What a screening tool actually reads in your resume

A resume parser does one simple thing: it opens your file, pulls text out of it, then tries to recognise a name, an email, employers, dates and skills. Anything that hinders that extraction costs you before a human ever opens your file.

The five things that break extraction

1. Columns. A two-column resume is read in zigzag by a machine: the Skills line mixes into the Experience line. It is the leading cause of unreadable files.
2. Tables, text boxes and images. Text inside a table or a floating box comes out in the wrong order, or not at all. A heading set as an image never comes out.
3. Headers and footers. Many parsers ignore them. An email placed in the footer can vanish entirely.
4. Fonts without a character map. The nastiest fault, because it is invisible to the eye: the PDF displays perfectly, but the bytes inside are glyph indexes, not letters. The extracted text comes out as gibberish.
5. Invented section headings. "My journey" instead of "Experience": the parser does not recognise the section and files its content anywhere.

The test nobody runs

Everyone shows an "ATS score". Almost nobody shows what a machine actually recovered from your file. Yet that is the only check that proves anything: reopen the PDF with a real extractor and verify, line by line, that the name, the email, every employer, every date and every skill come back out.

You can do it yourself, free, in thirty seconds: open your resume, select all, copy, and paste into a plain text editor. What you see is very close to what the machine sees. If the order is absurd, if words are glued together, if accents vanished, your resume has a problem its layout was hiding.

The format that passes everywhere

One column. Standard section headings. Dates written "March 2021 – February 2024". A PDF whose text is real text, or a Word file. No photo in countries where it is discouraged. And a file name that says who you are: it gets indexed too.