Every US and Canadian job seeker has heard some version of the same warning: "the ATS will reject you before a human ever sees your resume." Most of what circulates about how to beat it is folklore — stuff someone read on a forum in 2019 and never re-checked. Here's what an applicant tracking system actually does with a resume, based on what breaks and what doesn't when a resume gets machine-parsed at Workday, Greenhouse, iCIMS, and similar platforms.
An ATS doesn't reject you. It mis-reads you.
The phrase "ATS rejection" is misleading. In the large majority of setups, the applicant tracking system's job is narrower than people assume: pull structured fields — name, contact info, employer, title, dates, skills — out of an unstructured document, and hand a recruiter a searchable, filterable record. It doesn't usually run a pass/fail scoring model on you personally. What actually hurts candidates is a parser that reads a two-column layout left-to-right instead of top-down, stitches your most recent job title onto the wrong company, or drops a skills section entirely because it lived inside a text box the parser can't see into.
The practical effect is the same as a rejection — a recruiter searching for "Python" or "project management" never finds you, because the parser never extracted those words into a searchable field — but the mechanism is different, and so is the fix. You're not being judged and found wanting. You're being misread.
What actually breaks parsing
- Multi-column layouts. Parsers generally read left-to-right, line-by-line. A two-column resume gets its columns interleaved — a bullet from your "Skills" sidebar can end up mid-sentence inside your most recent job description.
- Text inside images, text boxes, or headers/footers. Anything that isn't part of the document's actual text flow — a name banner built as a text box, a skills list in the page header — is frequently invisible to the parser. It looks fine to a human, and reads as blank to the machine.
- Tables used for layout. A table used to align a job title on the left with dates on the right often parses as a jumbled row of cell contents rather than the sentence a human reads.
- Nonstandard section headings. "Where I've Made Impact" is a fine heading for a human. A parser looking for a work-history section is trained on "Experience," "Work History," or "Employment" — an unusual heading can mean the whole section gets miscategorized.
- Dates in inconsistent or unusual formats. "Since early 2023" doesn't parse into a start date the way "03/2023" does, and a missing end date can make a current role look finished.
What keyword-stuffing actually does
The advice to cram a job description's exact keywords into a resume comes from a real observation — recruiters do search parsed resumes by keyword — but the common execution of it backfires. A wall of skills with no context reads as noise to the same recruiter who's searching, and on some platforms a resume that scores as keyword-dense but content-thin gets flagged rather than surfaced. The version of this that actually works is using the same terms the job posting uses, in context, inside real bullet points: not just "Python" but "built a Python ETL pipeline that cut nightly batch runtime by 40%." That parses as a skill and reads as an accomplishment, which is the whole point.
The format that survives both a parser and a recruiter
A single column, standard section headings, dates as MM/YYYY, and real text (not text-inside-an-image) will parse cleanly on essentially every ATS in wide use across US and Canadian hiring — Workday, Greenhouse, Lever, iCIMS, and the smaller platforms built on similar conventions. We wrote a longer, more detailed breakdown of exactly this in A Resume Format That Autofill Tools and Recruiters Can Both Parse — it's the same underlying parsing problem, just from the formatting side rather than the "what breaks" side.
Where autofill fits into this
This matters beyond how a recruiter finds you — it's also exactly what determines how cleanly a browser-based autofill tool can read your resume back out. ApplyCandid's deterministic fields (name, email, phone, work authorization, and similar factual data) get matched from your uploaded resume and profile, entirely on-device, with no AI call involved. A resume that parses cleanly for a human ATS parses just as cleanly for that step — the same structural habits that keep a recruiter's search from missing you are what let an autofill run pull the right answer into the right field on the first pass, instead of leaving a blank for you to fix by hand.
Stop guessing which resume version parses cleanly.
Upload once. ApplyCandid fills the deterministic fields on any ATS from the same parsed profile.