Correction, 17 October 2026. Earlier versions of this article described a test of 20 Canva templates run through "a Workday-style parser configured with the same library stack Workday uses internally", with 14 failing, and promised screenshots. We have no records of that test, the article contained no screenshots, and we do not know what software Workday runs internally: our scanner is our own parser. We have removed those claims. This version reports a test we can show and the failure patterns behind it.
- A Canva-style resume with a sidebar scored 80 out of 100 in our scanner, with one critical flag: the two-column layout scrambled the reading order.
- The parser read the whole sidebar first: contact, skills, languages and education all came before the profile and the work history.
- Icons became stray characters in front of the email, phone and city, and a languages table glued into "EnglishNative" and "FrenchB2".
- The words survive; the order and the clean fields do not. Use a single-column file for applications and keep the designed one for people.
Canva templates are some of the most downloaded resume designs anywhere, and they are the files I see go wrong most often in our scanner. So instead of arguing about it, I built a resume that looks like the popular marketplace designs and ran it through the parser to see exactly what comes out.
What I tested
The test file is a one-page PDF for a fictional marketing manager, built to copy the look people recognise from template marketplaces: a coloured sidebar holding contact details, skills, a languages table and education; small icons in front of the email, phone and city; a decorative name band; and a main column with the profile and work history. On 17 October 2026 I scanned it on the live site, the same scanner every visitor uses.
The result: 80 out of 100, with one critical finding (the two-column layout scrambled the reading order) and one warning (year-only dates such as "2021–Present", which some systems read less precisely than month and year).
An 80 sounds fine. It is not the whole story, because the score counts what is wrong, and what is wrong here is the part a recruiter's software leans on most.
What the parser actually read
This is the start of the text our parser pulled out, in the order it came out:
DANIEL OKAFOR Marketing Manager · Growth & Brand CONTACT 'daniel.okafor@example.com &+44 7700 900123 #Manchester, UK SKILLS Brand Strategy Paid Social ... LANGUAGES EnglishNative FrenchB2 YorubaNative EDUCATION BA (Hons) Marketing ... PROFILE Growth-minded marketing manager with 7+ years ... EXPERIENCE Marketing Manager 2021–Present
Three things went wrong, and each one is typical of designed templates.
1. The sidebar was read before the experience
On the page, your eye goes to the name, then the profile, then the jobs. The parser does not see the page; it reads the text in the order it sits in the PDF, and in this design the sidebar text sits first. So contact, skills, languages and education all arrive before the profile and the work history. In other files the sidebar gets interleaved with the jobs instead, which is worse: skills land between bullets and dates drift away from the roles they belong to. The full mechanism is in are two-column resumes ATS-friendly?
2. Icons turned into stray characters
The little envelope, phone and pin icons are characters from an icon font. In the extracted text they came out as ', & and # glued to the front of the email, phone number and city, and the decorative bullets came out as '¤. A system that tries to recognise an email address or phone number by its shape can miss one with junk stuck to the front of it.
3. The languages table glued together
The languages sat in a two-column table, language on the left and level on the right. The parser read each row as one string, so "English" and "Native" became EnglishNative. A recruiter searching for "English" or "French" may not match a word that no longer exists on its own. Tables in general are covered in do ATS read tables?
Real files tell the same story. In 8,220 resumes I scanned in October 2026, 13.9% of Canva files had their columns read out of order, three to four times the rate of Word or Google Docs files.
Other Canva features that cause the same kind of damage
The test file did not include every popular flourish. These are the others I see most often in scans, and why they are risky:
- Name as an image or heavily styled graphic. If the name is a picture, there is no text to extract, so the name field comes back empty.
- Skill bars and rating dots. "Python ●●●●○" carries the level as a shape. The parser keeps the word, drops the level, and the shapes can confuse where the skills section ends.
- Contact details in a header or footer band. Some parsers skip header and footer text. You cannot tell which system an employer uses, so keep contact details in the body.
- Creative section headings. "My Story" or "What I've Built" may not be recognised as a summary or experience section. "Experience", "Education" and "Skills" always are.
Does this matter for Workday?
Workday is used by many large employers, so it is the system people ask about most. I cannot run files through Workday's own parser, and nobody outside Workday knows exactly how it is built, so I will not claim a Workday result. What I can say is that the failures above come from how the PDF itself is built, not from one vendor's quirks: the text order, the icon characters and the glued table cells are in the file before any ATS opens it. The Workday-specific advice is in the Workday ATS guide and will Workday read my PDF resume?
What to do if your resume is a Canva template
- Keep two versions. The designed Canva file for people: networking, interviews, email. A single-column file for online applications.
- Make the application version plain. One column, no sidebar, no tables, no icons, no skill bars. Name and contact details as ordinary text at the top. Standard headings.
- Write skills and languages as text. "English (native), French (B2)" survives; a two-column table may not.
- Use month and year for dates. "Jan 2021 – Present" is read more reliably than "2021–Present".
- Check the exact file you will send. Run the copy-paste test, or upload it to our free scanner and read the extracted text for scrambled order, stray symbols and glued words.
If you would rather not rebuild it by hand, the $5 rebuild turns your existing content into a single-column file, and you can then pick from 13 designs, each one checked in our parser against your own content before it is offered. For how AI-written resumes fare with the same problem, see I tested ChatGPT and Claude resumes.
The honest verdict
A designed Canva resume does not vanish in an ATS; in my test every word was still there. What it loses is order and clean fields: the sidebar before the experience, symbols stuck to the contact details, words glued together. That is enough to make a parsed profile harder to search and harder to read. For applications, use a single column and check the file. Keep the design for people.
→ Scan your Canva resume free and see the text an ATS would get.
