If you're going to make a reductionist argument like this, then consider the byte size of a text transcript versus the byte size of an audio recording of the same speech act. The latter will be orders of magnitude larger. Thus it strictly contains much more information.