The short version

Aim for under 18MB for personal Gmail and similar, and under 10MB if the recipient is on a corporate address. The stated limits are higher than the working ones because attachments are encoded for transit and grow by about 37 percent.

Then pick the technique that matches the file: remove blank pages from scans, re-export from the source for Office documents, compress images for everything else.

The attachment bounced, or the send button greyed out, and now the file has to get smaller. What most people do next is run it through a compressor, look at the result, and run it through again when the number is still too high.

That works for some files and does nothing for others, and the difference is predictable. It is worth thirty seconds of diagnosis before you start.

What number you are actually aiming at

Email attachment limits are not the numbers the providers advertise. Attachments cannot be sent as raw binary, so they are encoded into a text-safe format for transit, and that encoding inflates them by roughly a third. The advertised cap applies to the encoded size, not to the file on your disk.

AnatomyThe number your provider measures
On your disk22 MBOn the wire, after encoding30.1 MB+37%25 MB capthe cap applies to this bar, not the one above
The cap applies to the encoded attachment, which is about 37 percent larger than your file.
Gmail's stated cap
25 MB

Applies to the encoded attachment on the wire, not the file in your folder.

Encoding overhead
+37%

Base64 expansion plus MIME headers. Fixed, unavoidable, applies everywhere.

Safe working target
18 MB

Raw file size that clears personal accounts reliably. Use 10MB for corporate.

The recipient's server has the final say
Your provider accepting the send does not mean the message arrives. Every server in the chain applies its own cap and the tightest one wins. Corporate Exchange installations commonly cap inbound attachments at 10MB, and many of them reject silently without bouncing. If a large file has to reach a corporate inbox and you hear nothing back, do not assume it landed.

Work out what kind of file you have

Two questions settle which technique will help. Is the document scanned or born-digital? And does its size look proportionate to its page count?

Try it

Will your attachment actually send?

Enter a file size to see what it becomes on the wire and which inboxes will take it.

What the answer tells youDiagnosis
  • Scanned, very largeBlanks and downsampling
  • From Word or Docs, largeRe-export from the source
  • Few pages, enormousOne oversized image inside
  • Many pages, text onlyAlready small, split instead
  • Has photographsCompress, check quality after
  • Already compressed onceGo back to the original

Five techniques, most effective first

Remove pages that carry no information

On a duplex-scanned document this is frequently the biggest single saving available, and it is the only one that costs nothing at all in quality. Half the pages are photographs of blank paper at 20 to 80KB each. Removing them can halve the file before any compression happens.

Re-export from the original source

PDFs generated from Word, Pages or Google Docs vary enormously in size depending on the export settings used. A document exported with print-quality image embedding can be several times larger than the same document exported for screen viewing. If you still have the source file, re-exporting properly is faster than compressing and produces a cleaner result, because the images are downsampled once from the originals rather than re-encoded from already compressed copies.

Compress the images

The general purpose answer, and the right one when the document contains genuine photographic content. Downsampling resolution to match the destination does most of the work, and re-encoding at lower quality does the rest. This is where a compressor earns its place.

Split the document

Two 9MB emails reliably beat one 18MB email that bounces. This is standard practice in legal work, where disclosure bundles are routinely delivered as numbered parts. It is also the only technique that does not degrade anything.

Send a link instead

The fallback when the file genuinely cannot shrink and cannot sensibly be split. It works, and it changes the recipient's experience: a link rather than an attachment, sometimes a sign-in prompt, and a file that breaks if you ever move it.

Order matters more than people expect
Removing blank pages then compressing gives a better result than compressing then removing blanks, because the compressor spends effort on pages you were going to delete. Re-exporting from the source beats compressing the bad export for the same reason: you avoid stacking two lossy passes on the same images.

Keeping it readable

There is a floor below which a document stops being useful, and hitting an arbitrary size target is not worth crossing it.

Sensible compression

Smaller and still usable

  • Text stays crisp at 100 percent zoom
  • Resolution matched to how it will be read
  • One pass, from the original file
  • Checked on the page with the smallest text
  • Original kept in case you need to retry
Compressing to hit a number

Smaller and no longer useful

  • Grey halos around small text
  • Photographs turned blocky and posterised
  • Multiple passes stacking artifacts
  • Checked at fit-to-window, where damage hides
  • Original overwritten, no way back
Check before you send, not after
Open the compressed file, go to the page with the smallest text, and zoom to actual size. If it reads cleanly, send it. If not, the recipient will be the one to discover that, and by then it is an email thread rather than a ten second check.

Stopping the problem recurring

Most oversized attachments are made rather than found, and a few defaults prevent almost all of them.

Four settings worth changing once

Scan in greyscale. Unless colour carries meaning, greyscale is roughly a third the size for an identical looking document.

Scan at 300 DPI, not 600. Six hundred is an archival setting. For anything that will be read rather than reproduced, it is four times the pixels for no visible benefit.

Turn duplex off for single-sided originals. It halves the page count and the file size in one setting.

Export for screen, not press. Unless a printer is genuinely involved, the press-quality preset in Office and design tools is wasted weight.

Get the file under the limit

Compress in your browser and check the result before you send it. Nothing is uploaded.

Open Compress PDF

Common questions

Getting under a specific number.

01What size should I aim for?

Under 18MB for personal accounts such as Gmail, and under 10MB if the recipient is on a corporate address. Both are below the advertised caps because attachments grow by about 37 percent during encoding, and the limit applies to the encoded size.

02Why did my email fail when the file is under the stated limit?

Because the limit applies to the attachment after encoding, not to the file on your disk. A 22MB PDF becomes roughly 30MB in transit, which is over a 25MB cap even though the file itself is not.

03What is the fastest way to shrink a scanned document?

Remove the blank pages first. Duplex scanning of single-sided paper doubles the page count with photographs of empty paper, and deleting them often halves the file with no quality cost at all. Compress afterwards.

04Is it better to compress or to split?

Split when the content genuinely needs its resolution, such as drawings or photographs, or when compression has already reached the point where text is suffering. Two smaller emails always arrive. One oversized email does not.

05Why is my PDF huge when it is only three pages?

Almost certainly one oversized image. A photograph placed at full camera resolution and scaled down on the page keeps every one of its pixels in the file, even though only a fraction of them are visible. Compression fixes this well because there is so much redundancy to remove.

06Does compressing twice help?

No. The second pass works on the artifacts left by the first and treats them as detail worth preserving, so you lose noticeable quality for very little size. If the first result is still too large, return to the original and compress harder in a single pass.