Are Canva-style and two-column resumes ATS-friendly? We tested four layouts

Design templates are everywhere: a dark header, a tinted sidebar, skill bars, a photo circle. They look great, and they're one search away. The question everyone asks is whether software can read them.

Rather than repeat advice, we tested it. We built one fictional resume in four layouts, saved each as a real-text PDF, and ran every file through a resume parser to see what came out.

The test

The person is fictional: Kavya Raman, a data analyst in Bengaluru with one job and one internship. All four files hold the same words, character for character. Only the layout changes.

1. One column. Name, contact line, then Summary, Experience, Education, Skills, Certifications, top to bottom.
2. Two columns with a sidebar. Contact, skills, education and certificates in a tinted left column; name, summary and experience on the right.
3. A design-tool style layout. Built to look like a typical template: a full-width header band, positioned text boxes, skill bars, a date timeline in its own column, and spaced-out capital headings. We built it ourselves in HTML, so it is like a Canva export, not one.
4. A table layout. The Word habit of putting dates in a left cell and the role in the right, plus a table for education and skills.

Each was saved to PDF by Chromium, the way a browser's "Save as PDF" writes real, selectable text. Then we read each PDF with pdf.js, the open-source library Resumeize uses for uploads, and our own import, which rebuilds sections, roles, dates and skills. For a second opinion we also ran pdftotext, a widely used open-source extractor.

The results

Almost every word made it out of every file. What broke was order and grouping: which words belong to which role, and which heading a line sits under. That's exactly what a recruiter's search and an application form rely on.

What broke in each layout

Two columns: the same file, two different failures

Our parser saw the gap between the columns and read the sidebar first, then the main column. Every role, date and skill was right, but the first line of the file was "CONTACT" and the name came after the certificates, so the phone number was taken as the name and the real name was filed under Certifications.

pdftotext read the same page straight across instead, row by row. Sidebar lines landed in the middle of the job: the six skills, the degree, the school and the certificate came between the Experience heading and the Data Analyst job's first bullet.

Same PDF, two parsers, two different errors. That's the real risk of columns: you can't know which way the employer's software reads.

The design-tool layout: both roles lost

Text boxes placed side by side are read by height on the page, not by box. The dates sat in their own little column to the left of each job, so each range broke over two lines, and its second half, "– Present", was taken for a bullet point. The "About me" text box ran alongside the experience and was interleaved with it, line by line. Of five bullets, only one came out in one piece.

The spaced-out headings broke too. The name, set in wide capitals, came out as "KAVYARAMAN" in our parser and "K AV YA R A M A N" in pdftotext, and "WORK EXPERIENCE" became "WORKEXPERIENCE", so the Experience section wasn't recognised at all. Our import found no dated roles, and the internship's lines ended up mixed into the Skills list.

The table: dates parted from their roles

Our parser decided the table's left column was a sidebar and read it first: every section label and both date ranges, then all the content. Experience was left holding two date ranges with no titles, and everything else, from the email address to both jobs, the degree and the skills, was filed under the last heading it had seen: Certifications. pdftotext read the same table in nearly the right order, with only the degree's dates out of place.

One column: nothing to fix

Both parsers read the one-column file in order: name first, every role with its title, employer and dates, every bullet whole, every section where it belongs.

Why design templates trip parsers

A PDF doesn't store "this is my sidebar". It stores pieces of text and where each sits on the page. A parser has to guess the reading order from positions. One column leaves only one sensible guess. Columns, text boxes and tables leave several, and each parser picks differently.

Employers say the same. Roche's careers site, for one, warns applicants against tables, text boxes, logos and graphics, notes that some systems "read columns straight across", and says header and footer text sometimes gets dropped completely.

Human readers aren't kinder. In the Ladders eye-tracking study, resumes with multiple columns and cluttered layouts did worse with recruiters, while simple layouts with clear headings did better. How recruiters read a resume covers what they look at first.

If you design in Canva anyway

Canva says it estimates the reading order automatically when you export a PDF, and lets you set it yourself: in the Layers panel the bottom layer is read first, and the PDF download has a "Match reading order to layers" setting. We didn't test Canva's own export, but if you use it:

• Choose a one-column template. If it has a sidebar, keep your name above it, across the full width.
• Set the reading order: in the Layers panel, move what should be read first (your name) to the bottom of the list, then contact details, then each section in order.
• Use plain headings without wide letter-spacing: Experience, Education, Skills.
• Keep each role's title, employer and dates in one text box, not in separate boxes.
• Skip skill bars and star ratings. They're shapes, and say nothing a parser can read.
• Download as PDF and run the copy-paste test below before you apply.

The simpler fix

You don't have to choose between looking good and being read. A clean one-column layout with good type, spacing and bold titles looks professional and parses the same way everywhere. Our free ATS-friendly Word templates are built exactly that way, and what an ATS actually does explains what happens to your file after you click Submit.

We keep the script and the four PDFs so the test can be re-run; they're available on request. The resume is fictional throughout.