Why Are PDFs So Large? Understanding PDF File Sizes

How PDF Files Are Structured

The PDF format was created by Adobe in 1993 and became an open standard (ISO 32000) in 2008. It was designed with one primary goal: to ensure a document looks identical regardless of what device, operating system, or software displays it. To achieve this, a PDF must be entirely self-contained — it carries everything needed to render the document.

At a structural level, a PDF consists of a header, a body (containing objects like images, fonts, and page content), a cross-reference table (an index of all objects), and a trailer. The body is where the size lives: each embedded resource — an image, a font, a colour profile, an annotation — adds to the total.

Unlike a Word document, which references system fonts and links to embedded objects at relatively compact sizes, a PDF must be a closed universe. This self-containment is its greatest strength (portability) and its greatest weakness (file size).

The Biggest Culprit: Embedded Images

In the vast majority of large PDFs, images are the primary contributor to file size. Here is why:

Resolution Mismatch

Professional-grade images are often captured or created at 300–600 dpi (dots per inch), which is the standard required for high-quality print output. When these images are embedded in a PDF, the full resolution data is preserved. A single 300 dpi photograph at 15cm × 10cm requires approximately 4 MB of uncompressed data. A brochure with ten such images could carry 40 MB of image data before compression.

For a document that will only ever be read on screen, 72–96 dpi is entirely sufficient. That same photograph at 96 dpi is only about 400 KB — a 90% reduction. Compression tools reduce PDF size primarily by downsampling these high-resolution images to screen-appropriate resolutions.

Scanned Documents

Scanned PDFs are particularly large because every page is stored as a full-page image rather than as structured text and vectors. A scanned A4 page at 300 dpi produces an image of roughly 2480 × 3508 pixels — about 8.7 megapixels. Twenty such pages create a file that is enormous before any other factors are considered.

Uncompressed Image Streams

Some PDF creation tools embed images using lossless or uncompressed encoding, even when lossy compression (like JPEG) would produce visually identical output at a fraction of the size. This is particularly common with PDFs exported from CAD software, some older versions of Microsoft Office, and some scan-to-PDF workflows.

Font Embedding and Subsetting

PDF's portability guarantee requires that fonts be embedded in the file. If a font is not embedded, the viewer will substitute a different font, potentially changing the document's appearance entirely. Font embedding is therefore usually a feature, not a bug — but it adds weight.

A typical OpenType font file is 150–500 KB. A document using a regular, bold, italic, and bold-italic weight of the same typeface carries four separate font files — potentially 2 MB of font data before any text or images are considered. A document using multiple typefaces can easily accumulate 5–10 MB in font data alone.

Font subsetting addresses this by embedding only the specific characters (glyphs) used in the document, rather than the entire font. If your document uses 250 of the 900 glyphs in a font, subsetting reduces the embedded font data by roughly 72%. Most modern PDF creation tools enable subsetting by default, but some do not.

Metadata and Hidden Data

Beyond visible content, PDFs can contain substantial hidden data:

• Document metadata: Author, title, creation date, modification history, software used. Minor contributors to file size but worth stripping for privacy.
• Revision history: PDF supports incremental updates — each edit adds a new "revision" to the file rather than replacing the old content. A heavily edited document can contain many previous versions, all stored in the file.
• Comments and annotations: Review comments, sticky notes, and markup added during collaboration.
• Colour profiles: ICC colour profiles embedded for print accuracy can add 500 KB–2 MB to a file.
• Thumbnails: Some PDF creators embed page thumbnails for fast preview generation in document management systems.

Removing unnecessary metadata and revision history is a lossless operation — it reduces file size without affecting the visible content in any way.

How PDF Compression Works

PDF compression is not a single operation but a collection of techniques applied to different parts of the file:

Image Resampling

The most impactful technique. Images are downsampled from their original resolution to a lower one (commonly 72–150 dpi for screen use) and then re-encoded using JPEG compression. The JPEG quality setting balances file size against visible quality. At quality 80 (out of 100), JPEG compression is nearly imperceptible to most viewers while reducing image data by 80–90%.

Flate (zlib/DEFLATE) Compression

Text streams, graphics commands, and other non-image content in PDFs are compressed using Flate compression — the same algorithm used in ZIP files. This is lossless and typically reduces text and vector content by 50–80%.

Object Optimisation

Compression tools also reorganise the PDF's internal structure, removing duplicate objects, optimising the cross-reference table, and eliminating dead references. This structural housekeeping can reduce file size by 5–20% without affecting content.

Understanding these mechanisms explains why compression works differently on different PDFs: an image-heavy PDF compresses dramatically, a text-only PDF compresses moderately, and a PDF that was already well-compressed may barely shrink at all.