What gets compressed losslessly
- Embedded fonts that aren't subset (we keep only the glyphs actually used)
- Document metadata you don't need (Author, Title, Comments history)
- Thumbnails (PDF readers regenerate these anyway)
- Duplicate XObjects (the same logo embedded 50 times becomes one shared object)
- Unused color profiles
- Form field defaults that aren't visible
What gets compressed lossily
Mostly images. Photos and scanned pages are downsampled — typically from 300+ DPI to 150 DPI for screen-quality output. JPEG quality is also lowered slightly (e.g., from 95 to 75 on the 0–100 scale). Both are visible if pushed too aggressively, but at default settings the difference is hard to spot at normal zoom.
What never gets touched
- Text: stays as vector text, crisp at any zoom
- Vector graphics (logos, line art, shapes): not rasterized
- Form fields: preserved
- Annotations and comments: preserved
- Bookmarks and links: preserved
- Digital signatures: preserved (compression after signing breaks the signature)
Realistic size reductions
- Image-heavy reports / presentations / scanned documents: 50–80% smaller
- Mixed text-and-image: 20–40% smaller
- Pure text PDFs (Word doc exported as PDF): often already optimized — 5–15% reduction is the most you'll get
When to skip compression
- PDFs going to commercial print (offset, high-DPI digital): always send the original
- Documents under 1 MB: usually not worth the effort
- Already-compressed PDFs: re-compressing produces little benefit and can introduce visible artifacts
- PDFs with digital signatures: compressing breaks the signature