the algorithm ate it.” The largest share, by a wide margin, came down to a genuine qualifications mismatch, the kind no formatting fix addresses. A meaningfully smaller portion came from formatting and parsing failures: text trapped in a header the system skips, dates in inconsistent formats, content scrambled by a layout that wasn’t built with parsing in mind. A still smaller slice came from missing keywords specifically, and the remainder split between timing (applying days after a role already filled internally) and sheer volume overload.

That breakdown matters because it reorders the priority list most job-seeker advice gets backwards. Formatting genuinely matters, but it’s a smaller piece of the puzzle than the volume problem or the qualifications-fit problem, and no amount of resume polishing solves either of those.

Which formatting myths are real, and which are leftover folklore?

This is where the research gets genuinely interesting, because testing has actually been done here, not just opinion, and some long-repeated advice turns out to be outdated.

PDFs versus Word documents is the clearest example. For years, conventional wisdom insisted PDFs were risky and DOCX was the only safe choice. Direct testing across multiple major platforms in 2026 found the gap between PDF and DOCX parsing accuracy sits within one to three percentage points, essentially noise. The only real PDF failure case is a scanned image with no underlying selectable text, which has nothing to do with PDF as a format and everything to do with how it was created.

Two-column layouts get a similarly outdated reputation. The received wisdom says columns confuse parsers categorically. Testing tells a more specific story: on platforms with clean underlying document structure, properly built two-column layouts actually scored as well as or better than single-column ones. What actually breaks parsing isn’t columns themselves, it’s tables misused to fake a column layout, floating text boxes, and decorative graphic elements layered into the design. A clean, natively-formatted two-column resume and a single-column resume built from a messy table structure are not equally safe, and the column count alone doesn’t tell you which is which.

Keyword stuffing is where the most outdated advice persists. Older ATS platforms genuinely worked on simple term-matching, where literal word overlap was most of what mattered. Modern systems increasingly use the same family of language-understanding models behind today’s AI tools, which means they can recognize that “led cross-functional teams” and “managed collaborative projects across departments” describe a similar competency, even without exact word overlap. Practically, this cuts both ways: keyword stuffing matters less than it used to, because the system understands context now, but it also means an awkward, unnaturally crammed list of skills can score worse than the same skills woven naturally into real accomplishment statements, since the newer matching approach is sensitive to that kind of artificial padding.

White text keyword-hiding remains genuinely bad advice, and not because the ATS is clever enough to penalize it directly. The actual risk is much simpler: when a recruiter exports, copies, or prints the parsed version of your resume, the hidden text becomes visible plain text. What was invisible to you while editing becomes glaringly visible to the human who matters most, and it reads as exactly what it is.

Is there a quieter problem underneath all of this worth naming?

Yes, and it’s worth being direct about it: most of the specific statistics circulating in this space right now, including several used in this piece, originate from a fairly small cluster of sources, and a meaningful number of those sources are resume-building companies with an obvious commercial interest in making ATS seem both intimidating and beatable with the right tool. That doesn’t make their underlying data wrong. Independent recruiter interviews and direct parsing tests are genuinely more rigorous than a decade-old unsupported sales claim. But it’s worth reading this entire category of content, including pieces like this one, with the awareness that a lot of “ATS mythbusting” content is itself produced by companies selling the fix to the myth they’re busting. The most reliable individual data points here tend to be the ones traceable to a named study with a disclosed sample size (the 25-recruiter interview study, a national HR association’s multi-thousand-respondent survey) rather than to a single company’s internal blog claiming a round, convenient percentage.

So what’s actually worth doing with this information?

Format cleanly because parsing failures are real and easily avoidable, not because a robot is hunting for an excuse to discard you. Keep contact information in the document body rather than a header or footer, since that’s a genuine, well-documented blind spot for many parsers. Use conventional section labels. Mirror the language of the job posting where it’s honestly true of your experience, since modern semantic matching rewards natural overlap more than older systems did, while penalizing the obviously artificial kind. And recognize that if you’re not hearing back, the most statistically likely explanation isn’t a hidden formatting trap, it’s that several hundred other people applied to the same posting and a human being, not an algorithm, simply hasn’t gotten to you yet, or already found someone else first.

The ATS isn’t your enemy, and it isn’t quite your friend either. It’s mostly what it sounds like: a filing and ranking system standing between you and an overworked human, doing far less editorial judgment than either the scariest myths or the calmest reassurances usually suggest.

FAQ

Does an ATS really reject 75% of resumes automatically?

No. That figure traces back to an unsupported 2012 marketing claim from a now-defunct company, with no study or methodology behind it. Research based on structured interviews with recruiters has found that the large majority configure no automatic content-based rejection at all, relying instead on human review and narrow eligibility filters like required certifications.

Can an ATS read PDF resumes?

Yes, in nearly all modern systems, as long as the PDF contains actual selectable text rather than being a flattened image. Direct testing has found PDF and Word document parsing accuracy differ by only a few percentage points on current platforms. The only consistent PDF failure case is a scanned or image-based file with no underlying text layer.

Do two-column resumes get rejected by ATS?

Not categorically. Clean, natively formatted two-column layouts have tested as well as or better than single-column layouts on several major platforms. What actually causes parsing problems is using tables to fake a column structure, floating text boxes, or decorative graphic elements, not the presence of columns by itself.

Why didn’t I hear back if the ATS isn’t auto-rejecting me?

The most statistically common reasons have little to do with the software. A genuine mismatch between your qualifications and the role accounts for the largest share of rejections in available research, followed by formatting issues that cause parsing errors, and simple application volume, since popular postings can draw hundreds or thousands of applicants that a human reviewer physically cannot get through individually.

Does keyword stuffing help you pass an ATS?

Less than it used to, and it can actively hurt. Older systems relied heavily on literal keyword matching, but current systems increasingly use language-understanding models that recognize related phrasing even without exact overlap, while also being more likely to flag oddly stuffed, unnatural keyword lists as a negative signal rather than a positive one.

Read about: How Gen Z Is Redefining the Traditional 9-to-5 Job