- Not everybody can speak
- Not everybody can speak clearly enough for voice recognition to work well
- Not everybody wants to enter text in a environment quiet enough for voice recognition to work well
- Speech input is clumsy at handling homonyms, accents, formatting, indentation, and multi-key keyboard shortcuts, among other things
- Not everybody works in an environment where speaking out loud is possible
- The "Up to 99%" accuracy quoted one of those articles actually isn't very good. Especially since the word-errors that are produced by voice recognition can be harder to spot, and clumsier to correct, than the character based typos from keyboard input
…and probably some more.