Why is my PDF so large?

Where the bytes are

We opened the PDFs of our tests and added up the bytes of each kind of content. In the scanned book page, the one picture is 98% of the file. In a PDF made of photos it is 100%. These files are big because their pictures are big, and nothing else matters much.

The scanned NASA report has 150 pages, each one a black and white picture. The pictures are 79% of it, and the rest is the structure that holds that many pages together. The IRS form has no pictures at all: its 13 embedded fonts are 13% of the file, and the form fields and page text make up the other 87%.

The lecture notes mix everything: 19 pictures are 65% of the file, 46 embedded font programs are another 12%, and the 117 pages of text and drawings are the rest. The table under "What we measured" has every file.

Pictures: the weight of our heaviest PDFs

A photo straight from a phone has millions of pixels, far more than a page needs on a screen or even in print. When such a photo goes into a PDF at full size, the PDF carries every pixel. A scan does the same: scanning at a high resolution, in color, makes each page a large picture.

This is where the Recommended and Strong modes of Compress PDF can help, because pictures can lose detail nobody sees. A PDF with no pictures can only get lighter by a rewrite of its structure, and how much that gave on our files is in the next section.

Fonts, forms and structure

A PDF carries the fonts it needs, so it looks the same on any computer. A document set in many typefaces, or one that embeds whole fonts instead of only the letters it uses, gets heavier for it. Forms add their own weight: every field is an object, and some forms carry a second description of themselves for older software.

None of this can be squeezed the way a photo can. A lossless rewrite packs the structure more tightly and drops what is repeated, and what that gives depends on how the file was written. Our long text-only PDF, which our test script wrote without packing anything, lost 38% of its size to it. The lecture notes, a real document, lost only 12.0 KB of 5.3 MB.

What each mode of Compress PDF did

Lossless rewrites the file without touching any picture. It took 14% off the scanned report and 11% off the tax form. On the PDF of photos, whose weight is in its pictures, it saved only 827 bytes of 2.9 MB.

Recommended also recompresses the color and gray photos inside, at no more than the resolution a page needs. On the PDF of photos it took 67% off, and on the scanned book page 31%. On the scanned report it gave the same result as Lossless, because black and white scans are stored in a way this mode does not recompress.

Strong turns every page into a picture. On the PDF of photos it took 88% off, and on the scanned book page 68%. It has a cost: the text can no longer be selected or searched, and form fields, links and bookmarks are lost.

When the file comes back the same

Compress PDF never hands back a file bigger than the one you gave it. When a mode cannot make the file smaller, you get the original back, byte for byte. Of the 10 files we compressed, Lossless gave one back unchanged and Recommended one: a PDF that was already as small as the rewrite makes it.

Strong gave seven of the ten back unchanged. Text pages drawn as pictures are heavier than the text itself, and a black and white scan stored as a picture of a page is already smaller than a photo of the same page. In our tests, Strong did better than Recommended only on the PDF of photos and the scanned book page.

If it is still too big

If the PDF is mostly text and still over the limit after Lossless, split it into parts and send them separately, or remove the pages that are not needed. If it is mostly photos, compress the photos before they go into the PDF, at a size that fits the page.

For a scan you are about to make, a lower resolution and gray instead of color make each page a smaller picture from the start. And when the file has to stay whole, a link to a shared copy can go where an attachment cannot.

What we measured

First, what each kind of content weighs in six of our test PDFs. Then each mode of Compress PDF on the 10 PDFs of our tests the tool takes, with the share of the original size that is left.

What takes the space

Compress PDF, mode by mode

Measured on November 2, 2026. The scanned pages, the NASA report, the IRS form and the lecture notes are real documents; the other test PDFs were written by our own test script. What takes the space was read with pdf-lib, the library the engine opens files with. Lossless and Recommended ran in Node with the engine of this site, which gives the same bytes as the browser; Strong draws the pages on the canvas, so it was measured through the Compress PDF page in Chromium 156 and in WebKit 27.2 on Windows. The table shows Chromium.

Sources

• Gmail Help: Send attachments with your Gmail message. Read on November 1, 2026.
• Microsoft Support: Sending limits in Outlook.com. Read on November 1, 2026.
• Adobe: Why is my PDF so big?. Read on November 1, 2026.