A PDF that takes forty seconds to open, stutters when you scroll, or closes itself on your phone is doing something people reasonably mistake for corruption. It is not damaged. It is asking for more memory than the reader has, and the reader is losing.
The unhelpful part is that file size alone tells you very little. A 40 MB document can open instantly while an 8 MB one crawls, because the two files are large for entirely different reasons — and those reasons determine which fix will work and which will just cost you quality for nothing.
What "too large" actually means
"Too large" is not a single threshold; it is four different ceilings, and which one you hit determines what you see:
| Limit | Roughly where it bites | Symptom |
|---|---|---|
| Email attachment caps | 25 MB Gmail, 20 MB Outlook, often less on corporate servers | The message will not send |
| Web upload limits | 10–50 MB on most portals and government forms | Upload rejected before it starts |
| Mobile reader memory | Varies; image-heavy files struggle well before 100 MB | Opens then closes, or never opens |
| Desktop reader performance | No hard limit — degrades gradually | Slow scrolling, long open times, spinning cursor |
The first two are hard limits that reject the file outright. The second two are soft, and they depend far more on what a page contains than on the total megabytes. A thousand-page text document is easy work; a fifty-page catalogue of full-bleed photographs is not.
Why big PDFs choke a reader
A PDF page is a set of drawing instructions, and rendering one means executing them. When a page references a 25-megapixel photograph, the reader must decompress that image into memory in full before it can scale it down to the size of your screen — a 4 MB compressed image can occupy 100 MB uncompressed. Scroll through ten such pages and a mobile reader with a few hundred megabytes to work with runs out and is killed by the operating system.
Two other patterns produce the same result by different routes. Vector artwork — maps, CAD exports, complex charts — can hold hundreds of thousands of individual paths on a single page, all of which must be drawn every time that page appears; those files are often small on disk and brutally slow to render. And documents assembled by repeatedly merging other documents can accumulate dozens of copies of the same embedded fonts, inflating the file without adding a single visible thing.
What is actually taking up the space
In practice, the weight is almost always in one of these five places:
- Scanned pages at unnecessary resolution. By far the most common cause. Scanners default to 600 dpi in full colour, which produces roughly ten times the data of the 150–200 dpi that is entirely sufficient for reading and printing text.
- Photographs embedded at camera resolution. A phone photo dropped into a report is 12 megapixels; the page displays perhaps one. The other 90% of the pixels are stored, transmitted and decompressed for no benefit.
- Scans saved without compression. Some scanners and "print to PDF" paths store page images as raw bitmaps. These files are enormous — tens of megabytes per page — and compress dramatically.
- Duplicated embedded fonts. Every merge can add another copy of the same font set. Invisible, and occasionally a third of the file.
- Dense vector artwork. Maps, floor plans and CAD exports. Modest on disk, punishing to render, and the one cause that image compression does not help at all.
Work out where the weight is
Two quick calculations tell you which fix to reach for, and save you from degrading a file that was never going to get smaller:
- Divide file size by page count. Over about 1 MB per page means images dominate and compression will be dramatically effective. Well under 500 KB per page on a file that still runs slowly means the problem is vector complexity or page count, and compression will barely move it.
- Try to select text on a page. If you cannot, the page is an image — a scan — and image compression is exactly the right tool. If the text selects cleanly and the file is still huge, the weight is in fonts or vector content instead.
That second check has a useful consequence. A scanned document has no text layer at all, so it cannot be searched and its contents cannot be copied. If you are going to process the file anyway, running it through OCR PDF adds a searchable text layer underneath the page images without changing how they look — which turns an unwieldy scan into something you can actually work with.
Make a large PDF usable, in three steps
For the image-heavy files that make up most cases, this takes under a minute in the browser:
Upload the file
Go to pdfdocshift.com/compress-pdf and drop the document onto the upload area. Files up to 200 MB are accepted, which covers essentially every case where a reader has been struggling.
Choose a quality level to match the destination
High quality downsamples conservatively and suits anything destined for print. Medium targets around 150 dpi and is the right default for documents that will be read on screen and emailed. Low is for cases where fitting under a hard limit matters more than fidelity.
