Compressing a PDF for an online form

Government portals, exam registration systems, visa applications, tender submissions and HR systems share a habit: they demand a document under a small specific size, and they reject anything over it without telling you by how much.

The sizes are almost always 100KB, 200KB, 300KB or 500KB. Your document is 4MB. Here is how to close that gap without handing in something nobody can read.

First, check whether you are allowed to lose the text

This is the decision that everything else follows from, and it takes five seconds: try to select a line of text in your PDF.

If text selects, the document has a real text layer, and you have the good case. A structural rewrite — compressing object streams, dropping revision history, removing orphaned objects — frequently saves 30–70% with nothing lost at all. Many documents reach a 500KB target this way and never need anything lossy.

If nothing selects, it is a scan, and every option from here is a resolution decision. Scans have their own guide because the trade-offs are genuinely different.

A tool should always try the lossless route first and measure it. Rasterising a document that could have been compressed losslessly destroys search, copy-paste and screen reader access permanently — and for a document you are submitting to an institution, searchability is sometimes the thing they actually need.

What the budget per page really is

The useful way to think about a target is bytes per page, because that is what determines whether it is comfortable or hopeless:

Target	1 page	5 pages	20 pages
100KB	100KB/page — fine	20KB/page — tight	5KB/page — text only
200KB	roomy	40KB/page — workable	10KB/page — marginal
500KB	roomy	100KB/page — comfortable	25KB/page — tight
1MB	roomy	roomy	50KB/page — workable

For reference: a page of pure text with embedded fonts is 3–8KB. A page of text with one photograph is 100KB–2MB. A scanned page at 150dpi greyscale is 30–60KB.

So "compress this 20-page scan to 100KB" is a request for 5KB per page, which is below what any image of a page can do legibly. If that is your situation, the honest answers are to split the submission, re-scan at a lower setting, or go back and check whether the limit applies per file or per upload.

The order of operations

1. Rewrite losslessly and measure. Costs nothing, sometimes finishes the job.

2. Downsample the embedded images, not the page. If the document is text plus photographs, the photographs are the whole file. A 4000-pixel-wide image on a page that displays it at 500 points does not need to be 4000 pixels. This keeps the text as text.

3. Convert colour to greyscale, if the content allows. Roughly a 60% saving on scanned pages. Check first whether the form requires colour — many ID and certificate uploads specifically do.

4. Only then rasterise, and spend resolution before quality. If the page must become an image, keep as many pixels as the budget allows and hold the JPEG quality high. Blur from too few pixels looks worse than blur from a slightly harder quality setting, so the search should move resolution first.

The format trap

Several portals specify both a size limit and a format, and they are not always compatible with each other in the obvious way.

• "PDF only, under 200KB" — straightforward, the above applies.
• "JPEG only, under 200KB" for a document — then do not make a PDF at all. Export the page as an image and treat it as a photograph of text, which has different rules.
• "PDF, under 200KB, must be machine-readable" — this rules out rasterising entirely. If the lossless pass cannot reach 200KB you cannot satisfy the form by compressing, and the real fix is a smaller source document.
• A minimum as well as a maximum. Less common for PDFs than for photos, but it happens. Compressing to the smallest possible file is a rejection when a floor exists; landing just under the ceiling clears both.

Check the stated requirement for all of it, not just the number. The specifications we have verified against a primary source are here, along with an honest account of the ones we will not publish because no primary source states them.

Why measuring matters more here than anywhere else

PDF size is not predictable from the settings. Two documents at identical render scale and quality can differ tenfold, because one is text on white and the other is full-bleed photographs.

That means any tool computing a setting from your target is estimating, and estimating fails in both directions. Over, and the form rejects you again. Under by a lot, and you have handed in a 40KB illegible document when 200KB was allowed — readability you were entitled to and gave away.

Lossless rewrite only         780KB   over 200KB
scale 2.0 (144dpi) q0.78      1.2MB   over
scale 1.1 (79dpi)  q0.78      310KB   over
scale 0.85 (61dpi) q0.78      190KB   under, best so far
scale 0.95 (68dpi) q0.78      240KB   over
scale 0.90 (65dpi) q0.78      198KB   under, and the best found

Compress a PDF to 100KB, to 200KB or to 500KB runs that search in your browser and ships the best result that genuinely measured under your number. The document is never uploaded, which for the things people put in portals — passports, degree certificates, bank statements, medical reports — is not a minor detail.

Before you submit

• Open the output and read it. Not skim it — actually read the smallest text on the busiest page. A file that fits and cannot be read has failed in a way the upload form will not catch.
• Check the page count survived. Some compressors silently drop pages past a limit.
• Check the filename. A surprising number of portals reject spaces, non-ASCII characters, or names over 50 characters, and report it as a generic upload failure.

FAQ

How do I compress a PDF to 100KB?

If the document has selectable text, start with a structural rewrite — compressing the object streams and dropping revision history often saves 30–70% losslessly, and a short text document can reach 100KB that way with nothing lost. If it is a scan, 100KB means roughly 100KB for one page or 20KB each across five, which puts you at low resolution; convert to greyscale first, then let resolution fall as far as legibility allows.

Why does the portal reject my PDF when it is under the limit?

Usually something other than size. Common causes: the form wants a different format, the filename contains characters the validator refuses, the page count exceeds a separate limit, the PDF is encrypted or password-protected, or there is a minimum size as well as a maximum. Some validators also reject PDFs above a certain version number, which a modern exporter may produce by default.

Will compressing a PDF make the text unselectable?

Only if the tool rasterises it, and a good tool tries not to. A lossless structural rewrite keeps every glyph as a glyph. Rendering pages to images will reach any target but permanently ends search, copy-paste and screen reader access — so it should be the last resort, used only after a lossless pass has been measured and found insufficient.

What is the smallest a readable PDF can be?

About 3–8KB per page for pure text with embedded fonts, which is why a 20-page text document under 200KB is comfortable. For a scanned page, 20–30KB each is roughly the floor for legible 12pt print in greyscale. Below that you are under 100dpi and body text begins to break down regardless of what tool you use.