how does pdf comparison work

This tool doesn't extract and diff text the way a code-comparison tool does. Instead, it renders each page of both documents to an image at the same scale, then compares those two images pixel by pixel. Any pixel that differs beyond a small tolerance gets flagged, and the flagged regions are highlighted in red on the output.

This approach has a real advantage: it catches visual changes a text-only diff would completely miss, like a moved image, an altered signature, a different stamp, or a formatting change that doesn't touch the actual text content. It also works identically on scanned documents, which have no extractable text at all — a text-based diff tool simply couldn't compare those.

The trade-off is that a pixel comparison is sensitive to rendering differences that aren't meaningful changes to a human reader — for instance, if one PDF was generated at a very slightly different internal scale or with different font rendering hints, you might see minor highlighted differences even where the actual content reads identically. In practice this is uncommon for two versions of the same document, but it's worth knowing the comparison is genuinely visual, not semantic.

Unchanged regions are shown faded rather than fully removed from view, specifically so you can see the surrounding context around a flagged change — a change highlighted with no visible surrounding text or layout would be much harder to interpret than one shown in its original position on the page.