There are two completely different things called a PDF
This is the whole reason most PDF compressors are worse than they look, and it is worth two minutes before you press anything.
A scan is a photograph of paper. Each page is one large image, often captured at 300 dots per inch, and a ten page scan can easily be 25MB. There is no text in it at all, only a picture of text.
A text PDF came out of Word, a drawing package or an invoicing system. It stores the actual letters and the actual lines, which takes almost no space. A forty page report might be 400KB.
Compression that works brilliantly on the first is destructive vandalism on the second. So this looks at your file before offering anything, and tells you which one you have.
What most tools do instead
They redraw every page as a picture at a lower resolution, whatever you gave them. On a scan that is exactly right. On a text document it makes the file bigger, and it destroys the text layer permanently: the words can never be searched, selected or copied again, and no tool can put them back.
You usually find out weeks later when you try to search a contract for a clause and get nothing.
How this tool works
Before anything is offered, the document is looked at. Ten pages are sampled for how much real text they carry, and that is weighed against the file size per page. Those two numbers separate a scan, which is photographs of paper, from a text document straight out of Word, and they need completely different treatment.
Redrawing renders each page to a canvas at your chosen resolution and re-encodes it as JPEG. Enormously effective on a scan, and it permanently destroys the text layer of a text document, which is why the tool says which kind it found before you choose.
The lossless pass rewrites the file with object streams switched on and clears the metadata, keeping every page exactly as it was. Modest gains, sometimes none, and the only honest option for a document that was already efficient.
"Why won't my PDF compress?"
Because it is already good. This is the most common complaint about every PDF compressor and the answer is nearly always that there is nothing to take out.
If your 800KB document refuses to shrink, that is the tool being honest. Squeezing it further means throwing away the text, which is available here and clearly labelled as what it is. When a tidy-up finds nothing, the small saving is explained rather than dressed up, and if the file actually came out larger the download button is not offered at all, because handing you a worse copy of your own document would be a strange thing to do.
The two things this offers, and when each is right
Tidy up rewrites the file structure more efficiently and strips the metadata. Every page stays exactly as it was, so nothing can be damaged and text stays text. Expect a few per cent, sometimes nothing at all. That is not the tool failing, it is the file already being well made, and it is more useful to be told that than to be handed a big number achieved by damage you find out about later.
Redraw the pages draws every page again as a picture at a resolution you choose. It makes a scan small, and on a 300 dpi scan it routinely takes off most of the file size. It is also one way: the text layer goes, and the resolution you pick is the resolution you have from then on.
Neither can make the document worse without saying so. If the result comes out larger than what you put in, which is the normal outcome of redrawing a text document, the save button is not offered at all.
Picking a resolution
150 dpi is the usual answer. It reads comfortably on screen, prints acceptably, and is about where a 300 dpi scan stops looking obviously degraded. Drop to 96 only for something nobody will print. Go to 300 only if the original was higher than that, because redrawing at a higher resolution than the scanner used adds file size and no detail whatsoever.
Try greyscale on any scan of black text. The colour in those files is paper tint and scanner noise, and throwing it away often takes another third off.
Then open the result and read the smallest print on it. Resolution is a judgement about how much detail you needed, and only you can make it. Your original is untouched, so if it went too far, run it again higher.
The metadata is worth having gone on its own
Author, the software that created it, creation and modification timestamps, and often the original file path. Those fields travel with a document and routinely say more about where it came from than anyone meant them to.
If you are sending a quote, a drawing or a report outside your own business, clearing them is worth doing whether the file gets smaller or not.
What redrawing costs you
Worth being blunt about, because no other free compressor will be. Redrawing does not compress your text. It replaces your text with a photograph of your text.
Afterwards nobody can search the document for a clause, select a paragraph to quote, copy a reference number out of it, or have it read aloud by a screen reader. If it goes into a document management system, that system can no longer index a single word of it. None of that comes back, by any means, ever.
On a scan you lose none of that, because there was never any text in it to lose. A scan is already a photograph and you are only choosing how large a photograph. On anything that came out of Word or a drawing package you are destroying the most useful property the document had, and usually making the file bigger while you do it.
So this looks first, says which kind you have, marks the option that suits it, and puts the character count in front of you before it will redraw a document that has real text in it.
Keep the original
Good habit regardless. For a signed contract, a certificate or anything that might one day be evidence, keep the full file and send the tidied one.
No size limit, because nothing is paying for a server
The analysis and the compression both happen on this page using your own machine. No upload, no size limit, no watermark, no account and nothing logged.
Common questions
Why will my PDF not compress?
Almost certainly because it is already efficient. A PDF made by Word or a drawing package holds text as text, which is tiny. There is nothing to squeeze. Files that compress dramatically are scans, where each page is a large photograph.
Does this lose any quality?
That depends which of the two you choose, and the tool marks which one it recommends for your file. Tidy up loses nothing at all: the file is rewritten more efficiently and the metadata removed, every page stays exactly as it was, and the saving is usually a few per cent. Redraw the pages does lose quality, on purpose, because that is the only way a scan gets substantially smaller.
How do I get a scanned PDF under an email limit?
Redraw the pages at 150 dpi, and tick greyscale if it is a scan of black text. That routinely takes a 25MB scan to under 5MB. Most mail servers stop somewhere between 10MB and 25MB, and encoding adds roughly a third on top, so aim well under whatever limit you were given.
Will I still be able to search the text afterwards?
After a tidy up, yes: every page is untouched, so the text stays selectable and searchable. After redrawing, no. Redrawing turns each page into a picture, so the words become part of an image and can never be searched, selected, copied or read aloud again. No tool can put them back. That is why this checks whether your document has a text layer and puts the character count in front of you first.
How do I know which one to choose?
The tool marks one of them as best for your file, having looked at it first. Little text and large pages means a scan, so redrawing is right. Real text means it is already efficient, so tidying up is right. You can always pick the other one, and if the result comes out bigger than the original the save button is not offered at all.
Is my file uploaded to compress it?
No. Both the analysis and the compression happen on this page using your own machine, which you can confirm in your browser network tab. No size limit, no watermark and no account.