We exported a PDF from the builder and fed it back to the checker

A PDF is not a document with text in it. It is a set of instructions for drawing glyphs at coordinates, and the text layer is a courtesy the exporter may or may not have left behind. Every claim about what an applicant tracking system reads depends on that layer, and almost nobody checks it.

So we did the round trip: built a résumé, downloaded the PDF this product renders, and put that exact file back into the ATS résumé checker and the builder's own import. The numbers below were measured on 26 September 2026.

How the file was read

The text layer is pulled out in the browser by a PDF engine bundled with the page, before anything is sent anywhere. Where a page has no text layer at all, optical character recognition runs over the image instead, in the browser, before the review opens. Our export needed none: one page with a real text layer.

What came back

A complete recovery: both employers with their date ranges and all seven achievement sentences in the right order, the degree with its school, all nine skills. The file card reads 170 words where the pasted source has 162, and the difference is not a mistake: our renderer writes a few words into the text layer that are not printed on the page, a plain copy of the name first and the section word beside each block, so that parsers which rebuild a page from its geometry keep every line under the right heading.

The score, pasted and exported

100 × 30% + 100 × 30% + 71 × 40% = 88 for the file, and exactly the same for the text it was made from. That equality is the property worth having: the document you send is scored as the document you wrote.

What the round trip found on the way

It was not equal when we started on 26 September, and the reasons were ours. Imported back into the builder, our own PDF read nine roles instead of two: the section word written before each bullet for other parsers was being taken as a new section, so every bullet became a role of its own. The degree came back as two degrees, the upper-case name on the page became a headline, and the domain of the email address was filed as a portfolio link that nobody wrote. The checker, meanwhile, counted those extra words as readable content and scored the file one point above the text.

All of it was fixed the same day, in the reader and the scorer rather than by hiding anything: the section word inside its own section is read as a marker and nothing more, a line equal to the name is not a headline, a second email is never a website, and the markers are not counted as volume. A test that runs the browser's own reading of that PDF now guards each of them.

An earlier round trip, on 12 September

The first time this page measured the trip, the export scored three points below the pasted text. The renderer then drew its bullet marks as layout rather than as characters, so the text layer held none, and the achievement rule's verb list did not include rebuilt, wrote or mentored. Both were fixed on 13 September. The practical lesson survives the repair, for the reason every résumé guide already gives: start the line with the verb.

What this says about your own file

Run the file you actually send. A résumé that pastes at 400 words and uploads at 90 has a text layer problem, and no amount of rewriting will move the score until the export changes. The counts are printed for exactly this, as described in what the checker read, and the layouts most likely to survive the trip are the ones marked in what ATS-safe means on a design.

What a scanned page does to the same test

A PDF that was printed and photographed has no text layer at all. The reader falls back to optical character recognition in the browser, and the review then shows what recognition recovered. On a clean scan of this same export it recovered every line and lost the name, which is measured in rescuing a scanned résumé; on a creased phone photograph it loses more, and its errors land in exactly the places that hurt, which is proper nouns, job titles and figures.

Where a scan is the only copy of a résumé, the repair is to rebuild the document rather than to improve the scan. Import it, let the reader do what it can, correct the fields on the review, and download a fresh PDF whose text layer is real. The round trip above is then available as a check, and it takes a minute.

What we are not claiming

That our reading is an employer's reading. It is one parser, ours, and a different one will differ at the edges. What the round trip does establish is that a document downloaded from this product still contains its own text when it arrives, and reads back as the same résumé, which is the minimum a résumé builder owes you and is not universal. The other download, the Word file, never goes near a server at all: we measured that too.