How to Compress a PDF Without Losing Quality
Short answer
Truly lossless compression — removing duplicate objects, subsetting fonts, discarding edit history and rebuilding the cross-reference table — changes nothing you can see and typically saves 5–30%. Anything beyond that involves re-encoding images, which is lossy by definition. The honest question is not whether you lose quality but whether you will notice, and at 150 DPI on a document you will not.
On this page
“Without losing quality” is doing a lot of work in most PDF compression advice, and it is usually not true. Almost every tool that promises it is downsampling your images and hoping you do not look closely.
The honest version has two parts. A genuinely lossless set of operations exists, and it is worth doing first. Beyond that, compression trades quality for size — and the useful question becomes which losses are invisible rather than whether any loss occurred.
What is genuinely lossless
These operations cannot change a single visible pixel, because they only touch structure:
| Operation | Typical saving | What it does |
|---|---|---|
| Remove duplicate objects | 2–15% | The same image or font stored several times |
| Subset embedded fonts | 1–10% | Keeps only the characters actually used |
| Discard edit history | 0–40% | Superseded content from incremental saves |
| Rebuild the xref table | 1–5% | Object streams instead of a plain table |
| Strip metadata and thumbnails | 0–3% | Cached previews, XMP packets |
| Remove unused objects | 0–10% | Orphans left by editing software |
Together these typically recover 5–30% on a text-based PDF, occasionally much more on a document that has been edited repeatedly.
That last case deserves attention. Most PDF editors save by appending changes and marking the old content superseded rather than removing it — so a document through ten rounds of edits can contain ten copies of itself. A lossless rebuild discards all of it, which both shrinks the file and removes text you thought you had deleted.
Our PDF compressor does exactly this set of operations under Lossless rebuild, in your browser, with no upload.
The limitation is important: on a scanned PDF this saves almost nothing. Scans have no structural redundancy — each page is one photograph. If lossless rebuild disappoints you, that is the diagnosis, and compressing a scanned PDF is the guide you actually need.
What is lossy, stated plainly
Everything else. The table that matters:
| Change | Lossless? | You notice when |
|---|---|---|
| Downsample images to 300 DPI | No | Professional printing only |
| Downsample to 150 DPI | No | Zooming past ~200%, or fine print |
| Downsample to 96 DPI | No | Immediately on any modern screen |
| Colour scans → greyscale | No | Only if colour carried meaning |
| JPEG quality 85% | No | Essentially never on a document |
| JPEG quality 60% | No | Visible haloes around text |
| Flatten form fields | No | You can no longer fill the form |
| Flatten annotations | No | Comments become uneditable marks |
Note the pattern: the first step down is nearly free and each subsequent one costs more for less. Going from 600 DPI to 150 removes 94% of the pixel data; going from 150 to 96 removes another 59% of what is left, which is a much smaller absolute saving for a much more visible cost.
That asymmetry is the whole strategy. Take the first big step and stop.
The best method is not compression at all
If you still have the original Word, Pages, PowerPoint or InDesign file, re-exporting beats every compression tool — and it is genuinely lossless in the sense that matters, because you never create the excess data.
In Word: File → Save as Adobe PDF, or Export → Create PDF, and choose Minimum size or Standard (publishing online) rather than Best for printing. That single choice often halves the file.
While you are there, fix the images in the document itself: select one, then Picture Format → Compress Pictures, choose Web (150 ppi), and tick Delete cropped areas of pictures. Cropped-away pixels are stored in the file until you do that explicitly.
The reason this beats post-hoc compression is generation loss. Compressing an exported PDF re-encodes images that were already encoded once; exporting correctly encodes them once, from the original.
A practical sequence
- Run a lossless rebuild first. It costs nothing and you learn something — if it saves 25%, your PDF had structural bloat; if it saves 2%, your PDF is images and you know where to go next.
- If that is enough, stop. You have lost nothing.
- If not, check whether you have the source document. Re-export at screen resolution if you do.
- If not, downsample to 150 DPI at 75–80% quality. Check the result at 100% zoom on the page with the smallest text.
- Only if still too large, try greyscale, then 96 DPI, checking readability at each step.
Keep the original throughout. Compression is not reversible, and the one thing you cannot recover from is compressing over your only copy.
When lossless is not negotiable
Some documents should not be lossily compressed at all:
- Legal and evidential documents. Artefacts can be mistaken for content, and altered pixels can matter. Signatures, stamps and handwriting are particularly vulnerable.
- Medical and scientific imaging. Compression artefacts can look like findings.
- Documents for professional printing. A print shop needs 300 DPI. Send the large file.
- Archival copies. Keep one untouched original; compress copies.
- Anything you will edit again. Compress only at the final export.
For these, the lossless rebuild is your entire budget. If that is not enough, share a link rather than degrading the document.
Checking your work
Do not trust the file size alone — look at the result:
- Open the compressed PDF at 100% zoom, not fit-to-page.
- Go to the page with the smallest text — footnotes, disclaimers, form labels.
- Compare side by side with the original.
- If it is going to be printed, print one page.
If it looks fine at 100%, it is fine. Zooming to 400% and complaining about softness is judging it at a magnification nobody will use.
What does not work
- Zipping a PDF. Internally compressed already; a zip gains a few percent and makes the file harder to open.
- Running several compressors in sequence. Generation loss compounds and the savings do not.
- “Lossless” online compressors that shrink a scan by 90%. That is arithmetically impossible. They are downsampling and calling it lossless.
- Uploading confidential documents anywhere to save a few megabytes. Contracts, payslips and medical letters belong on your own machine — which is why our compressor runs in the browser.