What an ATS actually does to your resume

It is a database, not a judge

An Applicant Tracking System is the database an employer’s hiring software keeps. When you apply, it parses your resume. Software reads the file and tries to pull out your name, contact details, employers, titles, dates, education and skills, and drops each into a field.

A recruiter then searches that database. This is the part most people get wrong: there is usually no robot scoring you and rejecting you. There is a search, and you are either findable in it or you are not. If parsing garbles your resume, you are in there as noise, and no search finds noise.

What it reads well

• Ordinary section names, not clever ones. “Professional Experience”.
• A single reading order down the page
• Real selectable text, never a picture of a resume
• A recognizable date range next to each employer
• Words that genuinely match the posting, used in context

What genuinely breaks it: layout

Text in tables, headers and footers, or a two-column sidebar can come out interleaved, so your job titles land inside your skills list, and your dates bind to the wrong employer. The document still looks perfect to you. It is unreadable to the parser.

That is the one honest trade-off in resume design. Two-column layouts look better to a person and are riskier with software. Use one when you know a human screens first: a referral, a recruiter you have spoken to, a small company. Use a single column when you are applying cold.

The part that does reject you automatically

Everything above is about your resume, and your resume is not where automatic rejections happen. They happen on the form you filled in on the way to uploading it.

Most application forms carry knockout questions, and unlike the parse, these genuinely do reject you on the spot, with no human involved. They are usually some of:

• Are you legally authorised to work in [country]?
• Will you now or in the future require sponsorship?
• Do you have a bachelor’s degree?
• How many years of experience do you have with [thing]?
• Are you willing to work on site in [city]?
• What are your salary expectations?

This is the robot people are afraid of, and it is not reading their resume. It is reading six answers given in about nine seconds, on the way to the part they spent three weeks on.

What to do about it: slow down on those six questions and answer them deliberately. Do not round your years of experience down out of modesty — if you have three years and eleven months, that is four. Do not leave a salary field blank when it accepts a range, because a blank is sometimes scored as zero and sometimes as a refusal. And do not answer “no” to a willingness question you would in fact say yes to if asked properly; if the honest answer is “it depends”, the form cannot hear that, so answer for the situation you would actually accept.

Lying on them is a different matter, and it is a bad trade. Work authorisation and degree answers are verified later, and being caught costs you the offer at the point where you have already turned other things down.

The salary box, which is the one that costs most

It is a knockout question like the others, and it is the one people answer worst, because it is the only one where being honest and being strategic pull in opposite directions.

Leaving it blank is rarely safe: some systems reject the form outright, and some record it as a zero. Writing “negotiable” in a numeric field does the same. If the field takes text, a range is better than a number, and the range should start at what you would actually accept — because the bottom of your range is the offer you will get.

Anchoring low is the expensive mistake, and it is the common one. A number below the band the role was budgeted for does not make you the cheap, easy hire; it often reads as a mismatch of level and gets filtered in the same pass. If you genuinely do not know the band, the honest research is fifteen minutes: look at what the same title pays at three comparable employers in the same city, and take the middle.

Several US states and cities now require the employer to publish the band on the posting. If it is there, use it — asking for something inside a range they have already printed is not a negotiation, it is agreement.

Which file to send

PDF preserves your layout everywhere and is the right default. There are two exceptions worth knowing.

If the posting or the upload field says Word is preferred, send Word. Some older systems parse .docx more reliably than PDF, and a company that says so is telling you what theirs does. And if you have exported a PDF from a design tool rather than a word processor, check it is real text and not an image — select some and try to copy it.

One more thing nobody warns you about: several large systems, Workday among them, ask you to upload the resume and then present the parsed result as a form for you to correct. That form is what the recruiter sees, not your beautiful PDF. It is worth the four minutes to fix it, because a resume that parsed badly and was then left uncorrected is a resume you effectively did not send.

Myths worth dropping

“Colour and fonts get you rejected.”

They do not. Colour, typeface, bullet shape and margins are for the human who reads you after the search. Nobody was ever filtered out for using navy.

“Hide keywords in white text.”

The parser reads the text, not the colour, so a recruiter sees a wall of hidden keywords in the plain-text view. It reads as an attempt to cheat, because it is one.

“Stuff in as many keywords as possible.”

Search matches on context, and a keyword you cannot speak to survives about five minutes of an interview. Match the words you can defend.

“A tool can tell you your pass rate.”

No tool has access to an employer’s configured screening rules, which differ per company and per posting. Any specific percentage is invented. What can honestly be checked is whether your document parses cleanly and whether it uses the words the posting uses.

A checklist you can run yourself

1. Open your PDF, select all, and paste into a plain text editor.
2. Read what comes out in order. That is roughly what the parser sees.
3. If titles, dates and employers are interleaved or out of order, your layout is the problem, before a single word of it is.
4. Check your contact details survived. They are often in a header, and headers vanish.
5. Check the section names are ordinary ones.

Costs nothing, takes two minutes, and catches most of what actually filters people out.