February 3, 2026 · 3 min read
How PDF Compression Actually Works (and Why Some Files Won't Shrink)
What's really inside a bloated PDF, what compressors do to make it smaller, and how to choose settings without wrecking your document.
You've seen the mystery: two PDFs, both twenty pages, one is 300 KB and the other is 45 MB. Understanding why is the key to compressing files intelligently instead of just mashing the "compress" button and hoping.
What makes a PDF big
A PDF page can contain three kinds of content:
- Text and vector graphics — instructions like "draw the word 'Invoice' in Helvetica at this position." These are tiny. A whole page of text costs a few kilobytes.
- Embedded fonts — the actual letterforms, usually tens to hundreds of kilobytes per font, shared across all pages.
- Images — grids of pixels. These are where the megabytes live.
That 45 MB PDF? It's almost certainly a scan: every page is one giant photograph, often stored at 300+ DPI in full colour, even if the original paper was black text on white. Twenty pages × 2 MB per photographed page = a file your email server rejects.
What compression actually does
Honest PDF compression for image-heavy files does two things:
- Reduces resolution. A page scanned at 300 DPI contains four times as many pixels as the same page at 150 DPI. For on-screen reading, 120–150 DPI is plenty.
- Re-encodes with stronger JPEG compression. JPEG quality settings between 60–80% look nearly identical to the original for documents while using a fraction of the bytes.
Our Compress PDF tool does exactly this, with presets that map to real situations: Light (150 DPI) when fine detail matters, Recommended (120 DPI) for email, Extreme (96 DPI) when only the size limit matters. There's also a grayscale switch — colour data in a scan of a black-and-white document is pure waste, and dropping it saves another meaningful chunk. (If grayscale is your only goal, the dedicated Grayscale PDF tool gives you finer control.)
The trade-off most tools hide
Rebuilding pages as optimized images has a consequence: text in the output is no longer selectable. Most compression services do this and don't mention it. We mention it in the tool itself, because the trade-off is usually fine — the files that need compressing are scans, and scans never had selectable text to begin with — but you should know it's happening.
If you compress a scanned document and then need searchable text, run the result through an OCR tool, which adds an invisible recognized-text layer back on top.
Why some PDFs barely shrink
If you feed a compressor a born-digital PDF — one exported from Word with real text and no big images — not much happens. The text instructions were already tiny; there's no 2 MB photograph to shrink. A 400 KB text PDF might compress to 350 KB, and converting its crisp vector text to images could even make it bigger and uglier.
Rule of thumb:
| Your PDF | Expect |
|---|---|
| Scanned paper, phone "scans" | 60–90% smaller |
| Slides with photos | 40–70% smaller |
| Text exported from Word/Docs | Little change — skip compression |
A checklist before you compress
- Check what kind of PDF you have. Try selecting text. Can't? It's a scan — compression will work great.
- Pick the weakest setting that fits your limit. If the file needs to be under 10 MB and Recommended gets you to 4 MB, don't use Extreme.
- Keep the original. Any honest compressor (ours included) writes a new file and never touches your source — but it bears repeating: compression is lossy, archive the original.
