Why is my PDF so big?

A PDF is a container, not an image format. Ten pages of text is about 40KB. Ten pages of scanned paper is 40MB. Same page count, same apparent content, a thousand times the size.

So the first question is never "how do I compress this" — it is what is actually in it, because the answer decides whether making it smaller is free or destructive.

The four things that make a PDF large

1. Embedded raster images

By far the most common cause. Every photograph, logo, chart exported as a picture, or scanned page is stored as a full raster image, usually at whatever resolution it arrived at.

A single phone photo dropped into a report is 4MB inside the PDF just as it was outside it. Ten of them is 40MB, and the document looks like ten pages of text with pictures.

This compresses extremely well, because the images are almost always at far higher resolution than the page needs.

2. Scanned pages

A scan is the same problem taken to its conclusion: there is no text in the document at all, only a photograph of text. A page scanned at 600dpi in colour is about 100 megapixels of image data for something that reads as a page of words.

This compresses well too, but the trade-off is different — and worth its own guide, because the usual lossless pass cannot help you.

3. Embedded fonts

Every font used must be carried inside the file so it renders the same everywhere. A full CJK font is 10–20MB; a normal Latin font is 50–400KB per weight.

A document using six weights of two families can carry 2MB of fonts before any content. Subsetting — embedding only the glyphs actually used — usually cuts this by 90%, and most modern exporters do it automatically. If yours did not, this is pure waste.

4. Structure nobody asked for

Revision history, form fields, embedded colour profiles, JavaScript, annotations, thumbnails, and — the big one — incremental saves.

A PDF can be edited by appending changes rather than rewriting the file. Save twenty times in some tools and the file contains twenty generations of the document, all of it still there. This is why a file sometimes grows every time you touch it without the content changing.

This is free to remove. Entirely. It is the only category where compressing costs you literally nothing.

How to tell which one is yours

A rough diagnostic without any special software:

Pages x 5KB  ~= your file size?   -> Plain text. Already small.
Pages x 50KB ~= your file size?   -> Text with some graphics.
Pages x 500KB or more?            -> Embedded photos, or a scan.
File shrinks hugely on "Save As"? -> Incremental save bloat.
Can you select the text?          -> No = it is a scan.

That last one is the most important question on this page. Try to select a sentence with your cursor. If you can, the document has real text and a lossless pass can help. If you cannot, it is images of text, and every option from here involves a decision about resolution.

Why "lossless first" is not marketing

PDFs can be made smaller two fundamentally different ways, and the difference is permanent.

The lossless route rewrites the file's internal structure: compressing the object streams, removing orphaned objects, dropping the revision history, deduplicating repeated resources. Text stays text. Search works. Copy-paste works. Screen readers work. Typical saving on a text-heavy document with save bloat: 30–70%.

The lossy route renders each page to a bitmap and rebuilds the PDF from those images. It always works and can reach any target — and it destroys the text layer permanently. The document becomes pictures of pages. Search finds nothing, text cannot be copied, and anyone using a screen reader gets silence.

A tool that rasterises a document it could have compressed losslessly has taken something irreplaceable in exchange for nothing. So the order should always be: try lossless, measure the result, and only rasterise if the number still is not met. The full argument is here.

What rasterising actually costs, if you get there

If lossless is not enough, the question becomes how much resolution to keep. The relevant number is DPI, and a PDF's internal grid is 72 points per inch, so:

Render scale	DPI	Looks like
2.0	144	Crisp on screen, fine for print of a scan
1.5	108	Comfortable reading, text clearly sharp
1.0	72	Legible, slightly soft on small type
0.6	43	Readable headings, body text struggling

Resolution should be spent before quality: blur from too few pixels and blur from a hard JPEG setting look different, and running out of pixels is the uglier of the two. So a sensible tool searches scale first, holding quality high, and only drops quality when the scale floor is reached.

Hitting a specific number

PDF size is not predictable from the inputs. The same scale and quality on a text-heavy page and a photographic page differ by an order of magnitude, so any tool that computes a setting from your target is guessing.

What works is measuring. One render pass costs real time, but it costs far less than handing back a file that gets rejected:

Lossless rewrite only        2.4MB   over 1MB, must rasterise
scale 2.0  quality 0.78      3.1MB   over
scale 1.3  quality 0.78      1.4MB   over
scale 1.0  quality 0.78      940KB   under, best so far
scale 1.15 quality 0.78      1.1MB   over
scale 1.07 quality 0.78      990KB   under, and the best found

Compress a PDF to 1MB, to 500KB or to 2MB does exactly that, lossless pass first, in your browser. Nothing is uploaded — which for contracts, medical records, bank statements and ID scans is usually the point.

FAQ

What makes a PDF file large?

Embedded raster images, overwhelmingly — a single phone photo inside a document is several megabytes. After that: scanned pages, which are nothing but images; embedded fonts, especially CJK fonts or many weights without subsetting; and accumulated revision history from incremental saves, which can double or triple a file with no visible change to it.

Why does my PDF get bigger when I save it?

Most likely incremental saving. The PDF format allows a program to append changes rather than rewrite the whole file, so each save adds a new generation while keeping the old one. Twenty edits can mean twenty copies of the document inside one file. A "Save As" to a new file, or any tool that rewrites the object streams, discards all of it losslessly.

Can I reduce a PDF without losing quality?

Often, yes — if the document has real text. Rewriting its internal structure compresses the object streams and discards revision history, orphaned objects and duplicated resources, typically saving 30–70% on a text-heavy file with nothing lost at all. If the PDF is a scan, there is no text layer to protect and the only lever is image resolution, which does cost something.

How do I know if my PDF is a scan?

Try to select a line of text with your cursor. If the selection highlights words, the document has a real text layer. If nothing selects, or the whole page highlights as one block, it is an image of a page — and that changes every option available to you, because there is no text to preserve and resolution is the only dial.