Skip to content

Straight answers

What an ATS actually does

Most advice about applicant tracking systems is repeated, not checked. Some of the most confident claims trace back to a sales pitch from 2012. Here is what the evidence supports, what it does not, and the short list of things that genuinely change whether software can read your resume.

Every figure on this page is attributed

Where this advice comes from

ATS folklore spreads because it is useful to repeat. It explains rejection without requiring anyone to look at the resume, and it sells things. When HR.com asked recruiters where the auto-rejection claim comes from, 68% pointed at job seekers repeating it to each other and 20% pointed at coaches and resume companies recycling old advice. We build resume software, so that second number is about people like us.

Six things you have probably been told

The claim An ATS automatically rejects 75% of resumes before a human sees them.

What the evidence says

This is the single most repeated claim in the category, and it is fabricated. It traces to a 2012 sales pitch from a company that folded the following year, and it has been passed along ever since without anyone re-checking it. Asked directly, 92% of recruiters said applicant tracking systems do not automatically reject resumes. Most of these systems are databases with a search box: a recruiter runs a query and reads what comes back. Nothing is thrown away on its own.

Source: HR.com recruiter survey, 2026

The claim PDFs get rejected. Always send a Word file.

What the evidence says

Parsing tests put PDF and Word at roughly the same extraction accuracy. What actually breaks is a PDF with no text layer — an export that is really a picture of a resume — because there is nothing in the file for a parser to read. That is a property of how the PDF was made, not of the format. If you can select the text in your own PDF, so can the software.

Source: Enhancv parsing tests, 2026

The claim A two-column layout will break the parser.

What the evidence says

This one is real, and much smaller than advertised. Testing puts two-column layouts at around 86% extraction accuracy against roughly 93% for a single column. That is a modest difference, not a wall. A two-column resume is a legitimate choice, particularly if it is the one a human will actually enjoy reading — just know you are trading a few points of parsing reliability for it.

Source: Enhancv parsing tests, 2026

The claim Stuff in keywords, or hide them in white text, to beat the filter.

What the evidence says

There is no filter to beat. The search results a recruiter looks at are read by that recruiter, so a paragraph of comma-separated keywords is not a clever exploit — it is the first thing they see. White-text keyword hiding is a known trick, it is trivially visible the moment anyone selects the text or opens the file in a parser, and it reads as dishonesty rather than resourcefulness.

Source: Applicant tracking system, Wikipedia

The claim The system scores you and ranks you against everyone else.

What the evidence says

Some platforms do offer matching and ranking features, and adoption of automation is rising. But many applicant tracking systems in daily use are still straightforward trackers, and a score inside one is an aid to a recruiter rather than a gate. The decision is still someone reading a screen and choosing who to call.

Source: HR.com recruiter survey, 2026

The claim If you are not getting callbacks, your resume must be failing the ATS.

What the evidence says

This is the one worth sitting with, because the real change has nothing to do with parsing. Applications per role have roughly tripled since 2021 to 300 or more, LinkedIn reported a 45% year-on-year jump in applications, and AI-written applications have converged enough that employers are moving back toward live tests and referrals to tell candidates apart. If you sent 200 applications and heard nothing, the most likely explanation is that you were one of several hundred people for each of them, not that a parser mangled your dates.

Source: Bloomberg and LinkedIn application data, 2026

What actually matters

A short list, and it is genuinely short. Everything here is either mechanical or free.

Export a PDF with a real text layer. Open it and try to select your own name. If nothing highlights, no parser can read it either.
Use plain section headings. "Work Experience", "Education", "Skills". A parser maps those to its own fields; an invented heading may not be mapped at all.
Match the job title where it is honest to do so. It is one of the few fields mapped directly, so it carries more weight than any single keyword.
Keep one date format throughout. Employment dates are among the few fields every parser looks for.
Put your email and phone in the body of the document, not in a header or footer image.
Use the words the posting uses, where they are true of you. That is not keyword stuffing; it is answering the question that was asked.

Our own correction

UiResume used the 75% claim in its own guide until we checked it and could not find a source that survived contact. It is gone from the site, and the marketing copy no longer carries any figure we cannot attribute. We are naming it here because a resume company quietly deleting a statistic is not the same as a resume company telling you it was wrong.

Check yours instead of guessing

Our checker reads your PDF in your browser — the file is never uploaded — and tells you whether it has a real text layer, whether the layout is single or two column, and which sections a parser will find. That part needs no job description and no account.

Run the free check See ATS-friendly templates

Common questions

Do applicant tracking systems really reject resumes automatically?

Overwhelmingly, no. In a 2026 HR.com survey, 92% of recruiters said their applicant tracking system does not automatically reject resumes. Most systems store and search applications; a person decides who to contact. The widely repeated claim that 75% are auto-rejected traces to a 2012 sales pitch by a company that folded in 2013.

Should I send a PDF or a Word document?

Either. Parsing tests put PDF and Word at roughly the same accuracy. What matters is that the file contains real text rather than an image of a resume. If you can select the text in your own PDF, a parser can read it.

Is a two-column resume a bad idea?

It is a small trade-off, not a mistake. Testing puts two-column layouts at around 86% extraction accuracy against roughly 93% for single-column. If the two-column version is clearer for a human reader, that is a reasonable thing to trade a few points for.

Does hiding keywords in white text work?

No, and it is a bad idea. The text is visible to anyone who selects it or opens the file in a parser, and the recruiter reading the result sees the keyword block for what it is. It reads as an attempt to deceive rather than as a clever workaround.

Why am I not getting callbacks then?

Usually volume. Applications per role have roughly tripled since 2021 to 300 or more, and AI-assisted applying has made submissions look increasingly alike, which is why some employers are moving back toward tests and referrals. That is a competition problem rather than a formatting one, and the usual answer is fewer, better-targeted applications and a direct approach to the hiring manager.