Why your PDF is 40 MB

A four-page document should not be forty megabytes, and when it is, one of four things is almost always responsible. Which one decides what will fix it — and whether anything on this site is the right tool at all.

Most answers to this question are an upload button. This one is a diagnosis, including the cases where the honest advice is to use a program we have nothing to do with.

What a PDF actually stores

A PDF is not a picture of a document and it is not a document in the way a Word file is. It is a graph of objects — text-drawing instructions, embedded fonts, images, form fields, metadata — plus a table saying where each one begins. Everything in it is stored, not generated, and the file is the sum of those parts.

That matters because it means the PDF format itself is almost never the problem. A forty-megabyte PDF is forty megabytes of something, and finding out what is the entire job.

The four causes, in order of how often they are the answer

Scanned pages. Each page is a photograph of paper. At 300 DPI in colour, one A4 page is around 25 megabytes uncompressed and a few megabytes as JPEG. Ten pages of that is your whole file, and there is no text in the document at all.
Images at print resolution nobody downscaled. Somebody dropped 12-megapixel photographs into a report where they appear two inches wide. The PDF stores the full-resolution originals, because it has no way to know you will never zoom in.
Embedded fonts. A PDF carries the fonts it uses so it looks the same everywhere. Fully embedded rather than subset, a handful of typefaces with several weights each can add several megabytes to a document with almost nothing in it.
Revision history from incremental saves. PDFs can be updated by appending changes rather than rewriting, so a document edited fifty times may carry fifty generations of content, including things you deleted. This is the surprising one, and it is why a file sometimes shrinks dramatically when simply re-saved.

Working out which one you have

Two checks, thirty seconds:

Try to select a word. If text highlights, the document contains real text and your problem is images, fonts or history. If nothing highlights, it is a scan, and cause one is your answer.

Divide the file size by the page count. Under 100 KB a page is a normal text document and there is nothing to fix. One to three megabytes a page is a scan, or a document full of photographs. Over five megabytes a page means high-resolution colour scanning, and the fix is at the scanner rather than here.

What each fix costs

If it is a scan, compressing works well and costs you nothing that was there: the pages were already pictures. Compressing it re-renders each page at a lower resolution, and a 30 MB scan commonly comes back at 3 to 5 MB with no meaningful loss.

If it is a text document, browser-based compression is the wrong tool and it is worth being blunt about why. A browser cannot reach inside a PDF and re-encode the JPEGs buried in it; that needs a real PDF library operating on the file's internal structure. What a browser can do is render each page and save the result as an image — which shrinks a scan and destroys a document, turning selectable text into pixels and taking searching, copying, links, form fields and screen-reader access with it.

For that case, three better options in order: export again from the source with images at 150 DPI; split out only the pages you actually need, which is often the whole answer for a form; or run Ghostscript locally, which recompresses images and subsets fonts while leaving text as text. It is free, it runs on your machine, and it does the job properly. We do not do it and there is no point pretending otherwise.

When not to compress at all

A PDF of real text that is already under a megabyte is as small as it usefully gets, and running it through an image-based compressor will make it larger as well as worse: a picture of a page of text takes more space than the text did. If the document is going to be archived, signed, read by a screen reader, or searched, the text is the thing you are protecting, and file size is not worth trading for it.

Getting a large PDF somewhere it will not fit

If the obstacle is an upload limit rather than the file itself, splitting it is frequently the answer — most forms want three specific pages, not the whole bundle. If the pages need to go somewhere that does not take PDFs, exporting them as images at 144 DPI produces something far smaller than the original at screen quality. And if you are assembling a PDF from photographs, compress the images first: doing it in that order costs much less quality than compressing the finished document.

Questions

Why is my PDF larger than the images I put in it?

Usually because the images went in at their original resolution and the PDF added its own structure around each one. A PDF does not compress what it is given; it stores it and describes where it goes. If you assembled it from photographs, the file is the sum of those photographs plus overhead, which is exactly why compressing the images before assembly works far better than compressing the finished document.

Does compressing a PDF lose the text?

In a browser, yes, and that is the honest limitation. Shrinking a PDF here means rendering each page and storing the result as a picture, so real text becomes pixels: unsearchable, uncopyable and invisible to a screen reader. On a scan there was never any text to lose. On a report there was, and the right answer is usually to reduce the file another way.

How can I tell if my PDF is a scan?

Try to select a word in any PDF reader. If a highlight appears, the file contains real text and it is a document. If nothing highlights, or the whole page highlights as one block, it is a picture of a page and there is no text in the file at all. This single test tells you which of the fixes below will work, and it takes two seconds.

What is the best way to shrink a PDF without losing quality?

Go back to whatever produced it and export again with images at 150 DPI rather than 300, which is usually the whole problem. Failing that, Ghostscript on your own machine will recompress the embedded images and subset the fonts while leaving the text as text. No browser-based tool can do either, and anything claiming to is doing it on a server with your document.