How applicant tracking systems actually read your resume

An applicant tracking system does not read your resume and decide you are unqualified. It is a database with a workflow attached. It parses your file into fields, stores the text, indexes it, and presents recruiters with a searchable, sortable, filterable list of applicants. The decisions are made by people using that list.

This matters because it changes what you should optimise for. You are not trying to fool a gatekeeper. You are trying to be correctly represented in a database, and then to rank well when a recruiter searches it. Those are two separate problems and almost everything useful you can do addresses one or the other.

When you upload a file, the system tries to work out which text is your name, which is your email, which lines are job titles, which are companies, and which dates belong to which role. It is pattern matching against an expected document shape, and it is where most silent damage happens.

It goes wrong more often than the confidence of the upload screen suggests. Some leading platforms fail to correctly parse 20% or more of the resumes they are given, and the rate climbs with complex layouts, columns, tables and graphics. One document in five is misread before anybody has an opinion about the person who sent it.

Things that reliably confuse a parse: text inside text boxes or shapes, tables used to align columns, a two-column layout where reading order is ambiguous, headings rendered as images, icons in place of the words "email" or "phone", dates written as "Spring '19" or "3 yrs", and creative section names like "Where I've Been" instead of "Experience".

The result of a bad parse is not a rejection email. It is a record where your most recent role is blank, your dates are in the wrong century, or your skills section is one long unsearchable string. Nobody tells you this happened, which is why the single most useful diagnostic available is to import your own resume into any parser and look at what comes out.

Once the text is indexed, recruiters search it. Sometimes that is a literal keyword search ("Kubernetes" and "Terraform" and a location), and sometimes the system scores every applicant against the requisition automatically and sorts by that score. Both behaviours reward the same thing: your document containing the language of the posting, in the places the system weights most heavily.

The headline under your name is weighted disproportionately in most systems, which is why a headline reading "Senior Product Designer" when the posting says "Senior Product Designer" is one of the cheapest wins available. "Creative problem solver" in the same position is close to worthless.

Skills sections are indexed but weakly weighted, because everybody stuffs them. The strongest placement for a term is inside the bullet describing work you did with it: it is indexed, it is corroborated by context, and it survives a human reading it.

Use a standard, single-column structure with conventional section headings. Put the exact job title from the posting in your headline. Write dates as month and year, in one consistent format. Export a PDF with real selectable text, not a scan, not an image, not a design tool's flattened output. Name the file something a human can read.

Then tailor. Not by rewriting your history, but by reordering bullets so the relevant ones come first, adjusting the headline, and making sure the terms that describe things you have genuinely done appear in the sentences where you did them. Three or four changes per application is normal and takes ten minutes once you know which three.

White text keyword stuffing. It is detected trivially, it looks like fraud, and the recruiter who finds it will remember your name for the wrong reason.

Cramming every keyword from the posting into a skills list. It ranks you into an interview you cannot pass, which costs you more than the rejection would have.

Removing all formatting into a plain wall of text. Parsers are better than that, and a human still has to read it. Structure helps both.

Obsessing over a score from a free checker. Most of them measure word count and section presence. A number without a list of specific fixes is entertainment.