Short answer: yes, an ATS reads the text inside columns and tables. Words almost never go missing. What breaks is the order. The program dumps the file's contents in the sequence the file hands them over, and in a two-column document that sequence is not the one you see on screen. That is why a two-column resume can deliver the sidebar languages before your name, or six section headings in a row with nothing underneath them.
Below is the real extraction from a two-column resume, line by line, plus the two-minute check that tells you whether yours is affected.
The misunderstanding: "reading" is not the same as understanding the layout
A PDF does not store columns. It stores fragments of text, each with a position. When the machine extracts the content, it walks those fragments and writes them out one after another. If the file was built with the sidebar first and the main column second, that is what comes out, even though you see the two columns side by side.
The consequence matters more than it looks. A parser does not store your resume as running text: it tries to sort it into fields (name, title, company, dates, education, skills), and to do that it leans on how close each section heading sits to the text that follows. When the order breaks, that closeness breaks with it, and the data ends up filed in the wrong place. For the detail on which fields it tries to fill, see what a resume parser reads.
The proof: a two-column template, extracted as-is
This extraction comes from our test with real templates, where we exported each resume with its own placeholder content and ran the PDF through text-extraction software. It is a two-column marketing template with a sidebar. This is how the extracted text opens, verbatim (the template ships in Spanish; the finding is about layout, not language):
Four observable things, none of them opinion:
- The document opens with the languages. Those three lines are sidebar content, and they arrive before the job title and before the person's name.
- Six section headings arrive stuck together, in a row, with no content. Contact, Languages, References, Education, Experience, Courses. The bullets that follow can no longer be attributed to any of them.
- Labels get separated from their values. Further down the same extraction, "Mail:" and "Dirección:" arrive back to back and the email address shows up later. A parser expecting a label-value pair matches them wrongly.
- The skill level bars come out as loose numbers: 40%, 50%, 74%, with no skill next to them.
That is the whole problem. No text is missing. It is all there. It just arrives in an order you cannot reconstruct a person from.
The full test, with four templates and their complete extractions, is in is your Canva resume ATS-friendly?.
What about tables?
Here it is worth being exact about what we tested and what we did not. Our extraction used templates laid out with text boxes and columns, not with tables. But the mechanism that fails is the same, and the outcome is usually worse, for a specific reason: a table is walked cell by cell, and the walk order depends on how the table was built, not on how you read it.
In practice that produces two typical kinds of damage:
- Rows that read as a single line. If you put the job title in one cell and the dates in the next, they can arrive glued together ("Data Analyst 2021 - 2024 Acme Inc.") or several lines apart.
- Columns read end to end before moving to the next one. All the dates first, all the titles after, at which point there is no way to tell which date belongs to which job.
It is also why university career services that publish ATS guidance, such as UConn and UMN Duluth, put tables and columns on the same avoid list as text boxes, icons, and anything placed in a header or footer.
The copy-paste test: two minutes, nothing to install
You do not need a tool to find out whether your resume is affected:
- Open your resume as a PDF.
- Select all the text (Ctrl+A or Cmd+A).
- Copy it.
- Paste it into a plain text editor, like Notepad or TextEdit in plain-text mode.
What shows up there is, in essence, what the machine receives. If it reads top to bottom in the order you expect, your layout holds. If your name appears halfway down, if you see section headings in a row with no content, or if dates sit far from their jobs, you now know exactly what reaches the parser.
One detail that fools people: text being selectable does not mean it comes out in order. Those are two different problems. A scanned resume, or one exported as an image, hands over no text at all; a two-column resume hands over all of the text, scrambled. The second case is the more common one and the harder to spot by eye.
What to do if your resume is in two columns
You do not need to rebuild it blindly. In order of effort:
If the copy-paste test comes out in order, change nothing. Two-column documents that export cleanly do exist. What you cannot do is assume it.
If it comes out scrambled, move the content to a single column. That is the change that fixes the problem at the root, because it removes the ambiguity instead of working around it: with one column there is only one possible order and it matches the one you see. You can copy the structure section by section from our free ATS resume template.
Keep the good-looking version for something else. If you like your two-column design, hold on to it for emailing a specific person, for your portfolio, or for a careers fair. The file you upload to a portal is a different file, and it does not have to be the same one.
A note on skill level bars: even if your layout holds, those bars tell the parser nothing. They are graphics. The machine keeps the loose number or nothing at all. If a skill matters, write it out in words in the skills section and back it up in a job description.