How Applicant Tracking Systems Really Work in 2026

The myth that will not die

You have read the story. A robot reads your resume, scores it against the job description, and if you fall below some percentage it is rejected before a human ever sees it. Seventy-five percent of resumes never reach a person, the statistic usually goes.

That figure has been repeated for years without anyone producing the study behind it, and the mechanism it describes is not how most companies use the software.

What an applicant tracking system does at the overwhelming majority of employers is more mundane. It takes your application, parses it into structured fields, and puts it in a database. A recruiter then searches that database. That is it.

The distinction matters because it changes what is worth fixing. If you believe a scoring algorithm is judging you, you will spend your time on keyword density. If you understand it is a database and a search box, you will spend it on being parsed correctly and being findable, which is where the real losses happen.

What actually causes automatic rejection

Automatic rejection does exist. It just does not usually come from your resume.

It comes from the application form. Those questions you click through before uploading anything are called knockout questions, and they are the genuine automated filter. Do you have the right to work in this country. Do you hold the required licence. Do you have at least five years of experience. Are you willing to work on site.

Answer one of those in a way the employer has marked as disqualifying and your application is filtered, regardless of how good the resume is. This is explicit, it is configured deliberately, and it is the thing people are actually experiencing when they say the robot rejected them.

So read those questions carefully, and answer them accurately. Where a question has a free-text box alongside it, use it. "Four years of direct experience plus two in an adjacent role" is a better answer than a bare number that trips a filter.

Parsing, which is where resumes are genuinely lost

The real damage happens quietly. Your resume goes in, the parser cannot make sense of the layout, and what lands in the database is a mangled record with your employment history scrambled or missing.

You are not rejected. You are simply not found, because the fields a recruiter searches are empty or wrong.

The layout choices that cause this are consistent and avoidable:

• Two-column layouts. Parsers often read straight across the page, interleaving the sidebar and the main body into nonsense.
• Text inside images or graphics. Extracts as nothing at all. A resume designed as a single image contains no readable text.
• Contact details in the header or footer. Some parsers skip those regions entirely, which is how a resume arrives with no phone number.
• Tables used for layout. Frequently collapse in unpredictable ways.
• Creative section headings. "My Journey" instead of Experience means the parser cannot identify the employment block.
• Dates in inconsistent formats. The parser needs to work out where each role starts and ends.

The fix for all of these is dull and effective: one column, standard headings, real text, consistent dates, contact details in the body.

How recruiters actually search

Once your resume is in the database correctly, the question becomes whether you surface when someone looks.

Recruiter searches are usually short. A job title, one or two hard skills, sometimes a certification or a named tool. Rarely anything elaborate.

That tells you which terms are worth having on the page and where. Job titles, named platforms, qualifications and settled technical vocabulary. It also tells you what is worthless as a keyword: nobody searches a candidate database for "hard working" or "results-oriented".

Recency matters too. Many systems weight or filter by how recently a resume was added, so an application from eight months ago may not appear in today's search even though it is stored. If a company reposts a role you applied to, applying again is reasonable rather than desperate.

Which file format to send

PDF, unless the employer asks for something else, and make sure it is a text PDF rather than a scan or an export that flattened everything into an image.

The old advice to always send a Word document dates from a period when parsers handled PDFs poorly. That has not been broadly true for years. What still matters is that the file contains selectable text.

A quick check: open your PDF, try to select a sentence, and copy it into a text editor. If it pastes as readable text in a sensible order, the parser will manage. If nothing selects, or the order is scrambled, that is exactly what the database will receive.

One clear takeaway: most resumes are not rejected by software, they are stored badly and never surfaced.

The tricks that do not work

Every few months a technique circulates promising to defeat the system.

White text on a white background, listing keywords invisibly. Parsers strip formatting, so the text usually arrives fully visible, and a recruiter reading a paragraph of stuffed keywords draws the obvious conclusion.

Keyword blocks in one-point type. Same outcome.

Pasting the entire job description into the document. Produces a resume that matches everything and reads as absurd to the human who opens it.

These share a common flaw. They are aimed at a scoring algorithm that mostly is not running, and they fail at the step that is actually happening, which is a person reading the document.

Where AI screening genuinely is being used

It would be wrong to say nothing has changed. Some larger employers now run models over applications to rank or summarise them, and AI screening is more common in high-volume hiring than it was.

What is worth understanding is that these systems generally assess relevance in a way closer to how a person reads than to keyword counting. A resume written to be clear, specific and evidenced tends to do well with them, and one stuffed with terms tends not to.

Which means the advice does not fork. The document that survives a recruiter's thirty-second read is the same one that survives an automated summary. That is convenient, and it is a good reason to stop optimising for an imagined robot.

Our guide to what actually beats ATS in 2026 goes further into how the screening layer has shifted.

What to do when you have applied and heard nothing

Silence is the most common outcome of an application and it tells you almost nothing about why.

Before assuming the software is at fault, check the two things you can actually verify. Open the confirmation email and see whether the system echoed your details back correctly, because a mangled name or a missing phone number is a parsing failure you can see. Then look at whether the role is still posted weeks later, which often means the search is ongoing rather than that you were dismissed.

Where a company reposts a role you applied to, apply again with a version tailored more closely. Many systems weight recent applications, and a resume submitted four months ago may simply not appear in today's search.

Following up directly is worth doing once, briefly, to a named person rather than a generic inbox. It does not defeat any filter, but it puts a human in the loop who can look you up, which is a different mechanism entirely from anything the software is doing.

A short checklist that covers most of it

• One column, no sidebar, no text boxes.
• Standard headings: Experience, Education, Skills.
• Contact details in the body of the document.
• Month and year on every role, formatted the same way throughout.
• Job titles that a recruiter would recognise, with an explanation where your internal title is unusual.
• Named tools and certifications spelled exactly as the industry writes them.
• A text PDF you can select and copy from.
• Knockout questions on the form answered accurately, with context where a box allows it.

Our free ATS scanner checks the parsing side and compares your document against a job description, which catches the two failure modes that matter without you having to guess.

Questions people ask

Do applicant tracking systems reject resumes automatically?

Not usually on resume content. Automatic filtering normally comes from knockout questions on the application form, such as work authorisation or a required licence. The resume is stored and searched rather than scored.

Is it true that 75% of resumes never reach a human?

That figure circulates widely without a credible source and does not describe how most employers configure the software. The more common outcome is a resume that parsed badly and was never surfaced in a search.

Should I send a PDF or a Word document?

PDF, unless the employer requests otherwise, and make sure it contains selectable text. The advice to avoid PDFs is out of date for the systems in common use.

Will a two-column resume get me rejected?

It will not trigger a rejection, but it frequently parses in the wrong order, which leaves you unfindable. A single column removes the risk for no real cost.

Does AI screening change what I should do?

Not much, and that is the useful part. Systems that summarise or rank with a model respond to clear, specific, evidenced writing, which is what persuades a human reader too.

Related reading

• Writing a Resume With Little or No Work Experience
• The Career Change Resume: Making Unrelated Experience Count