How to Get a PDF Under an Email Attachment Limit

Every so often a document refuses to go anywhere. The mail client accepts the attachment, thinks about it, and returns a message about exceeding the maximum size. The document has not changed since yesterday; the limit has simply caught up with it.

Your limit is smaller than the number you were given

Email cannot carry binary data. Attachments are converted to text using Base64 encoding, which represents every three bytes as four characters — an increase of about 33% before headers are even counted. When a provider advertises a 25 MB limit, it is describing the encoded message.

So the practical ceiling for a single attachment is closer to 18 MB. Two more details make it stricter still: the recipient's server has its own limit, often lower than yours and applied on arrival, and forwarding a message re-encodes everything, which is why a thread that worked for three replies suddenly bounces on the fourth.

Providers change these values and administrators override them, so treat the table as orientation rather than gospel. The safe habit is to target about 70% of whatever number you were told.

Step one: find out where the weight is

Before compressing anything, work out what kind of document you have — the answer determines which tool can help at all.

• Can you select the text with your cursor? If yes, it is a real text PDF and much of its size may be structural waste. Lossless compression has something to work with.
• Does the text refuse to highlight, and does zooming in reveal soft edges? It is a scan. The bytes are image data, and only re-encoding those images will help.
• Divide the size by the page count. Under 100 KB per page is already lean. Over 1 MB per page means full-page images — a scan, a photo brochure, or a deck full of screenshots.

Step two: lossless compression

Start here, always, because it cannot make anything worse. Lossless compression repacks the file's internal structure — bundling scattered objects, rebuilding the cross-reference table, discarding data left behind by earlier edits — without touching a single visible pixel or character.

On a Word export or a much-revised report this frequently removes 15–40%. On a scan it removes almost nothing, which is not a failure: it is the file telling you the weight lives in the images.

Step three: re-render to a target

If the file is still too heavy, the next lever costs you something. Shrinking to a target size renders each page as an image and compresses that image, searching for the highest quality that fits under the number you set. Layout and page count survive intact; selectable text does not.

Set the target below the real limit — 18 MB for a 25 MB mailbox — and check the output before deleting anything. Look specifically at the small print: footnotes, stamps, handwritten signatures, table rules. These degrade before body text does.

For a colour scan, expect an 80–95% reduction with little visible change on screen. For a text document that was already efficient, expect a poor trade — sometimes even a larger file, because a page of crisp text is cheaper to store as text than as an image of text.

Alternatives worth preferring

Compression is not always the right answer, and reaching for it reflexively can produce a document nobody can read comfortably.

• Split it. Two 12 MB files usually beat one degraded 18 MB file. Contracts, reports, and anything with appendices split along obvious lines.
• Remove what nobody asked for. Blank separator pages from a scanner, duplicated cover sheets, and appendices attached out of habit are free savings.
• Send a link. Any cloud drive will do. For genuinely large documents this is the professional answer, and it is what most recipients quietly prefer.
• Fix it at the source. If the file came from your own scanner, rescanning at 200 DPI in greyscale may produce something a tenth of the size, with no compression step needed. See why scanned PDFs are so large.

What to do about "compress PDF online" results

Most of the tools at the top of that search upload your document to a server, process it there, and delete it after some retention window. For a marketing brochure this is unremarkable. For a payslip, a passport scan, a medical report, or a client's contract, you have handed a copy of the document to an organisation you have not evaluated, and under the GDPR you may have made a decision on someone else's behalf.

Browser-based tools — including the ones on this site — avoid the question because there is nothing to send. The test is straightforward: load the page, disconnect from the internet, and try again. If it still works, the file never went anywhere. We wrote more about this in what happens to a PDF you upload.