Does an ATS read tables, text boxes and headers?

Element by element: what survives extraction, what comes out scrambled, and the one that can make your phone number disappear completely. Plus the ten-second check.

Last reviewed 13 Aug 2026

Short answer, element by element: tables are read but often scrambled, text boxes are frequently skipped entirely, headers and footers are commonly ignored, hyperlinks keep their visible text and often lose the URL, and italics and real bullet points are fine. The word doing the work in every one of those is often: there is no single ATS, and the parsers behind Workday, Greenhouse, Lever, iCIMS, Taleo and Naukri do not behave identically.

So the useful question is not “is this element supported”. It is what happens to my page in the bad case, and whether that bad case loses information a recruiter needs.

Tables

A table is read. The problem is the order. A parser walks the cells, and a two-column table of Skills | Python, SQL comes out as a run of short fragments — sometimes column by column, sometimes row by row. Your dates end up detached from the job they belong to, and the system that tries to compute “years of experience” from your date ranges gets nonsense.

Tables used for a visible grid of data (a list of certifications with dates) usually survive. Tables used invisibly to lay out the whole page — the classic Word trick — are the ones that break things. If you cannot tell which you have, click inside your resume in Word and press Ctrl/Cmd + A: table borders light up even when they are set to “no border”.

Text boxes

This is the genuinely dangerous one. In a .docx, a text box is not part of the document flow — it is a floating object anchored to it. A large share of parsers read the flow and ignore the anchors, so text in a box can be completely absent from what the employer receives. Not reordered. Absent.

Designed templates love text boxes for the name block and the contact strip, which means the field most likely to vanish is your phone number.

Headers and footers

Treat the header and footer margins as decoration only. Many parsers skip them, and there is no upside to the risk: nothing in a resume needs to repeat on every page except, arguably, your name. Put your contact details in the body of the first page, where every parser looks.

A page number in a footer is harmless. An email address in one is a coin flip.

Hyperlinks

Extraction gets the visible text of a link reliably and the underlying URL unreliably. That is why linkedin.com/in/your-name written out as text beats the word LinkedIn hyperlinked to it. A human clicks either one; a parser only records the first.

The same rule applies to a portfolio, a GitHub profile and a certificate. Write the address you want recorded.

Italics, bold and bullet points

All fine, with one caveat that is about how the file was made rather than about the ATS. Bold and italic are formatting on top of characters, and the characters extract normally. Bullet points extract normally too — when the bullet is a real list marker.

We know that one precisely, because it is a defect we hit in our own renderer. A bullet drawn with an absolutely-positioned decoration instead of a list marker moves the line into a different paint layer, and PDF text is written in paint order — so every bullet extracted after every employer name, in a file that looked perfect on screen. It is now an automated test that runs against all 33 of our templates on every change.

What about graphics, icons and skill bars?

An icon is an image and carries no text, so a phone icon in place of the word “Phone” costs nothing — the number beside it is still text. A skill rating bar is different: it encodes the one thing it claims to say (how good you are) as a picture, so the extracted text is the skill name and nothing else. Every reader, human or machine, is left with an unqualified list.

The ten-second check that beats every rule above

Open your PDF, select all, copy, and paste into a plain text editor. That paste is approximately what the employer’s system receives. Read it in order. If your phone number is missing, if your dates are floating in a block of their own, or if nothing pastes at all — that is your answer, and it is worth more than any list of supported elements.

If nothing pastes, your resume is an image. That is a total failure and it is common: it happens with scans, phone photos, and design-tool exports set to rasterize.

Where this site fits

Our free checker does that paste for you and grades what came out — it shows the extracted text and the eight categories it scores, without an account. If you build here instead, the templates are the same ones the extraction test runs against: no text boxes, no layout tables, contact details in the body, links written as addresses.