Why is my PDF so large, and what actually makes it smaller?

A four-page PDF that weighs 38 MB feels like something has gone wrong. Usually nothing has: you just cannot see what is taking up the room.

What is actually in there

The proportions surprise people.

Text costs almost nothing. A page of prose is a few kilobytes. The characters are stored as codes with positions, and the font is stored once for the whole document even if it appears on four hundred pages. A 500-page novel as a text PDF can be under 2 MB.

Vector graphics cost almost nothing. Lines, boxes, logos drawn as shapes. These are instructions, not pixels. A complex diagram might be 50 KB.

Images cost everything. A single phone photo at full resolution is 3–6 MB. One scanned page at 300 DPI in colour is 5–8 MB before compression. Four scanned pages and you are at 30 MB without a word of text anywhere.

So the practical rule: if a PDF is large, it contains images. Everything else is rounding error. That immediately tells you where to look and what will and will not help.

The two things that make a PDF enormous

It is a scan. Every page is a photograph. There is no text at all, only pixels, which is also why you cannot select anything in it.

Images were embedded at full resolution. Someone dropped 12-megapixel photos into a Word document, scaled them down to postcard size on the page, and exported. The PDF still contains the full 12 megapixels; it just draws them small. This is extremely common and the file can be ten times larger than it needs to be.

What compression actually does

Compress PDF mainly re-encodes the images: lowering their resolution to something appropriate for the size they are actually displayed at, and increasing JPEG compression. It also strips duplicated resources and unused objects, but that is a rounding error next to the images.

Which means:

• A scan compresses dramatically. 40 MB to 4 MB is routine, and often looks no different on screen.
• A text-only PDF barely compresses at all. There is nothing to squeeze. If a 900 KB text document comes back at 850 KB, that is the honest maximum, not a failure.

Expecting the second case to behave like the first is where most disappointment comes from.

Choosing how hard to compress

More compression means more visible degradation, and where the line falls depends entirely on what the document is for.

The failure that matters is compressing a document that will be printed, then discovering at the print shop that the images are soft. Always keep the original until the job is done.

When compression is not the problem

Sometimes "too big" is the wrong diagnosis. Two other shapes of the same complaint:

The document is too long, not too heavy. A 200-page report that is mostly whitespace is a different problem. It is not about megabytes, it is about how many sheets it is. SmartShrink combines pages that are mostly empty, which reduces the page count rather than the file size. Measured on a realistic report: 200 pages to 83, with the text at its original size.

There is content that should not be there. Embedded attachments, an unused layer, a full-resolution logo repeated on every page. These are rarer, but if a text document is inexplicably large this is where to look.

Practical ways to avoid the problem

Most oversized PDFs are created rather than discovered:

• Scan at 300 DPI, in greyscale for plain documents. Colour at 600 DPI is four times the data for a page of black text.
• Resize images before inserting them. Scaling a photo down in Word changes how it is drawn, not what is stored.
• Export properly. Word's "Minimum size" option and Acrobat's "Optimized PDF" both downsample as they go.
• Do not stack conversions. Every round trip through another format tends to grow the file and lose a little quality.

If it still will not fit

When the file genuinely cannot get small enough:

Split it into logical parts and send them separately, but only if the parts stand alone. Four chunks someone has to reassemble is worse than a link.

Send a link instead. Most email limits are 20–25 MB; cloud storage has no such problem and gives you a record of who opened it.

Ask what they actually need. Often the answer is three pages of a sixty-page document, and extracting those solves the problem completely.

Finding out what is actually taking the space

Guessing wastes time. Two checks tell you almost immediately.

Divide the size by the page count. Under 100 KB a page is a text document and there is little to win. Over 500 KB a page means images, and that is where all the size is. Several megabytes a page means full-resolution photographs.

Look at where the size is concentrated. Split the document in half with Split PDF and compare the two halves. If one is ten times the other, the problem is a handful of pages rather than the document as a whole, usually a scanned appendix or a page with a large photograph on it.

That second check turns a vague "the file is too big" into "pages 22 to 24 are the file", which is a problem with an obvious fix.

The case nobody expects: fonts

Occasionally a modest-looking document is unexpectedly large and there are no images at all.

The usual cause is fully embedded fonts rather than subsets. A complete CJK or Indic font family can be several megabytes on its own, and a document that embeds four weights of one is carrying twenty megabytes of letterforms it barely uses.

You can see this in the Fonts tab of the document properties: entries saying "Embedded" rather than "Embedded Subset". Re-exporting from the original document with subsetting turned on is the real fix, and it is a setting in the export dialog of most word processors.