The PDF file format: a 2026 primer

PDF has been around for 35 years and is still the default document format for contracts, invoices, e-books, government filings, and academic papers. Most people treat it as a black box. Here's what's actually inside one.

A short history

PDF was created by John Warnock, co-founder of Adobe, in 1991. The internal codename was 'Project Camelot.' Adobe released the PDF specification for free in 1993, which is unusual for a company that could have kept it proprietary — and that decision is the reason PDF won as a standard. In 2008, PDF was standardized internationally as ISO 32000-1. The current version, PDF 2.0, is ISO 32000-2:2020.

Source: ISO 32000-2:2020 and Adobe's own PDF history page. Cross-referenced on 11 June 2026.

What's inside a PDF

A PDF is a structured collection of objects. The structure follows a typed object model sometimes called COS (Carousel Object System, in homage to the Project Camelot origin). There are nine basic object types:

1. Boolean — true / false
2. Number — integer or real
3. String — wrapped in parentheses, can be hex-encoded
4. Name — preceded by a slash, like /Title
5. Array — ordered list of objects
6. Dictionary — unordered key/value map (the workhorse of the format)
7. Stream — a dictionary followed by a length-prefixed byte sequence (used for page contents, images, embedded fonts)
8. Null — a single object
9. Indirect object — an object with a unique object number, addressable from anywhere else in the file

The first line of a PDF is a header like %PDF-2.0 that declares the version. The last line is %%EOF. Between them, you have a body of indirect objects and a cross-reference table that lets a reader locate any object by number without scanning the file linearly.

Versions in practice

The version number in the header doesn't dictate features — readers are supposed to inspect actual object types. But the version does hint at what's likely inside. A short tour:

The 14 standard fonts

PDF 1.0 defined 14 fonts that every conformant reader must support without embedding: Times-Roman, Times-Bold, Times-Italic, Times-BoldItalic, Helvetica, Helvetica-Bold, Helvetica-Oblique, Helvetica-BoldOblique, Courier, Courier-Bold, Courier-Oblique, Courier-BoldOblique, Symbol, and ZapfDingbats. If a PDF uses only these, it can be tiny — no embedded font data needed.

Encryption

PDF 2.0 defines 256-bit AES encryption as the recommended cipher. Earlier versions used 40-bit and 128-bit RC4 (deprecated) and 128-bit AES. When you password-protect a PDF, both a user password (required to open) and an owner password (required to change permissions like print/edit) are supported. The encryption applies to the document body but not the cross-reference table or header, which is a small leak but in practice not exploitable.

Subsets you should know about

ISO and other bodies have defined PDF subsets for specific use cases. The ones that matter in 2026:

• PDF/A — archival. Forbid external references, require embedded fonts, ban JavaScript. Required for legal and government archives in many jurisdictions.
• PDF/UA — universal accessibility. Require tagged content, structure trees, and reading order. The basis for WCAG-aligned document workflows.
• PDF/X — print production. Strict color management (CMYK, ICC profiles), high-resolution images, no transparency.
• PDF/E — engineering. For 3D and interactive machinery documentation.
• PDF/VT — variable data transactional. For batch-produced documents like statements and bills.

Why this matters for tools

When a tool like GetPDFPro merges, splits, or compresses a PDF, it's manipulating this object graph — concatenating page trees, rewriting the cross-reference table, optionally recompressing image streams. Tools that do this well preserve structure; tools that don't fall back to rasterization, which loses everything that makes PDF a structured format rather than a stack of images.

Practical takeaway: when choosing a PDF tool, check whether the output still has selectable text and a working outline. If it doesn't, the tool rasterized your document, and you've lost searchability, accessibility, and reusability.