Back to Blog

The First Line Read SKILLS | SUMMARY

November 18, 2026
Share:
The First Line Read SKILLS | SUMMARY

I drew a one-page, two-column resume for a made-up data analyst called Mara Quill. Skills down the left, summary and jobs down the right. Then I read it the way a plain copy reads a PDF, straight across the page, one line at a time.

The first line was "SKILLS | SUMMARY". The second was "- SQL | Data analyst who turns messy freight data into". Her first skill and the first half of her first sentence, glued together with a bar.

That's what a two-column resume can turn into when an application form asks you to paste it. The page I built for it is Resume to Plain Text. The file stays in the tab.


The box that wants your resume as text

A lot of job applications have one plain text box somewhere in them. Sometimes it's the whole resume step. Sometimes it sits next to the upload button as a backup, with a label like "paste your resume". The box takes characters and nothing else: no columns, no bold, no bullets that come from a symbol font.

So you open your PDF, select all, copy, paste. What lands in the box depends on two things you never chose. One is the order the PDF stored its text in. The other is what your PDF viewer does when it guesses lines from that text.

A one-column resume usually survives this. A two-column one often doesn't. The sidebar and the main column sit at the same heights on the page, so a reader that goes across the page, line by line, takes one line from each column and joins them.

The first line of the sample, read straight across, is "SKILLS | SUMMARY".

That's the whole problem in five words. The smaller messes come along with it.


A PDF has no lines, only pieces of text with positions

A PDF doesn't store "line 1, line 2". It stores runs of text, each with a position on the page, a font size and a width. A run can be a whole line, one word, or three letters of a word, depending on the program that made the file.

pdf.js, the library Firefox uses to show PDFs, hands those runs over in the order the file stores them. Rebuilding lines is your job. The page does it like this:

StepWhat it looks atWhat it decides
Rowseach run's height on the pageruns within half a font size of each other are one row
Spacesthe gap between two runs on a rowtouching runs join, a gap of a fifth of the font size gets a space
Wide gapsa gap of more than three times the font sizethe gap becomes " | ", so a date pushed to the right edge stays readable
Paragraphsthe space between two rowsmore than 1.8 times the font size leaves a blank line

The bar for wide gaps is a choice. A job title on the left and its dates on the right are one line on the page, and "Analyst, Acme Freight | 2021 - Present" reads better in a text box than the two jammed together.


Finding the gap between two columns

Once the rows exist, the page looks for a column gutter. It walks across the page one point at a time and counts how many rows have text crossing that spot. A spot that almost no row crosses is open. A stretch of open spots in a row is a candidate gutter.

A candidate only counts as a real column split if it passes all of these:

  • It's at least one and a half font sizes wide.
  • Its middle sits between 20 and 80 percent of the text's width, so a margin doesn't count.
  • No more than a fifth of the rows cross it. The name and contact line at the top usually do, and that's fine.
  • Each side has at least three rows of its own, and the right side has text on at least a quarter of all rows.
  • Each side is at least a fifth of the text's width.
  • Most rows on the right start at the same spot, within three points.

The last two are there for a one-column resume with a lot of structure. A narrow column of years down the left edge, with job lines next to it, is not a sidebar. Read as two columns, you'd get every year first and then every job, which is worse than the straight read. The width rule rejects it. Dates pushed to the right edge fail the other way: they end at the same spot but start wherever the date happens to start, so the alignment rule rejects them.

When a gutter passes, the page reads the left column top to bottom, then the right column, then carries on. A row that crosses the gutter, like the name, stays where it was and splits the page into blocks. The sample comes out as the name, the contact line, every skill, education and the certification, and then the summary and both jobs.


The bullets that show up as boxes

Most resumes use bullets, and a bullet is where a PDF gets odd. Some PDFs store a bullet as a normal round bullet character. Some store it as a character from a symbol font, a code that lives in a range Unicode keeps for private use. On the page it looks like a dot. In a text box it can come out as an empty square or a question mark, because the box has no idea what that code is supposed to be.

The page looks at the start of every line. A round bullet, a square, a diamond, an arrow, a check mark or one of the common symbol-font codes becomes the marker you picked:

PickA bullet becomesGood for
dash"- "most text boxes, and the default
star"* "boxes that render Markdown
dot"• "a box you know shows a real bullet

A dash or a star that was already there is left alone and doesn't count as a change. A bullet in the middle of a line, like "Python • SQL • Excel", becomes a bar, because those are separators in a skills list, not list items.

Other private-use characters, the ones that aren't bullets, are usually icons: a phone next to the number, an envelope next to the email, a pin next to the city. Those are removed. So are emoji. The contact line ends up as the words, which is all a text box can show anyway.


Quotes, dashes and the letters "fi"

Curly quotes, the long dash and the three-dot character are fine in most boxes and not fine in some. The page turns them into straight quotes, a hyphen and three dots by default. "2019–2021" becomes "2019-2021". You can untick Straight quotes and plain dashes to keep them.

