Applicants have been told for a decade that a robot throws most resumes away before a human ever sees them. That story is mostly wrong, and believing it leads people to spend their time fixing the wrong things.

An applicant tracking system — an ATS — is not a gatekeeper with an opinion. It is a database. It takes in applications, stores them, lets a recruiter search and sort them, and tracks what stage each candidate has reached. Workday, Greenhouse, Lever, iCIMS and Taleo are all doing that same job with different buttons.

Understanding what it does, and specifically what it does not do, changes what is worth your hour.

The 75-percent myth

You will see the claim that applicant tracking systems automatically reject three quarters of resumes. It has been repeated so often that it now reads as established fact. It is not. It traces back to vendor marketing and to a widely misread figure about how many applicants are not selected — which, for a role with two hundred applicants and one opening, is 99.5 percent whether any software is involved or not.

Most systems, in their default configuration, reject nobody on the strength of a resume. They rank, they tag, and they wait for a person. Someone does still read yours; the real questions are whether they read it early, late, or never, and whether what they read is what you actually wrote.

What genuinely filters you out

Three things do real filtering, and only one of them is about keywords.

1. Knockout questions

These are the questions in the application form itself, not on your resume: are you legally authorised to work here, will you require sponsorship, do you have five years of experience with this particular thing, are you within commuting distance, what are your salary expectations. Answers to these are frequently used to disqualify automatically, and that is a genuine automated rejection.

This is where most real auto-rejection happens, and almost nobody optimises for it. Answer the form as carefully as you wrote the resume. A rushed answer here undoes a perfect document.

2. The recruiter's search

A recruiter facing three hundred applications does not read three hundred resumes. They search the database — the same way you search your own email — for the two or three terms that define the role. If the posting is for "payroll" and your resume only ever says "compensation processing", you are in the pile, but you are not in the search results, and the pile is where applications go to be forgotten.

This is the real reason to match the language of the job description: not to please an algorithm, but to be findable by a person using a search box.

3. Parsing failure

Before any of that, the system has to convert your document into structured fields — name, employer, job title, dates, skills. When that conversion goes wrong, your resume is still in there, but the searchable version of it is mangled. This is the failure mode people are actually worried about when they worry about "the ATS", and it is the one you can test in two minutes.

The formatting that actually breaks parsers

Most resume advice on this topic is superstition. These are the items with a real mechanism behind them:

  • Text inside images. A resume exported as a picture, or a skills section drawn as a graphic, contains no text to extract. This is the one formatting mistake that is reliably fatal.
  • Multi-column layouts. A two-column design is read straight across both columns by some parsers, interleaving your job titles with your skills list into nonsense. Common in exactly the attractive templates people pay for.
  • Tables and text boxes. The same problem, and less obvious, because the document still looks fine to you.
  • Headers and footers. Some parsers ignore them entirely. Contact details belong in the body of the page.
  • Unusual date formats. "Spring '22 – present" is readable by you and not by a date parser. Write March 2022 – Present.
  • Invented section names. "Where I've Made an Impact" is a nice line and it is not a heading any parser recognises. Use Experience, Education, Skills.

What matters less than you have been told: the file format, provided the text is real text. A PDF exported from a word processor extracts cleanly in every modern system. If an employer's form specifically asks for a .docx, give it one; otherwise either is fine.

The two-minute test that beats all of this advice

You do not need to guess about any of it. Open your resume, select all, copy, and paste it into a completely plain text editor — Notepad, TextEdit switched to plain text, or the body of a blank email.

What appears is approximately what the parser sees. Read it in order. If your job titles are interleaved with your skills, if your dates have vanished, if your phone number is nowhere, you have found a real problem. If it reads as a slightly ugly but correct document, your formatting is fine and you can stop worrying about this entirely.

Almost everyone who runs this test learns one of two things: that their resume was never at risk, or that a template they paid for scrambles it completely.

Where AI helps, and where it makes things worse

The genuinely useful application is comparison, not generation. Put a job description and your resume in front of an assistant together and ask which requirements in the posting have no corresponding evidence in your resume, and which of your real accomplishments the posting cares about that you have buried on page two. That produces a short, specific list of edits, all of them about things you have actually done.

The unhelpful application is asking it to write the resume from nothing. It will produce fluent, generic, achievement-shaped sentences that read like every other AI-written resume a recruiter saw that week, and it will occasionally invent a metric you cannot defend in an interview.

And one to avoid outright: keyword stuffing — pasting the job description in white text, or listing skills you do not have. Parsers strip formatting, so white text arrives as ordinary visible text in the recruiter's view. It looks exactly like what it is.

What to do with an hour

  1. Run the copy-paste test. Fix the layout only if it fails. Fifteen minutes, and for many people the job ends here.
  2. Match the job title. If the posting says "Operations Coordinator" and your last role was internally called "Business Support Lead", the posting's title belongs somewhere on your resume — in a summary line, honestly described.
  3. Mirror the vocabulary you legitimately own. Their words for your real skills, not your words.
  4. Answer the application form as carefully as the resume. This is where the automatic rejections actually live.
  5. Apply through the company's own careers page where you can. Applications forwarded from aggregators more often arrive with fields missing.

None of this is about outwitting software. It is about writing a document that survives being converted into a database row, and then being findable by the person searching that database. The system is far less clever, and far less hostile, than the folklore around it suggests.