Why Is My PDF So Large?
Updated August 30, 2026
Twenty-two pages of plain text, and the file is 24 MB. Nothing on the page explains it. There is one test that identifies the cause in about ten seconds, and for most oversized PDFs the answer is the same.
The ten-second test
Open the PDF and try to select a sentence with your cursor, as though you were going to copy it.
If the text highlights, the document is born-digital — it was exported from Word, Pages, or a design tool, and its text is stored as text. A born-digital document of a few dozen pages is normally well under a megabyte.
If nothing highlights, or the whole page selects as one block, it is a scan. Every page is a photograph of paper. That is your answer, and everything below follows from it.
Scanned versus born-digital is the whole story
A born-digital page stores instructions: draw this glyph here, in this font, at this size. The whole page might be a few kilobytes, and the font is stored once for the entire document no matter how many pages use it.
A scanned page stores pixels. A single letter-size page at 300 dots per inch is roughly 2,550 by 3,300 pixels — about 8.4 million of them. In colour, before compression, that is over 25 MB for one page. Compression brings it down a long way, but the starting point is a different universe from a few kilobytes of text.
This is why page count is such a poor predictor. Ten scanned pages will beat five hundred pages of text, every time.
What actually drives size, in order
Most articles list causes without ranking them, which is unhelpful — the top two dwarf everything else so completely that the rest are barely worth your time.
- Scan resolution and colour mode. Both multiply. Doubling DPI roughly quadruples the pixels, and colour costs several times greyscale, which costs more than black and white. The US Bankruptcy Court for the Southern District of Indiana, which publishes filing guidance because oversized PDFs are a chronic problem there, tells filers to select black and white rather than greyscale or colour, and to lower quality to 300 dpi — those two settings, before anything else.
- Embedded images and graphics in a born-digital file. A report with photographs pasted in at full camera resolution carries every one of those pixels, even though it is displayed two inches wide.
- Embedded fonts. Real, but bounded: a font subset is tens to a few hundred kilobytes, stored once for the document.
- Metadata. Effectively nothing. Author names, timestamps, and revision history are measured in kilobytes against pages measured in megabytes. Advice to strip metadata to save space is misdirection.
Why compressing a text PDF can make it bigger
Browser-based compressors, ours included, work by rendering each page as an image and rebuilding the document from those images. On a scan that is a huge win, because the pages were images to begin with and they get re-encoded more efficiently. On a text document it is a catastrophe, because it converts compact drawing instructions into photographs of themselves.
We measured it. A 20-page text-only PDF of 74.7 KB, rendered at the settings an image-based compressor uses, became 1.9 MB at the lowest resolution and 4.9 MB at a higher one — 25 to 66 times the original, from a tool whose job is to make it smaller.
Our PDF Compressor checks for this and hands back your original untouched, marked "already optimized", rather than returning something larger. That is not the tool failing. It is the honest answer: a text-only PDF that is genuinely large is not large because of its text, so compression has nothing to work on. Look for an embedded image instead.
What to do, by cause
Match the fix to what the ten-second test told you.
| What you found | What actually works |
|---|---|
| A scan, and you still have the paper | Rescan in black and white at 300 dpi. This beats any compression applied afterwards, because it never creates the data in the first place. |
| A scan, and the paper is gone | Compress it. This is the case compression is built for — PDF Compressor rebuilds the pages at a lower quality you choose. |
| Born-digital with photographs | Re-export from the original document with images downsampled, if you still have it. Compressing the PDF works too, but costs you selectable text. |
| Born-digital, text only, still large | Something is embedded that you cannot see. Compression will not help. Check for a full-page background image or a pasted screenshot. |
| It only needs to be smaller for one upload | Send fewer pages. Split PDF extracts the pages that matter, which beats degrading all of them. |
| You only need the words, not the layout | PDF to Text pulls the text layer out as a .txt file — kilobytes instead of megabytes. Born-digital documents only; a scan has no text layer to extract. |
What you trade away
Every route above costs something, and it is better to choose the cost than discover it later.
Compression rebuilds pages as images, so text stops being selectable and searchable, and form fields are flattened. Keep the original for editing and send the compressed copy. Rescanning at lower quality is permanent for that copy — check small print and signatures are still legible before discarding anything. Splitting is lossless but changes what you are sending, which may not be acceptable for a filing.
And one thing none of them do: compression never makes a scan searchable. That needs OCR, which reads the picture and writes a text layer. Our tools do not do OCR, and a compressed scan is exactly as unsearchable as it was before.
When a limit is the real problem
Often the file is not too large in the abstract, just larger than one system will accept — and those limits are set per organisation, not by any universal rule. The Southern District of Indiana's bankruptcy court, for instance, caps electronic filings at 35.0 MB and tells filers to split anything larger into parts. A different district sets a different number.
So find the actual limit before optimising against a guess. For email attachments in particular the published number is not the whole story, because attachments are encoded in transit and grow on the way — getting a PDF under an email attachment limit covers what each provider really allows. If you have a specific target in mind, compressing to 100 KB, 200 KB, or 1 MB sets realistic expectations for what each is achievable from.
Common questions
- Why is my text-only PDF so big?
- Usually because it is not text-only. Check for a full-page background, a letterhead image repeated on every page, or a pasted screenshot. If you cannot select the text with your cursor, it is not a text PDF at all — it is a scan, whatever it looks like.
- What DPI should I scan at?
- For documents you need to read rather than reproduce, 300 dpi in black and white is the usual recommendation, and it is what the Southern District of Indiana's bankruptcy court tells filers to use. Go higher only for photographs or fine detail like signatures and small print.
- Why did compression make my file larger?
- Because the document was text, not images. Compression rebuilds pages as pictures, and a picture of text is far larger than the text itself — we measured a 74.7 KB text PDF ballooning to between 1.9 MB and 4.9 MB. Our compressor detects this and returns your original unchanged instead.
- Does deleting metadata reduce PDF size?
- Barely. Metadata is kilobytes; scanned pages are megabytes. If a file is oversized, the cause is on the pages, not in the document properties.
- Will compressing a scanned PDF make the text searchable?
- No. Compression only changes how the page images are stored. Making a scan searchable requires OCR, which reads the image and adds a text layer — our tools do not do that, and a compressed scan stays exactly as unsearchable as the original.
- Is it safe to compress a confidential document online?
- It depends on the tool. Most upload your file to a server and process it there. Ours rebuilds the PDF in your browser, so a contract or medical record never leaves your device — you can confirm that in your browser's network tab.