Why a scanned PDF is large, and how to compress it safely

Someone scans a six-page contract at 300 DPI, gets a 14 MB PDF, and uploads it to a portal with a 2 MB limit. The first instinct is to run the file through the compressor again and again. The file barely moves. That is not because the compressor is weak — it is because a scanned PDF is not a document in the sense most people assume. It is a stack of photographs wearing a PDF wrapper, and the compressor can only recompress photographs.

Start by inspecting the PDF and checking its actual byte count and page count. If the file contains unnecessary pages, select the required pages before compressing. After processing, use file preflight to compare the output with the receiving form’s requirements. A smaller PDF is useful only if the required content remains readable.

A scanned PDF is a container

Two kinds of PDF page

A PDF page can hold two very different kinds of content:

Page content	Typical source	Where the size goes
Vector text and fonts	Word, LaTeX, browser "print to PDF"	Font programs, embedded once and reused
One full-page bitmap image	Scanner, phone camera, MFP	The image's pixels, quality and colour

The second row is what a scan produces. Your scanner does not understand letters. It captures a raster image of the page — at 300 DPI on A4 that is roughly 2480 × 3508 pixels per page — and puts it in the PDF. Six pages means six of those images.

So the size question turns into an image question: how many pixels, how many colours, how much high-frequency detail, and what quality setting. The PDF format itself contributes very little. That is also why a text-native PDF exported from a word processor is often already small and gives you almost nothing to reclaim.

Why running the compressor twice does nothing

The page images are already JPEG

The most common wasted effort is compressing the same scan repeatedly, hoping each pass shaves off a bit more. It does not, for a concrete reason: the page images inside a scanned PDF are already JPEG-compressed. JPEG is lossy, and its output is not meaningfully more compressible than its input. Re-encoding the same JPEG at the same quality produces a file of roughly the same size; re-encoding at lower quality loses detail that is already gone.

What a lower scan DPI would have changed

The second pass cannot recover the two things that would actually help — pixels that were never captured at a lower resolution, and detail that the original encoding already discarded. Only a genuine re-encode at different settings, or fewer pixels, changes the arithmetic.

Identify what is taking the space

Before changing anything, find out what is in the file. Try selecting a sentence in your PDF reader. Selectable text means a text layer exists, but it is not conclusive: a scan can carry an OCR layer sitting invisibly over a large background image, and the image is still what fills the file. PDF inspection reports page count, page dimensions and embedded image and font facts — that is where the answer lives.

Typical sources of avoidable size in a scan:

• Overscanned pages, captured at 600 DPI when the content does not need it
• Photographic noise across white margins — scanner gain set too high
• Duplicated or blank pages included by accident
• Colour capture of a document that is black and white
• Desk edges, fingers or shadows included in the frame

Keep the original before processing. If a reduced version turns out unreadable, the original is the only way back.

Why handwriting resists compression

Every pencil stroke is an edge

A page of typed text compresses well because it is mostly flat white with a few predictable black strokes. A page of handwriting is the opposite: pencil or pen on textured paper produces thousands of tiny irregular edges, and every one of those edges is high-frequency content. Noise from a low-quality scanner adds more of the same, and it is indistinguishable to the encoder from a genuine pen stroke.

Once you see that, the trade-off becomes explicit. The compressor's only lever on such a page is smoothing away fine detail — and on a handwritten signature, that fine detail is the content. A signature reduced to a grey blur is worse than a rejected upload, because it may be accepted and then questioned later.

Reduce size in sensible steps

The five passes that are worth doing

1. Remove pages that genuinely are not needed, if the submission's rules allow it. This is the only step that reduces size with zero quality loss.
2. Open the PDF compressor and set the actual byte target from the portal's rules, not from guesswork.
3. Use strict mode when the limit is hard. If the tool cannot meet it, read the failure rather than submitting an unverified candidate.
4. Compare page count and page dimensions against the original. A page silently dropped is the worst kind of success.
5. Zoom in on small text, stamps, handwriting and thin rules. Then open the result in a second PDF reader if you have one, so a single renderer's quirk does not mislead you.

AttachReady can optimise PDF objects and recompress supported embedded images. Some masks and complex colour spaces are preserved rather than aggressively rewritten, so the result can come with warnings — read them. There is no universal "lossless to 200 KB" setting. A PDF that is already efficiently encoded may barely shrink, and that is a real answer, not a failure of the tool.

Page-level and document-level operations

Which operation can actually shrink the file

Knowing which operations act on pages and which act on the file as a whole saves a lot of confusion:

Operation	Scope	Effect on size
Select pages	Page level	Proportional to pages removed
Rotate	Page level	Essentially none
Merge	Document level	Adds up the inputs; never shrinks
Compression	Document level	Re-encodes embedded images
Re-scanning	Source	Changes everything downstream

The practical consequence: if you need a big reduction on a scan, the decisive move usually happens before the file ever reaches a compressor — at the scanner, in the choice of DPI and colour mode.

A verify-only path when you cannot risk quality

Split, inspect, compare, keep the original

Sometimes the right answer is not to compress at all. If the document is a signed contract, a certified translation or anything where legibility is legally relevant, try this instead:

1. Check whether the portal accepts multiple uploads. Splitting a six-page scan into three two-page files often fits limits with no re-encoding at all.
2. Inspect the file to confirm whether it is actually scan-heavy or text-native — a text PDF needs no image work.
3. If reduction is unavoidable, compress a copy and compare the output page by page against the original at 200% zoom before submitting.
4. Keep the original and never submit a result you have not opened yourself.

Protected, signed and unusual files

Encrypted PDFs must be decrypted by someone authorised to do so before processing. PDFs with active content, embedded attachments or executable or external actions may be rejected — that is a processing boundary, not a claim that accepted files are malware-free.

Rewriting a PDF can invalidate a digital signature. Keep the signed original and use the recipient's approved process when a signature must stay valid. Compression is not a substitute for re-signing or for obtaining a fresh official document.

When the target is unrealistic

A long, detailed, handwritten scan may not stay legible at a tiny byte limit. In that case, try a cleaner scan at a sensible resolution, convert to grayscale if colour is not required, split the document only if multiple uploads are allowed, or ask the receiving organisation about another submission channel. Do not delete pages that the submission requires just to make a progress indicator turn green.

For related workflows, see merge and select PDF pages and upload error diagnosis. The formats and limits page lists the ceilings that apply per access tier.