ATS-friendly resume format: the rules that actually matter

Why formatting affects parsing at all

A resume file does not contain a resume. It contains instructions for drawing one. This is most obvious in PDF, where the format stores glyphs at coordinates on a page rather than a sequence of sentences — the reading order a human perceives is inferred from position, not recorded in the file.

Extraction software has to rebuild the reading order from those positions. Where the layout is a single column top to bottom, that reconstruction is trivial and near-perfect. Where it is not — two columns, a sidebar, a table with invisible borders — the software has to guess whether to read across or down, and every wrong guess interleaves two unrelated pieces of text into one nonsensical line.

That is the entire mechanism behind "ATS-friendly formatting". Not a list of arbitrary preferences: a set of choices that make the reconstruction unambiguous.

The rules that actually matter

PDF or DOCX: which to send

Send what the application asks for. When it does not specify, either works — with one qualification each.

PDF

Preserves your layout exactly, which is why most people prefer it, and modern parsers handle text-based PDF well. The failure mode is a PDF that contains no text: exported from a design tool as a flattened image, or produced by scanning a printout. If you cannot select and copy text out of your own PDF, no parser can read it either.

DOCX

Generally the more predictable of the two, because the format records structure explicitly — this paragraph is a heading, this block is a table — rather than leaving it to be inferred from position. The trade-off is that layout can shift between Word versions.

What to avoid regardless: .pages, .odt, RTF and image formats. Support for these is inconsistent and there is no upside to finding out which side your employer is on.

Section headings that parse reliably

Parsers look for conventional headings to segment a document. Use the boring ones — this is not where personality earns anything.

• Work Experience— or "Experience", or "Professional Experience". Not "Where I've Been".
• Education— not "Learning" or "Academic Journey".
• Skills— not "Toolbox" or "What I'm Good At".
• Projects, Certifications, Publications — all standard and all recognised.

Within each role, keep the same order every time: title, employer, dates, then bullets. A consistent pattern is easier to segment than a varied one, and it reads faster for the human too.

Rules that are repeated but obsolete

Several widely repeated prohibitions describe parser behaviour from a decade or more ago. Following them costs you a worse-looking resume for no gain.

• "Never send a PDF." Text-based PDF parses reliably in current systems. The real rule is narrower: never send a PDF whose text cannot be selected.
• "Never use bold or italics." Character formatting is discarded on extraction and does no harm. Use it for the human.
• "Use one specific font or be rejected." Any common, embedded font extracts identically. The typeface is a design decision, not a parsing one.
• "Keep it to one page or it will not be read." A length convention for humans, with no bearing on parsing. Two pages is normal past a few years of experience.
• "Match keywords to exactly 60% density." No published scoring system works this way, and the edits it produces make a resume worse to read.

How to check your own file in five minutes

You do not need a tool for the first and most valuable test.

• The copy-paste test. Open your resume, select all, copy, and paste into a plain text editor. What you see is approximately what a parser sees. If columns interleave, sections vanish or your name is missing, you have found your problem.
• The heading test. In the pasted text, can you tell where experience ends and education begins, without the visual styling? If not, neither can a parser.
• The date test. Are all dates in one format, and does each role clearly have one?
• Then run a parse. A resume checker will show the structured fields it recovered, which catches boundary errors the copy-paste test cannot.

Once the file parses cleanly, formatting stops being the constraint and content takes over — which is what tailoring to the posting is for. tailorurCV stores your details once and renders them through several templates, so switching layout costs a click rather than a retype — worth knowing that the single-column templates are the safer choice when a posting matters and the two-column and sidebar ones trade a little parsing certainty for density.