What makes a resume ATS-friendly, and the 75% rejection myth

Most advice about applicant tracking systems starts with a statistic: three resumes in four are rejected by software before a person sees them. The figure is everywhere, and it has no source anyone can check. What is real is narrower and more useful: a parser can mangle a resume it cannot read, and the layouts that cause that are easy to avoid.

Where the 75% figure comes from

The claim traces back to a 2012 sales pitch from Preptel, a company that sold resume optimisation and closed the following year. It never published how the number was measured.

When Enhancv interviewed 25 US recruiters in September and October 2025 about how they configure their systems, 23 said they do not set up automatic rejection on resume content. The usual filters are knockout questions on the application form, such as work authorisation or a required licence, not a judgement on your layout.

So most resumes that go unanswered were read by nobody because there were 300 others, not because software deleted them. That changes what is worth optimising. The goal is not to beat a robot; it is to make sure that when a recruiter searches the system for “Figma” or “logistics”, the parser has put your words in the fields they search.

What a parser does with your file

A tracking system extracts the text from your PDF or Word file, then guesses which line is your name, which block is a job, and which string is a date. It works through the text in the order the file gives it, top to bottom.

That is where layout matters. A two-column resume looks orderly to a person, but the text inside the file can come out interleaved: a line of skills from the sidebar, then half a job title, then a phone number. The parser then files “Python” as your employer and your employer as a skill.

Layout choice	What can happen when it is parsed	Safer alternative
Two columns or a sidebar	Reading order interleaves the columns	One column, top to bottom
Dates in a table cell	Table skipped, or dates split from the job	Dates on the job line, right-aligned with a tab
Contact details in the document header	Header often dropped entirely	Name, email and phone as the first lines of the page
Text in a text box or shape	Box skipped or read last	Plain paragraphs
A PDF exported as an image	No text at all to extract	A PDF with real, selectable text
Icons for phone and email	Read as nothing, or as stray symbols	The words written out

None of these come from a controlled study; they are consistent across vendor documentation and recruiters’ reports, which is the best evidence there is for systems whose parsers are private.

The format that parses

Put your name first, then your job title, then one line of contact details: email, phone, city and a profile link. Then the sections, each under a plain heading a parser expects: Summary, Experience, Education, Skills.

For each job, the title and dates go on one line, the company and location on the next, then two to five bullets (how to write resume bullet points has rewrites). Start each bullet with what you did and end with what changed, and include a figure where you have one: “Cut the time to send an invoice from 4 minutes to 90 seconds” is searchable and checkable in a way “Responsible for invoicing” is not.

Use a standard font. The PDF standard fonts, Helvetica and Times, are present in every PDF reader and every parser, so nothing depends on an embedded font being decoded correctly.

PDF or Word

Send a PDF where a person will read it, because it looks the same on every screen. A text-based PDF parses in most current systems. Keep a Word file for portals that ask for one, or that reject a PDF on upload; some of the strictest systems handle .docx most consistently.

You will see figures claiming Word files parse 97% correctly and PDFs 72%. They come from tests whose method was never published, so treat them as marketing. Having both files ready costs nothing.

One page or two

The one-page rule is folklore for anyone past their first job. In a simulated hiring study ResumeGo ran with 482 recruiters and hiring managers, 5,375 of the 7,712 resumes they picked were two pages long, and two-page resumes were preferred 2.3 times as often overall: 1.4 times at entry level, 2.6 times mid-career and 2.9 times for managers. It is a vendor-run study, but it is the largest test of the question there is. Fit what earns its place, and stop at two outside academia; how long a resume should be has the detail by career stage.

Test your own resume

The test a parser runs is text extraction, and you can run it yourself. Open your PDF, select all the text and paste it into a plain text editor. Read what comes out from top to bottom.

If your name, email and phone appear first, every job title sits next to its dates, and nothing from a sidebar has been spliced into the middle of a job, a parser will read it the same way. If the text comes out scrambled, so will your record in the system. Running the PDF through PDF to text shows the same extraction without the copy and paste.

To see what a parser does with that text, open the file in the ATS resume checker, which shows the reading order and the fields a parser finds. It numbers every piece of text on your page in the order it is read, lists the name, contact details and jobs it found, and names each problem, such as columns read across or contact details only in the document header. It reads PDF and Word files in your browser, and gives no score, because no tracking system publishes one.

Building one that passes by construction

The resume builder on this site has 16 templates: 12 are one column, and the 4 with a sidebar each download as a one-column PDF too, with the same content. Contact details sit in the page body and every character is real text. Every build of the site renders a sample resume in each template, extracts it with pdf.js and fails if the 16 fields it checks come back out of order, on A4 and US Letter.

It downloads the PDF and a Word file free, with no account, and the resume stays in your browser. Instead of a score, it lists suggestions with their reasons: a job without dates, bullets that list duties rather than results, a summary over 80 words.