Why does my resume look wrong or get mangled when I upload it?

Because the software never sees your layout — it extracts a single stream of text from the file, and columns, tables, text boxes, and header regions can come out of that stream in the wrong order. Your PDF still looks perfect when you open it, but the version the system stored is a different document, which is why the damage is invisible from your side.

What extraction actually does to your file

A PDF is not a document in the way a word processor file is. It is a set of instructions for painting glyphs at coordinates on a page. There is no stored notion of "this is a column", "this is a bullet list", or "this line belongs to that heading" — those are things your eye assembles from position.

An extractor has to reconstruct all of that from the coordinates, and reconstruction is where it goes wrong. Two columns get read across instead of down. A bullet glyph from a decorative font is not recognised as a bullet. Lines that were visually separate get merged, or a single line gets split.

What we measured in our own parser

We took one resume, rendered it five ways, and ran each through our real extraction pipeline. It is one document in five renderings, testing our parser and nobody else's, and the full write-up with limitations is on the study page.

Three findings are worth carrying away:

Nothing was ever lost. The word count was 159 in every single run, before and after our fixes. A parse can be badly wrong while every word is still present — which is exactly why re-reading your own resume cannot catch this.
Structure was destroyed instead. Under the extractor we shipped before 10 August 2026, a 25-line resume collapsed to a single line, bullet detection found 0 of 9, and the score fell from 98/A to 67/C. Same words, unusable structure.
A perfect score did not mean a clean parse. The two-column rendering scores 98/A today, and its stored text still interleaves the sidebar into the body — a degree welded onto a job's start date. The scorer was measuring text that had already been scrambled.

The Symbol-font bullet trap

One specific defect is worth naming because it is invisible and common. Word's default bullet in some list styles is drawn from the Symbol font, and when that PDF is extracted the character arrives as U+F0B7 — a Private Use Area codepoint — rather than the bullet character "•".

To a parser that is not looking for it, that is not a bullet. It is an unknown glyph, and a bullet list becomes a run of unlabelled text. Our parser got this wrong until we fixed it, which is how we know to look for it.

How to see what your file extracts to

Select all the text in your resume PDF, copy it, and paste it into a plain text editor. That approximation takes about fifteen seconds and catches all three of the failure modes above: interleaved columns, merged lines, and mangled bullets.

If you would rather have it checked and scored, the free checker does the extraction in your browser and shows you the text it read — not only the number it produced.

What to change if it comes out wrong

The fixes are structural, and none of them require an ugly resume:

Move the body to a single column. Keep the sidebar for a header block if you like the look, but not for content that has to stay in order.
Take your work history out of tables and text boxes and set it as ordinary paragraphs.
Put dates on the same line as the role they belong to, not in a separate column.
Use a standard bullet character rather than a decorative one from a symbol font.
Keep anything that matters out of the page header and footer region.
Re-export and run the copy-paste test again. It either reads correctly now or it does not, and you can see which.

What we cannot tell you

We tested our own parser. Workday, Greenhouse, Lever, Taleo, and every employer's internal system are closed — they do not publish their parsers and we have no access to them, so we cannot say whether any specific one has the same weaknesses.

What the study does establish is that the class of failure is real and mechanical: a document that looks perfect on screen can extract into scrambled text. That is worth checking regardless of which system is on the other end, and it is the part you control.

Frequently asked questions

My resume looks fine when I open it. Is it still a problem?

Possibly, and that is the difficulty. In every run of our study the word count was identical — 159 words — while the structure was destroyed. Because nothing is missing, proofreading your own file cannot detect it. The copy-and-paste test can, in about fifteen seconds.

Are two-column resumes safe?

They are a risk you cannot see. In our study the two-column rendering scored 98 out of 100 while its stored text interleaved the sidebar into the work history. We cannot tell you what any specific employer's system does with columns, because none of them publish that — but a single-column body removes the variable.

Should I send a PDF or a Word document?

Follow the posting's instructions first; if it names a format, that answer beats any general rule. In our own testing both extracted acceptably once the known defects were fixed, with one caveat: Word's Symbol-font bullets arrive as Private Use Area codepoints rather than bullet characters, which is a real trap in .docx-derived files.

Why do my bullet points disappear?

Usually because the bullet glyph came from a decorative or symbol font and was not recognised as a bullet marker. Our own parser found 0 of 9 bullets in a resume that visibly had nine, which collapsed the bullet-impact sub-score to a floor of 30. Switching to a standard bullet character removes the problem.

Does this happen at Workday or Greenhouse?

We do not know, and we will not claim otherwise. Those systems are closed and publish nothing about their parsers. Our study measures our own pipeline; what it demonstrates is that the failure mode is mechanical and real, not that any particular employer suffers from it.