Where a Skill Has to Appear to Be Found
A skill only reaches your parsed record if its name exists as extractable text in a region the parser actually read. That rules out more of a typical resume than people expect: graphics, some headers and footers, and any sidebar whose text gets interleaved into the wrong lines. This post is a location-by-location audit of where a term survives, where it is at risk, and where it is simply gone.
Region 1: the body of the page — safe
Ordinary paragraph and list text, in a single column, flowing top to bottom, is the reliable case. It extracts in order, lands inside whatever section heading precedes it, and matches the vocabulary cleanly.
If you remember one rule, it is this: anything you need a machine to read belongs in the body text of a single-column layout. Everything below is a variation on the ways that condition gets broken.
Region 2: headers and footers — at risk
Word processors and page layout tools treat the header and footer as a separate region that repeats on every page. Extractors handle them inconsistently: some interleave the header text at the top of each page’s content, some collect it separately, and some skip it entirely.
That is fine for a page number. It is a real problem for the thing people most often put there — the contact block. An email address that lives only in a document header may not reach the record at all, and a candidate with no email address is a candidate nobody can contact.
Put your name, email, phone, and location in the first lines of the body. If you also want them repeated in a running footer for the benefit of someone holding a printout, that is harmless. Just don’t let the footer be the only copy.
Region 3: text boxes and shapes — at risk
A text box is not part of the document’s main text flow; it is an object placed on the page with its own content. Extraction may pull its text out at an unpredictable position — before the body, after it, or in the middle of an unrelated sentence — or omit it.
Text boxes are common in template-based designs, where a coloured band across the top of the page or a shaded “Profile” panel is implemented as a shape rather than as a paragraph. Visually you cannot tell. Mechanically it is a different kind of object.
The copy-paste test in how to test your resume like a parser would exposes this quickly: text boxes are usually the content that appears in a strange place or not at all.
Region 4: tables and multi-column sidebars — at risk
A table is real text, so it survives extraction — but its order frequently does not. Reading a two-column layout line by line across the page produces text where a skill name from the sidebar lands inside a sentence from the main column.
The skill may still match the vocabulary, because matching mostly ignores context. What breaks is everything that depends on adjacency: which section the skill was in, which job it belonged to, whether that date range attaches to that employer. You keep the term and lose the meaning. This is covered in detail in why your two-column resume breaks.
Single-cell tables used purely for alignment are a milder version of the same risk. If you are using a table only to right-align dates, a tab stop does the same job with nothing to shred.
Region 5: images and vector graphics — gone
Text drawn as a picture is not text. This includes:
- Skill bar charts, five-dot proficiency meters, radial “expertise” wheels.
- Word clouds.
- A logo that contains your name.
- Icons that substitute for labels — a small envelope glyph instead of the word
Email. - A resume exported as a single flattened image and wrapped in a PDF, which happens if the file went through a scanner or certain design tools.
All of it contributes exactly nothing to the record. The proficiency chart is the cruel case, because it is the part of the document the writer worked hardest on and it is the part that vanishes completely.
A whole-document image is worse than it sounds: the file parses to zero text, and the record ends up empty except for whatever a human later types in. Some pipelines run optical character recognition to recover text from images, which may rescue you — with recognition errors, and no reliable structure. Do not plan around it.
Region 6: document metadata and invisible text — do not
Some guides suggest putting keywords in the file’s metadata properties, or in white-on-white text, or in a two-point font behind a graphic. This site will not recommend it, and the reason is not squeamishness.
Extracted text is frequently what the recruiter reads in their own interface. Hidden blocks routinely surface there in plain view, stripped of the formatting that hid them, in front of exactly the person you were hiding them from. It reads as an attempt to deceive, because it is one, and it ends the application on the spot — decided by a human, not caught by a machine.
The premise of getting your resume parsed correctly is that the machine’s version and the human’s version say the same thing. Hidden text is the deliberate opposite.
A placement checklist
For each item you need a machine to have:
| Item | Put it |
|---|---|
| Name, email, phone, city | First lines of the body text, not the header |
| Skill names | A plain skills list plus the relevant experience bullets |
| Job title, employer, dates | Same or adjacent lines, in the body, one column |
| Section labels | Ordinary paragraph text using standard headings |
| Certification and licence names | Written out as text under a recognisable heading |
| Anything currently in a chart | Rewritten as words |
Two things that do not matter
To keep the fear proportionate: font choice and colour are irrelevant to extraction. A parser discards both. Choosing a different typeface does not make you more machine-readable, and neither does going grayscale. Use whatever a human finds legible.
Likewise, white space and margins are irrelevant, up to the point where they are implemented with tables or text boxes. A generous, airy single-column layout parses exactly as well as a cramped one.
The distinction that matters is not plain versus designed. It is text in flow versus text in an object.
Check it
Print your resume to plain text — select all, copy, paste — and then go looking for the six things in the checklist above. Anything you cannot find in the pasted version is not in your record either, no matter how prominent it looks on the page. Anything that appears in a bizarre position is in a text box or a table, and moving it into the body flow will fix it.