"Compress a PDF without losing quality" is two requests, and whether they can both be satisfied depends entirely on why your particular file is large. Some PDFs shrink by ninety per cent with nothing whatsoever discarded. Others are already as small as their contents allow, and every megabyte you remove comes out of the pages.

So the useful first step is not choosing a compressor. It is finding out which kind of file you have.

Find out what is actually taking up the space

A PDF is a container, and four things commonly fill it.

Images. Scans, photographs, screenshots, logos. On most large PDFs this is essentially the whole answer, and it is not close — a single scanned page at 600 dpi in colour can be larger than a hundred pages of text.

Embedded fonts. Each font carries its glyph data. Fully embedded fonts, and particularly fully embedded CJK fonts, run to megabytes; a subset containing only the characters actually used is a fraction of that.

Vector content. Maps, CAD exports and chart-heavy reports can contain hundreds of thousands of drawing operations. Technically text, practically enormous, and not much can be done about it without changing the drawing.

Invisible baggage. Revision history from incremental saves, unused objects, embedded file attachments, thumbnails, metadata, and — in a PDF exported from a design tool — occasionally the entire original artwork on a hidden layer.

There is a quick diagnostic that needs no software. Divide the file size by the page count. Text pages run a few tens of kilobytes each. Anything over about a megabyte per page is an image-dominated document, and images are where you should look. PDF Page Counter gives you page counts across a whole folder in the browser if you are triaging a stack rather than one file.

Lossless and lossy are different operations

Most compression tools offer one slider and mix these together, which is why results are so unpredictable. They are worth separating.

Lossless. Nothing on the page changes; the file just stores it more efficiently. Re-compressing the object streams, dropping unused objects, subsetting fonts, removing revision history, deduplicating a logo that was embedded once per page. The output renders identically, pixel for pixel, and the text stays selectable. This genuinely is compression without loss of quality.

Lossy. Image data is thrown away and cannot be recovered — downsampling to a lower resolution, re-encoding as JPEG at a lower setting, converting colour to greyscale. Often entirely reasonable, sometimes invisible, never reversible.

If "without losing quality" is a hard requirement, you want the first list only, and you should expect modest savings on a text document and dramatic savings on one that was assembled badly. If the file is image-heavy, lossless alone will not get you under an email limit, and the honest trade is to choose deliberately which quality to lose rather than letting a slider choose for you.

How far to downsample, if you are going to

When the images have to get smaller, resolution is the lever that matters, and there are well-established thresholds worth knowing:

300 dpi   print quality; the right target for scanned text
200 dpi   visibly softer on small print, still legible
150 dpi   fine for on-screen reading, poor for print
 96 dpi   screen only; small print becomes a guess

Downsampling a 600 dpi scan to 300 dpi typically removes three quarters of the file and nothing a reader will notice, because 600 dpi captured no additional detail from a printed page in the first place. Going below 300 on a document whose figures anyone might need to read later is where regret starts, especially if the document might ever go through character recognition — converting scanned PDFs to Excel explains why resolution and colour depth set the ceiling for that.

Two related choices. Colour to greyscale halves or thirds the size of a scan and loses nothing on a black-ink document, but check for a coloured stamp or signature first. JPEG quality around 70–80 is usually indistinguishable from the original on a photograph and visibly ugly on text, which is the reason a single slider serves mixed documents badly.

Ways to compress a PDF

Re-export from the source. Much the best option when it is available. If the PDF came out of Word, InDesign or a reporting tool, export it again with a smaller image setting. The source has the original images at full quality, so the export makes one good decision instead of a compressor making a second-hand one.

Print to PDF. On both macOS and Windows, printing a PDF to a PDF writer flattens it and frequently produces a much smaller file. It is lossy and it is blunt — it re-renders everything and may drop bookmarks, links and form fields — but it is built in and it takes ten seconds.

macOS Preview's Quartz filter. Export, then choose "Reduce File Size". Fast, aggressive, and on text-heavy documents sometimes increases the size, which is itself a useful signal: it means the file had nothing easy left to remove.

Adobe Acrobat. Has both a straightforward "Reduce File Size" and a PDF Optimizer that reports what each category is consuming and lets you act on one at a time. If you do this regularly, the report is the valuable part.

Ghostscript. The command-line option, with presets from /screen to /prepress. Predictable and scriptable, which matters if this is a recurring job.

Online compressors. Convenient, and worth one moment's thought about what you are handing over: the file goes to a machine you do not control, and how long it stays there is that site's policy rather than a property of the process. For a marketing brochure, who cares. For signed contracts, bank statements or anything with personal data in it, prefer something that runs locally. The same reasoning is why merging PDFs and splitting them are done in the browser on this site rather than on a server.

Cheap wins before you compress anything

Several of these remove more than a compressor will, and none of them touches image quality.

Remove pages you do not need. The blank back pages, the duplicated cover, the twelve-page terms appendix nobody reads. A 40% page reduction is a 40% size reduction and it is entirely lossless. Split PDF extracts the ranges you want to keep.

Check for a duplicated attachment. A PDF can carry embedded files, and a document that has a copy of its own spreadsheet attached is twice the size for no visible reason.

Re-save to flatten revision history. Every incremental save appends to the file rather than rewriting it, so a PDF edited twenty times carries twenty versions. A plain re-save or "save as" collapses that, losing nothing visible.

Flatten filled form fields once the form is final. Interactive form data is stored separately from the appearance of the page.

Send a link instead. Worth saying out loud, because it is often the real answer to "this will not fit in an email".

What compression will not fix

A scan is a scan. Compressing it smaller does not make it searchable, and compressing it harder makes recognition worse. If the goal is getting the figures out rather than emailing the file, compression is not on the path at all.

Vector-heavy drawings. A map with 200,000 line segments is large because of what it depicts. Simplifying the drawing is a design decision, not a compression setting.

Already-optimised files. Many PDFs produced by accounting and banking systems are already minimal. If a compressor returns a file 2% smaller, that is the correct answer, and the next tool promising 90% is either lossy or lying.

Common questions

Does compressing a PDF make the text blurry? Lossless compression does not touch the text at all. Lossy compression does not usually touch vector text either — it is the images that degrade. Blurry text after compression almost always means the page was an image to begin with, i.e. a scan.

Will compression break the text layer? Lossless, no. Re-rendering approaches such as print-to-PDF generally preserve selectable text but may discard links, bookmarks and form fields. If text selection matters, check the output before discarding the original — the quickest check is Extract Text from PDF, which tells you per page whether a text layer is still there.

Is there a point in compressing a stack before merging it? Yes, and the order matters: compress the parts, then merge. Merging copies pages as they are, so whatever you saved is carried into the result.

How small does a PDF need to be? Email attachment limits are commonly 10 MB or 25 MB, and the sending and receiving systems may differ. If you are close to a limit, go well under it rather than just under it.

Can I compress a password-protected PDF? Not without the password. Remove the protection first in whatever application applied it.

If the real problem is the contents

One thing worth checking before you spend an afternoon on this. If you are compressing a 200-page PDF of supplier invoices so that it can be emailed to someone who is going to read the totals off it and type them into a spreadsheet, the size is not the problem.

That job — one row per document, under column headings you chose — is what ExtractToExcel does, and the spreadsheet is a few kilobytes. Converting multiple PDFs into one spreadsheet describes the workflow, and the free tools page has the browser-side PDF utilities that are built here.