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 A smaller share traced it to career coaches recycling outdated advice.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

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

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

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’