Back in 94, when I first started using speech recognition (Dragon Dictate, word-at-a-time recognition) I had a curious experience. I could feel the difference between speaking spoken speech and speaking written speech. Now I've been using speech recognition for so long, I now swap back and forth relatively easily. This ability to switch how I speak comes in handy when writing dialogue or a speech.

The end result is when speaking, I use a slightly more formal version of speech, and that helps me organize my thoughts on the fly in conversation. When speaking written speech, I speak more formally than I do in conversation, but the pacing and pattern is different. I think ahead of what I want to say and then try out 2 or 3 different forms before I actually say what I want into the recognition environment. Then I let Grammarly tell me how I'm an uneducated hick that don't speak good.