What an ATS actually does with your resume, and what it does not

Before a recruiter ever opens your resume, software usually reads it first. An applicant tracking system, or ATS, pulls the text out of your file and sorts it into fields like contact details, work history, education, and skills. This step is called parsing, and it happens whether your resume looks perfect on screen or not.

Parsing is where formatting can cause problems. If the software cannot correctly identify where one section ends and another begins, it can misfile your experience, drop your contact details, or jumble your work history out of order, before anyone has read a word of what you actually did.

Many employers then use the system to filter or rank applicants. In a 2021 Harvard Business School and Accenture survey, more than 90% of employers who used this kind of software said they used it to initially filter or rank candidates. The filters are rules the employer sets, such as a required degree, specific skills, or no gap in employment longer than six months. There is no universal passing score.

Most parsers read a document the way we read a book, left to right, top to bottom, in one continuous stream. A two-column layout or a table breaks that assumption. The software can read straight across the page instead of down one column at a time, which mixes unrelated lines together.

Columns don't help with people either. In Ladders' 2018 eye-tracking study, cluttered layouts with long sentences, multiple columns and little white space performed poorly with recruiters.

The fix: use a single-column layout for your entire resume. If you want a clean visual layout, use simple line breaks and bullet points instead of a table.

Many ATS platforms focus on the main body of a document. Anything placed in a header, footer, or floating text box, like your name and phone number inside a designed header, may not be picked up reliably by the parser.

The fix: put your name, phone number, email, and location in the main body of the document, not in a header or footer.

The software looks for familiar words to classify your content, things like Experience, Education, and Skills. A creative heading like My Journey or Where I Have Been might read beautifully to a person, but it gives the parser nothing to match, so that section can be skipped or filed incorrectly.

The fix: stick to plain, expected headings. Save the personality for your summary and your interview.

A logo, a headshot, or a set of star ratings for your skills might look sharp in a PDF, but none of it is text. Parsers cannot read pictures, so any information locked inside an image is simply gone as far as the system is concerned.

The fix: describe your skill level in words, for example Advanced or 5+ years, instead of graphics, and keep photos and icons out of the document entirely for roles based in the US and most other markets.

The PDF versus Word debate has cooled off. Most modern systems handle both reasonably well, as long as the file has real, selectable text. What still causes real damage is a scanned or image-based PDF, made by photographing or scanning a printed resume. There is no text layer at all for the software to extract, so the whole document can come through blank.

The fix: if a job posting does not specify a format, a text-based .docx is the safest universal choice. If you use PDF, export it directly from Word or Google Docs, never from a scanned image.

Different ATS platforms have different known weak points. This is not a reason to guess which system a company uses, it is a reason to avoid the riskiest formatting choices across the board.

Parsing behavior changes as vendors update their software, so treat this as a general risk map, not a guarantee for any single application.

Myth: you need a perfect ATS match score to get noticed. There is no universal passing score. Employers set their own filters and rank you against other applicants for that specific role, so a well-tailored resume with an honest keyword match is worth more than an artificially inflated one.

Myth: 75% of resumes are automatically rejected by ATS before a human ever sees them. This is one of the most repeated numbers in job search advice, and it does not hold up. HR consultant Christine Assaf traced it back to Preptel, a job search services firm that went out of business in August 2013, and found no methodology that would let anyone verify it. Automatic filtering is real, but it comes from rules employers configure, such as knockout questions, degree requirements or employment gaps, not from a machine judging your formatting.

Myth: stuffing keywords in white text tricks the system. Hidden text is still text. It shows up in the parsed version a recruiter sees, and it makes an application look dishonest rather than qualified.

Myth: a paid ATS-friendly template guarantees you will pass. Parsability depends mostly on the underlying file structure, not the price tag. A plain, well-organized document you build yourself with a single column and standard headings is a perfectly good choice.

The origin of the most repeated of these claims is traced in full in where the 75% number came from. If you want to know what an employer has to tell you when software screens your application, that is in AI in hiring: what you are entitled to know.

We did not run our own parsing test, and this guide does not pretend to. It reads published research and links every figure to where it came from.

Two of those figures are weaker than they look, and it would be dishonest to quote them without saying so. The 7.4 second scan comes from a Ladders vendor study that is not peer reviewed, and the 2018 announcement does not give a sample size (the original 2012 version tracked 30 recruiters). The 88% is employers giving their own view of their hiring process, not a measured rejection rate. We have kept both because they are the best available, and labelled both because they are not conclusive.

One figure we removed. An earlier version of this page claimed that 25% of ATS platforms miss contact details placed in a header or footer. We could not find an independent study behind it, so it is gone. The underlying advice stands on its own: parsers read the document body, and a header is not reliably part of it.

Our own resume score is an AI estimate of how well a resume is likely to parse and match. It is not a test against any employer's ATS. Nobody outside Workday, Greenhouse or Taleo can run one, and anyone who says they do is describing something they cannot do.

The filter that genuinely ends an application on its own is not parsing at all: see knockout questions.