How ATS Resume Parsing Works, and Where It Breaks

An applicant tracking system processes a CV in four mechanical stages: extracting the text, splitting it into sections by headings, pulling out entities like email addresses, dates and job titles, and indexing the result for recruiter searches. Many common CV formatting rules stem from specific stages failing, such as PDFs with multi-column layouts, where text fragments can be merged into nonsensical lines. Parsers identify sections by matching short headings against known word lists, so conventional labels like Experience or Skills are recognised far more reliably than creative alternatives. Within sections, dates act as anchors that signal the start of a new role, which is why undated entries often get misread. The article's author illustrates these failures with a line-by-line checker they built.