An applicant tracking system reads your resume as text before anyone sees it as a page. The parser pulls out your name, your contact details, each role with its dates, your skills and your education, and files them in fields. If it cannot find a field, the field is blank, and a blank field is how a good resume fails a screen. The rules for a readable resume are short, they come from the vendors' own documentation, and most of the folklore about them is wrong.
Key points
- One column. A parser reads across the page, and two columns become one interleaved mess.
- Plain headings: Experience, Education, Skills. The parser looks for those words.
- Each role as title, company, dates, on one or two lines, before its bullets.
- No text in images, headers, footers or text boxes. The parser does not look there.
- A PDF from a word processor is fine. A PDF from a design tool is often not.
What the parser does
The parser turns the file into text, then looks for patterns: a block of contact details at the top, a section called Experience with role headings and date ranges, a section called Education, a list of skills. Each thing it finds becomes a field in your candidate record. The recruiter then searches and filters on those fields.
Two things follow. The parser has to be able to read the text at all, which is about the file. And it has to be able to tell which text is which, which is about the layout.
A resume the parser cannot read is a resume nobody read.
The nine rules
| Rule | Why |
|---|---|
| One column | Multi-column pages are read left to right across both columns. |
| Headings in plain words | The parser looks for Experience, Education, Skills, Certifications. "Where I have been" is not on its list. |
| Role heading: title, company, dates | Each on the same line or the next, before the bullets, every time. |
| Dates as month and year or year | "Mar 2021 to Jun 2023". Not "Spring 2021". |
| Standard bullets | A dash or a round bullet. Not a custom glyph that becomes a question mark. |
| No tables for layout | Cells are read in an order the parser chooses. |
| No text in headers, footers, images or text boxes | Many parsers skip them. Your name in a header is a name the parser never saw. |
| A standard font | Any common one. The parser reads the text, not the shape. |
| Export from a word processor | Word, Google Docs, or the builder on this site. A design tool's PDF often stores text as outlines. |
The first five are about telling the fields apart. The rest are about reading the text at all.
The five myths
"ATS rejects resumes automatically." It files them. A recruiter rejects them. A parser that found no dates gives the recruiter a candidate with no experience, and the recruiter rejects that. The machine is not the reader. It is the clerk.
"You need a 75 percent keyword match." No hiring system we have read scores a resume as a percentage. Recruiters search for terms, and a resume with the term is found. The number comes from resume-scanning products, which sell the number.
"PDFs do not parse." They do, when the PDF has a text layer, which every word processor writes. The ones that fail are PDFs exported from design tools with text as curves, and scanned PDFs, which are pictures.
"Keep it to one page or the ATS truncates it." Nothing truncates. Length is about the reader. The guide on length has the rule.
"Put keywords in white text." Parsers read it and the recruiter sees it in the text view. It is the one thing that gets an application binned on sight.
The file
A PDF exported from a word processor or from the builder on this site has a text layer and parses cleanly. A Word file does too. Some older systems ask for Word. Most accept either.
Name the file with your name and the role. The name is for the recruiter's download folder, not the parser.
Keep it under two megabytes. A resume is text. A resume that is twelve megabytes has a photo in it at print resolution, and the guide on how to write a resume says where the photo goes, which is nowhere.
The check on this site
The score on the resume builder runs the parse: it reads your resume the way a hiring system does and shows what it found. The "clean, standard structure" bubble is the layout rules above. The "easily contactable" bubble is whether the contact block parsed. A field the parser missed is marked, with the line that caused it.
The builder's export is one column, plain headings, standard bullets, and a text layer, so a resume built there passes the nine rules by construction. The score is for the ones you bring in.
Questions
Which hiring systems are strict?
The parse rules above cover all the ones we read from, which is most of them: Greenhouse, Lever, Ashby, Workday, Workable, SmartRecruiters, iCIMS, Taleo. Workday and Taleo also ask you to retype the resume into a form, and the guides for each one cover that.
Do colours or icons break the parser?
Colours no. Icons for phone and email yes, when the icon replaces the label. Write "Email" or write nothing, but write the address as text.
My designer resume has to look good. What then?
Two files. A plain one for the form, and the designed one linked from it as a portfolio. The parser reads the first. The designer who interviews you opens the second.
Does a two-page resume parse worse than one?
No. Page breaks are invisible to a parser. A role split across the break can confuse a weak one, so end page one at the end of a role.
Can I test my resume against a real parser?
The score on this site is one. Most hiring systems also show you what they parsed in the application form, as pre-filled fields. If the fields are wrong, the resume is wrong, and you can fix it before you submit.