What applicant tracking systems actually do

Applicant tracking systems have acquired a mythology: a machine that scores your résumé, ranks you against everyone else, and rejects you before a person ever sees it. This drives a small industry of keyword-stuffing advice, white text tricks, and services that promise to optimise your "ATS score".

The reality is more mundane. An ATS is primarily a database with a workflow attached. Its main job is to store applications, move them through stages, record who said what, and produce the audit trail an organisation needs. Ranking and parsing are features on top, and at many employers they are barely used. Recruitment systems should not be confused with employee monitoring software, which is designed for post-hire work visibility rather than candidate screening.

What the software actually does

• Stores the application and the uploaded file.
• Parses the file into structured fields — name, contact details, employment history, education — with imperfect accuracy.
• Runs knockout questions, which are the genuinely automatic rejections. "Do you have the legal right to work here?" "Do you hold a current registration?" A wrong answer here closes the application without human involvement.
• Allows the recruiter to search and filter the pool, usually by keyword.
• Tracks stage, notes, and interview scheduling.

Note what is not on that list: an automated quality score that decides your fate. Some systems offer match percentages and some recruiters glance at them, but the dominant mechanism is far simpler — a recruiter searches the pool for a term, and you are either in the results or not. NIST provides a useful reference for technology and security terminology.

Parsing failures, which are real

This is where applications genuinely die, and it is entirely preventable.

• Multi-column layouts. Parsers read left to right across the page and interleave the columns into scrambled text.
• Text inside images, logos, or graphics. Invisible to the parser.
• Tables used for layout. Sometimes fine, frequently scrambled.
• Text in headers and footers. Often ignored entirely — which is why contact details belong in the body of the document.
• Unusual fonts, or PDFs exported in a way that embeds text as outlines. Test by opening the PDF and trying to select the text. If you cannot select it, neither can the parser.
• Date formats that are not recognisable. "Jan 2019 – Mar 2022" parses. "'19–'22" often does not.
• Section headings the parser does not recognise. "Professional Journey" may not register as employment history where "Experience" or "Work Experience" would.

A single-column document, standard section headings, ordinary fonts, and dates in a plain format solves nearly all of it. This is the same advice as writing for a human skim, which is convenient. Questions about discrimination and fair hiring are covered by the EEOC.

Keywords, honestly

Keywords matter, but not in the way the folklore suggests. The recruiter searching the pool types the terms from the posting. If the posting says "accounts payable" and your résumé only says "AP", you may not appear. If it names a specific system and you have used it, the name should be on the page.

So: use the employer's vocabulary for things you have genuinely done. Write out abbreviations at least once with the short form in brackets. Include the specific tool names you have actually used. That is the whole technique.

What does not work, and carries real cost:

• White text or hidden keyword blocks. Recruiters see them the moment the file is opened or converted, and it is treated as dishonesty.
• A keyword salad section listing forty technologies you have touched once. It parses fine and reads as desperate to the human who arrives next.
• Listing skills you do not have because they appeared in the posting. This survives the search and fails at the screening call, having wasted a fortnight.

The application form itself

Many candidates upload a careful PDF and then rush the form, which is backwards — the form fields are what the system actually searches and filters on. If the parser has mangled your history into the wrong boxes, fix it. It is tedious and it is the highest-value fifteen minutes in the process.

Two fields deserve particular care. Knockout questions should be answered accurately, because a wrong answer here is one of the few genuinely automatic rejections and cannot be appealed. Salary expectation fields, where mandatory, are covered separately — but never leave one blank if it is required and never enter a placeholder like zero, which sorts you to the bottom of every filter.

Applying more than once

Most systems keep your record and show the recruiter your application history. This is worth knowing for two reasons: a previous rejection is visible, and so is a pattern of applying to twelve unrelated roles at the same employer in a month, which reads badly. Applying again after a genuine change — new certification, more relevant experience, a role that fits better — is normal and fine.

Where the human takes over

Once you are in the shortlist, the software stops mattering entirely. Everything after that point is people reading, calling, and forming opinions. The ATS is a filter at the door, not a judge in the room, and treating it as the main obstacle leads people to optimise a document for a machine that was never going to be the hard part.