I don't currently listen to "podcasts" (and am still not sure how they differ from previous web audio streams, except for the In name)--much as I don't listen to audiobooks.
And that's me.
I also would be unlikely to start doing podcasts, because speaking isn't normally a way I organize what I want to say (when I do a speech, there's almost always a full-text written version, even if I vary from it a lot).
And that's also me.
One early library-related podcaster has/had a blog that lapsed into inactivity. He's now doing a stream of podcasts. Apparently, talking through what he has to say is more natural for him than putting together blog entries or written journals.
And that's him.
Some librarians are excited about podcasts, both because they find audio speech a good way to take in information and because they believe it might be another way for libraries to spread the word.
And that's them.
If you're looking for an attack on podcasts, you've come to the wrong journal. Different people have different preferred learning styles or, for that matter, taking-in-entertainment styles. Different people have different preferred creation/organization styles. This is a case where "YMMV" becomes the heading I've used once or twice in C&I: "The way we're wired."
I've wanted to try speech-recognition software--but realized that I'm more likely to sit down and write through something than I am to sit down and talk through it (when it's something that belongs in print, that is). That's my style.
Of course, if we had highly accurate multivoice speech recognition software and, conversely, human-sounding text-to-speech software (which we may have, for all I know), people could mix-and-match to suit their own preferences: I could read these podcasts as text, and people could listen to Cites & Insights (which I believe they can anyway: I certainly don't disable TTS in the PDF files, although as a dumb XP user I also don't get how to start TTS).
Anyway, I think this falls into "to each their own"--and, to be sure, accessibility. My preference for text over speech as a source method is just that: Mine. I do not claim universality.
If podcasts work for you and yours, great.