Your CV has two readers and only one forgives you

Your CV has two readers, and they are not equally forgiving.

The first is a parser. It opens your file, tries to find structure, and extracts what it can into fields. It does not read. It matches. Whatever it fails to extract does not exist for the purposes of the next step, which might be a keyword filter, a ranking score, or a recruiter's search.

The second is a person, and they get your CV after the parser has already had its way with it. They spend something between six and ten seconds deciding whether to keep reading.

Most engineers optimise for the second reader and forget the first entirely. Two columns, because it looks designed. A skills matrix in a table, because it is compact. Icons instead of labels. Contact details inside a graphic. A PDF exported from a design tool where the text layer is decorative.

Every one of those is defensible as design. Every one costs you information at the parsing stage, invisibly and consistently.

I did not learn this as a candidate. I learned it on the hiring side, which is a much worse way to learn it. I marked a CV as thin because there was almost nothing in it. Later, for an unrelated reason, I saw the original file. Four roles, real detail, a strong project at a company I knew. Two-column layout, and the system had read down the left column then down the right, interleaving job titles with the skills list into something incoherent.

I scored a strong candidate as thin, in about eight seconds, and I was completely confident.

What this costs, and it is measured in months.

If your CV loses forty percent of its content at parsing, every application you send is a weakened version of you, and the weakening is invisible. You get no feedback saying your layout broke. You get silence, which you read as a market problem, so you apply harder with the same file.

Six months of applications with an unreadable document is not six months of market feedback. It is the same experiment run repeatedly with nothing changed.

The fix is two days. Run your current file through a parser and read the extraction, not the score. Then rebuild: single column, standard headings, real selectable text, no content inside images. Re-run it and confirm the losses are gone.