Because it was designed as a graphical output format, not an editable format.

Some of the "compression" tricks it allows one to use (eg. font subsetting, even remapping characters to use fewer bits to encode text) may make the data only keep the same appearance, and semantic encoding would be gone (for example, "A" may stand for "#").

It's actually quite similar in nature to TeX's DVI format (boxes and their positions), though obviously not a bitmap format but a vector one with all the deps embedded.

This means that, for instance, using non-default kerning and whitespace will lead to all text becoming box-per-character thrown around the page.