Ligatures are a quieter one. Some fonts draw "fi", "fl" and "ffi" as a single glyph, and some PDFs store that glyph as one character. Paste "Certified" into a box and a search for "certified" can miss it, because the "fi" is one symbol. The page spells those out.

Then there are characters you can't see. Zero-width spaces and soft hyphens come along with some exports and sit inside words. A non-breaking space looks like a space and isn't one. Tabs show up from tables. The page removes the invisible ones and turns the odd spaces into plain spaces.

Whatever is still outside plain ASCII after that, an accented letter in a name, an arrow, a trademark sign, the page lists with the line it's on. It doesn't touch them. Most boxes take "José" just fine. A symbol might come out as a question mark, and that list is where you'd look before you paste.


Lines that wrapped on the page

A PDF breaks a long sentence wherever the line ran out. Pasted as is, a bullet that wrapped turns into two lines, and the second one looks like a new bullet with no marker.

The page joins a line onto the one above it when it starts with a lowercase letter, as long as it isn't a bullet itself and the line above isn't a section heading. In the sample, "Data analyst who turns messy freight data into" and "weekly reports people read." become one line again, and so does a bullet that wrapped onto "now reads every Monday". A word broken with a hyphen at the end of a line is closed up without a space.

A line that starts with a capital stays on its own line. That's on purpose. "Acme Freight" followed by "Senior Analyst" is two lines, and gluing those would be wrong far more often than right. If you don't want any joining at all, untick Join wrapped lines.


Headings people type with spaces

Designed resumes love "S K I L L S", letters spread out with spaces. In a text box it's eleven characters of nothing a search would match. The page closes a line made only of single letters and spaces into one word, and treats a double space as the gap between words, so "W O R K H I S T O R Y" becomes "WORK HISTORY".

Known section headings, Summary, Experience, Education, Skills, Projects, Certifications and their common variants, get a blank line above them so the sections stay apart in a box with no bold. With Section headings in capitals ticked, they're also written in capitals. A sentence that starts with one of those words, like "Skills I picked up at work", isn't a heading and isn't touched.


Word files keep bullets somewhere else

A .docx is a different animal. Word doesn't store a list bullet as a character at all. It stores the paragraph as a list item and draws the bullet from numbering settings. That's why a lot of text exports from Word come out with no bullets: there was never a bullet character to export.

The page reads a Word file with mammoth, which turns it into simple HTML and keeps the lists as lists. Then a small reader walks that HTML and writes each list item with a bullet, numbers for numbered lists, and two spaces of indent for a list inside a list. Tables are read row by row, with a bar between cells. After that, the same cleaning runs as for a PDF.

The old .doc format and RTF aren't read. Open those in Word, Google Docs or LibreOffice, save as .docx or PDF, and drop that.


When there's nothing to read

Some PDFs are a photo of a page. A scanner or a phone camera made it, and there's no text layer, just pixels. pdf.js finds almost no text, so the page says this PDF has no text in it, only a picture of text, and stops. It doesn't guess the words from the picture.

The fix is upstream: export the PDF again from the document you wrote it in, or paste the text. The limits are 15 MB for a file, 30 pages for a PDF and 200,000 characters for a paste. A paste over the limit isn't cut. It stays in the box, the page says how many characters over it is, and the plain text waits until it fits.

Each pdf.js step has a 15 second limit, and the document is shut down when the read ends, so a file that stalls the reader gets an honest message instead of a spinner that never stops.


What it is not

It isn't a copy of any tracking system's parser. Every one of those has its own, and I'm not going to claim the page reproduces any of them. It does show where a layout can trip one up, like two columns read straight across or bullets that are symbol-font codes.

It doesn't score or rewrite the resume. The words that come out are the words that went in, in a better order, with cleaner punctuation. And it isn't magic on a heavily designed PDF. A resume built from text boxes scattered around a page can store runs in an order no gutter search will untangle. Untick Read two columns separately and you'll see the straight-across read, so you can compare.


Where the file goes

Nowhere. The PDF or the Word file is opened in the tab, and the text is built there. You can load the page, go offline and still convert a file.

Mixpanel and Google Analytics load with Softery.io, the same as every page on the site. They see that the tool was opened, which settings changed, and size bands like "under 1 MB". They don't get the file or the text. The five settings are saved in the browser, and the share link carries those five settings and nothing else.


What to do with your resume

  1. Open Resume to Plain Text and drop your PDF or .docx, or press Try the sample PDF first to see Mara Quill's columns.
  2. Look at the left box. That's the PDF in reading order. If your resume has a sidebar, check that the "What changed" sheet says the page was read as two columns.
  3. Read the right box from the top. The skills should come before the jobs, not woven into them.
  4. Pick a bullet. Dash is the safe choice for a plain box.
  5. Check the list of characters outside ASCII. Fix anything that matters in your original file.
  6. Copy plain text, or download the .txt, and paste it into the form.

If the box has a character limit, the Word and Character Counter tells you how far over you are. If you have a job posting in mind, the ATS Resume Scanner shows which of its words your resume is missing.


Last updated: November 18, 2026 | Reading time: 10 minutes

Written by Softery.io, after reading a made-up analyst's resume one wrong line at a time.