Check the result at full zoom
Download the file and look at the most detailed page — a page with small print, a signature or a stamp — at 100% zoom rather than fitted to the window. If the text is crisp, the compression was free. If it is soft, step back up one quality level and run it again.
Compress a PDF that is too large
Reduce file size by up to 70% with a quality level you choose. Free, no sign-up, files deleted after 2 hours.
Open Compress PDF →When compression is the wrong answer
Compression works on images. If the weight is not in images, it will cost you quality and return very little, and you need a different approach.
- Very long documents. A two-thousand-page file is slow because of its page count, not its bytes. Split PDF divides it into chapters or fixed-size chunks, each of which opens instantly — and splitting is usually the better answer for email limits too, since a recipient would rather receive three readable files than one that bounces.
- Dense vector artwork. Image compression has nothing to act on. Ask whoever produced the drawing to export a flattened or reduced-detail version; there is no meaningful client-side fix.
- Only a few pages are needed. Extracting the relevant pages beats compressing the whole document. A five-page extract from a three-hundred-page report is smaller than any compression will achieve, and more useful to read.
- The file is also damaged. If it is both huge and unreliable — opens sometimes, stalls other times — repair it first with Repair PDF. Compressing a structurally broken file tends to bake the damage in.
Keeping PDFs small from the start
Most oversized PDFs are made oversized at the moment of creation, by a default nobody chose deliberately:
- Scan text at 200–300 dpi in greyscale. Colour scanning of black-and-white paper triples the data for nothing. This one change is usually worth more than any compression pass afterwards.
- Resize images before placing them in a document. A photo scaled down to the size it will actually appear costs a fraction of the full-resolution original, and looks identical.
- Export rather than print to PDF where you can. Word, Pages and Google Docs all produce far more efficient files through Export or Save as PDF than through the print dialog.
- Merge once, at the end. Repeatedly merging and re-merging accumulates duplicate font data with every pass. Assemble the final document in a single operation.
- Compress before archiving, not after distributing. Once a large file has gone out to twenty people, the size problem has already been everyone's problem.
One last distinction worth holding onto: a file that is slow is a size problem, and a file that refuses to open at all is not. If your reader shows an error rather than struggling, start with what that error message means instead — compressing a file that will not open is not going to help.
Frequently asked questions
Mobile readers have far less memory to work with, and a PDF page must be decompressed in full before it can be displayed — a 4 MB compressed photograph can occupy 100 MB in memory. A desktop absorbs that easily; a phone runs out and the operating system closes the app. The file is not damaged, and compressing it to reduce the image data resolves it.
It depends on the limit you are hitting rather than on any single number. Email caps are hard: 25 MB on Gmail, 20 MB on Outlook, often lower on corporate servers. Web upload forms commonly stop at 10–50 MB. Reader performance has no fixed ceiling and degrades gradually, driven far more by what a page contains than by total size — a thousand pages of text is easier work than fifty pages of photographs.
Not the text, if the document has a real text layer. Text is stored as characters plus font instructions and is redrawn sharply at any size, so compression leaves it untouched. What compression reduces is image resolution — which matters if your pages are scans, because then the text is part of the image. For scanned documents, use a higher quality level and check a detailed page at 100% zoom.
That is the signature of dense vector artwork — maps, floor plans, CAD exports and complex charts, which can hold hundreds of thousands of individual paths that must all be drawn each time the page appears. Such files are small on disk and slow to render, and image compression has nothing to act on. The only real fix is a simplified export from whatever produced the drawing.
It depends where the weight is. Divide the file size by the page count: above roughly 1 MB per page, images dominate and compression will be dramatically effective. Below about 500 KB per page on a file that is still slow, the problem is page count or vector complexity, and splitting into smaller documents helps where compression will not. For email, splitting is often better regardless — three files that arrive beat one that bounces.
No. Compression works on embedded images and leaves the text layer intact, so a searchable document stays searchable. A scanned document has no text layer to begin with and remains unsearchable either way — if you want to search it, run it through OCR, which adds a text layer underneath the page images without altering how they look.
Ready to shrink your file?
Free, no sign-up, files deleted after 2 hours. Files up to 200 MB, with no installation.
Open Compress PDF →