pdflatex, just like TeX, comes from an 8-bit past, so it will frequently resort to custom encodings to fit all the glyphs in a single 255-character font. That

pdflatex, just like TeX, comes from an 8-bit past, so it will frequently resort to custom encodings to fit all the glyphs in a single 255-character font. That means that any 8-bit "character" (in PDF bytes, what is really copy-pasteable or readable by TTS system) can actually stand for any displayable character. Perhaps XeTeX is better? (I am not intimately familiar with what pdf(la)tex does, but this is how TeX works, and I've seen those PDFs, esp with eg. Cyrillic, which on select show up garbled text)