When compressing a PDF destroys selectable text

People search “compress PDF” when a file is too big to email. They usually want the same document, smaller. Our PDF Compress tool often delivers a smaller file. It also often delivers a different kind of PDF: pages that are pictures of pages. This guide is when that is fine, and when it is a mistake you cannot undo from the download.

What the compressor actually does

Each page is rendered with PDF.js at a chosen scale, drawn to a canvas, encoded as JPEG at a quality preset (High, Balanced, Small, Tiny), then placed into a new document with pdf-lib. The output is usually an image-only PDF. Text is not a text object anymore. You cannot select a paragraph. Search inside the file fails. Screen readers get little or nothing. Form fields and annotations are gone.

That pipeline is a known tradeoff, not a bug we forgot to mention. Rasterizing is how you shrink a 40 MB scan of a magazine. It is how you do not shrink a 200 KB text contract that was already distilled. If the source is already efficient vectors and fonts, JPEG pages can come out larger than the original.

Soft limits exist (~120 MB, ~200 pages) because rendering every page into bitmaps is RAM-heavy. They are not a pricing tier.

Use compress when

• The PDF is a scan or a photo-heavy export (phone scans, lookbooks, slide decks with full-bleed images).
• The recipient only needs to read or print it, not copy clauses or fill fields.
• You already tried “save as reduced size” in a desktop editor and still need a web-safe attachment.

Do not use this compressor when

• You need to copy text, quote a contract, or keep accessibility tags.
• The file has fillable forms, signatures, or layers you still have to edit.
• It is already a small born-digital PDF and you are compressing out of habit.

For those jobs, Merge & Split copies pages with pdf-lib and usually keeps text. That is the organizer (reorder, rotate, extract). It is not a compressor. Flipbook is a private preview, not an optimizer.

How to choose a preset

High is for pages you still want to pinch-zoom. Tiny is for “this must fit in an email cap.” Balanced is the default because it is the least surprising. If Tiny makes type unreadable, go back up a step. There is no universal kilobyte target: a 12-page photo PDF and a 12-page text PDF do not share a size curve after rasterizing.

Always open the download and try to select a sentence. If you cannot, and you needed to, you used the wrong tool — keep the original.