How ATS actually reads your CV — the parts you can control

There's a lot of ATS folklore online. Most of it is downstream of two facts: (1) every ATS is a database with a keyword search, not a robot with judgment; (2) the parser that fills the database rejects the same three PDF anti-patterns everywhere. Everything else is detail.

The three anti-patterns that always fail

1. Multi-column layouts. Two-column resumes look elegant to humans but scramble in every ATS tested — rows get read across columns, so your "React" skill ends up next to your MBA start date. The parser then indexes garbage.
2. Text inside images. Icons carrying meaning, skills lists rendered as graphics, contact info embedded in a header image — invisible to the parser. If your name is only in the image, the ATS files you under "unknown."
3. Tables for critical data. Tables scramble like columns — the ATS reads by row then column, but the visual grouping was meant column-first. Skills, dates, and job titles inside tables end up mismatched.

Two subtle killers most articles miss

1. Contact info in the header/footer

Every ATS strips header and footer elements before parsing (they're layout scaffolding, not content). If your email and phone live in those regions, the recruiter's system gets your resume with no way to reach you. Keep contact info in a plain-text block at the very top of page 1.

2. Custom fonts that don't embed

If you export a PDF using a font that Word or your PDF library didn't embed, the ATS renders "missing glyph" boxes for every character. Your name displays as ▯▯▯▯. Stick to Inter, Calibri, Arial, or Helvetica — the four fonts that ship with every PDF library on earth.

After the layout works, what actually ranks you

Once your CV parses cleanly, ranking inside the ATS depends on keyword match against the recruiter's Boolean query. The keyword extraction is straightforward: read the JD, pull the 8-15 nouns and named tools, verify each appears in both your Skills section AND at least one Experience bullet. That's the entire optimization loop.