What Is an ATS, and How Does It Actually Read Your Resume?

Almost every piece of resume advice mentions the ATS, and almost none of it explains what the thing actually does. The result is a lot of folklore: that it gives you a score, that it rejects you automatically, that there's a secret formula. None of that is quite right. An applicant tracking system is a database with a filing clerk attached, and once you know what it's filing, most of the advice stops being superstition and starts making obvious sense. If you already know the mechanics and just want the fixes, our ATS-friendly resume guide covers those directly, and our ATS-safe template collection is the shortcut — every layout there is built for the behaviour described below. When you want to test your own file, the five-minute self-test is free, and our explainer on "ATS scores" deals with the myth that's most likely to waste your money. The full ATS guide collection sits behind all of it.

An ATS is a filing system, not a judge

An applicant tracking system is software an employer uses to manage applications — collecting them, storing them, tracking who's at which stage, and letting recruiters search the pile. Workday, Greenhouse, Lever, Taleo and iCIMS are the names you'll meet most often. They are, at heart, databases.

That single fact clears up most of the confusion. A database doesn't have opinions. It doesn't score you out of 100. What it does is take an unstructured document — your resume, which to a computer is just a stream of text and shapes — and try to turn it into structured records: this string is a job title, that one is an employer, these are dates. Everything that goes wrong with resumes and ATS goes wrong at that translation step.

The four things it does to your file

1. It extracts the text

Before anything else, the system pulls raw text out of your file. A well-made PDF or a .docx contains real, selectable characters, and extraction is clean. A resume that's been flattened into an image — a screenshot, a scan, or a design exported as a picture — contains no text at all, and extraction returns essentially nothing.

2. It parses that text into fields

This is the interesting part. The parser reads the extracted text in order and tries to assign it to known fields: name, contact details, work history, education, skills. It leans on two signals to do this — recognisable section headings, and reading order.

Headings tell it where it is. "Experience" and "Work History" map to something the system understands. "My Journey" and "Where I've Made an Impact" do not, and the content underneath may never get filed as work history at all, however good it is.

Reading order tells it what belongs to what. The parser walks the text as a sequence. If your layout puts a sidebar next to your main column, the extracted sequence may interleave the two — a line of skills, a line of job description, another line of skills. To you it's two tidy columns. To the parser it's one scrambled stream.

3. It stores the result as a record

What ends up in the database isn't your resume as you designed it. It's a structured record: fields, values, dates. Your careful typography is gone. Whatever the parser got wrong is now simply what the system believes about you.

4. It surfaces you when someone searches

Then a recruiter searches — for a job title, a skill, a certification, a location. The system returns matching records, ranked by whatever criteria that employer configured. This is the step people imagine as "the ATS rejecting you." Usually nothing rejected you. You just weren't in the result set, because the record built in step 3 didn't contain the words being searched for.

Where it actually goes wrong

Three failures account for most of it:

• Content trapped in the wrong container. Text inside headers, footers, text boxes or graphics is frequently skipped. Contact details in a decorative header band are the classic version — the resume parses fine, and nobody can contact you.
• Reading order scrambled by columns. Any layout with a genuine sidebar risks this. It's why our ATS-safe templates are single-column throughout.
• Vocabulary mismatch. You wrote "people operations," the recruiter searched "HR." Both are correct; only one is in the record. Matching the language of the job posting is not gaming the system, it's making yourself findable — though there's a real limit, which our guide to keyword stuffing draws.

What this means when you write

Almost nothing about being ATS-readable conflicts with being good. Use standard section headings. Keep one clear reading order. Put real text on the page instead of pictures of text. Write dates in a computable format. Use the words the industry actually uses.

The one genuine tension is visual: heavily designed, multi-column resumes parse worst. That's a real trade-off, not a myth, and where it bites hardest is covered in our guide to satisfying both readers.

Frequently asked questions

Does an ATS automatically reject resumes? Rarely in the way people imagine. Some systems apply knockout questions — work authorisation, a required certification — and those do filter automatically. But the common experience of "the ATS rejected me" is usually not being surfaced in a recruiter's search, which is a visibility problem rather than a rejection.

Do all applicant tracking systems work the same way? No. Parsing quality varies considerably between vendors and versions, and each employer configures their own. That's exactly why there's no universal score to optimise against — and why building for reliable parsing across systems beats targeting any one of them.

Is a PDF or a Word file safer? Both parse well when generated properly. A text-based PDF is the safer default because it preserves your layout for the human reader afterwards. Our comparison of PDF and Word covers the exceptions, including postings that explicitly ask for .docx.

How do I know whether my own resume parses correctly? Copy the text out of your exported file and paste it into a plain text editor. What you see is close to what the parser sees. The full self-test takes about five minutes and costs nothing.