How AI actually reads your résumé.

Here's an uncomfortable truth about the modern job search: for most roles, the first thing to read your résumé isn't a person. It's software — an applicant tracking system, increasingly powered by AI — that parses, scores, and ranks you before a human spends a single second on your application.

That sounds dystopian. It isn't, really. But it does mean a great candidate can get filtered out for reasons that have nothing to do with whether they'd be great at the job. The good news: once you understand what the machine is doing, you can stop losing to it.

What the machine is actually doing

An AI résumé screen does roughly four things: it extracts your information into structured fields (name, roles, dates, skills), interprets what those fields mean, compares them against the requirements of the job, and ranks you against everyone else who applied.

Modern systems don't just match exact keywords anymore — they read for meaning. They can tell that "managed a team of six" implies leadership, or that "shipped a payments integration" implies certain technical skills. But they're only as good as the signal you give them. Bury the signal, and you rank low.

Where good candidates get filtered out

• Heavy formatting — multi-column layouts, tables, text boxes, and graphics that parsers scramble or skip entirely.
• Résumés saved as images (or scanned PDFs) where the text isn't actually selectable.
• Skills that are only implied by your experience but never stated plainly.
• Job titles that are creative or company-specific and don't map to the role you want.
• Career gaps or pivots with zero context, which the system can't interpret charitably.

How to write for the filter (without writing for a robot)

• Use a clean, single-column layout with real, selectable text. Boring parses best.
• Mirror the language of the job description — but only for skills you genuinely have.
• State your skills explicitly, then prove them in bullets with concrete outcomes.
• Use standard section headers (Experience, Skills, Education) so fields extract cleanly.
• Keep titles recognizable, and add a plain-English version in parentheses if yours is unusual.

Write for the human who reads it second. Make it legible to the machine that reads it first.

You can't beat the filter by gaming it. You beat it by being legible: clear about what you've done, clear about what you want, and easy for both software and humans to understand. Do that, and the machine becomes a door instead of a wall.