What is actually inside a PDF
Almost every surprising thing a PDF does — a redaction that fails, a file that will not shrink, text you cannot select — comes from one fact about its structure.
A PDF is a set of drawing instructions
Not a picture of a page, whatever it looks like. It is a list of instructions: draw this text in this font at this position, place this image in this rectangle, stroke this line. A reader just follows the list, and the page appears.
That is why a PDF looks identical everywhere, and why a few things that seem contradictory turn out to both be true at once.
Text and images are separate layers
A scanned page is, underneath everything, an image. If a scanner ran OCR on it, the recognised text gets added as a separate invisible layer sitting over the picture. So the page is a photograph, and it is also searchable, and both of those are true at the same time.
This explains the most dangerous mistake people make with a PDF, and I care about this one: drawing a black box over a number just adds another instruction on top of the text. The original text instruction is still sitting in the list. Anyone can select it, search it, or simply delete the box on top.
Annotations are not the page
Highlights, comments, signatures and shapes all get stored as annotation objects layered over the content, not merged into it. That is why they can be moved or deleted later, and why "flattening" exists as its own distinct operation — the one that actually bakes them in for good.
If you sign a document and email it, the recipient can usually still drag your signature off the page — unless it was flattened.
Metadata travels with the file
A PDF quietly carries its author, its creation and modification dates, the software that produced it, and sometimes the original filename and full path. Fine while it stays internal, awkward when the path is /Users/yourname/Clients/CompetitorName/. Worth checking before anything leaves.
Why the file is so large
Almost always the images. A single scanned page at 300dpi in colour already runs several megabytes; twenty of them and you have a file nobody can email. Scan in greyscale, or drop to 200dpi for ordinary text, and you typically cut it by three quarters with no difference worth noticing.
Fonts matter less, but are not free either — a fully embedded font set adds hundreds of kilobytes, which is why a text-only PDF is rarely as small as you would expect.
What this means in practice
- To actually remove content, it has to be deleted from the instruction list, or the page re-rendered without it. Covering it up never works.
- To make a signature permanent, flatten it.
- To shrink a file, start with the images.
- To check any of this yourself, open the exported file and search it.
Chartra — Scan, sign and redact documents. See what it does →