I don't know what you have been considering doing with testing, but off the top of my head I don't know of anyone doing content A/B testing (ie: which version of content is more effective). I would expect to see a more granular, dynamic approach in website verbiage as NLP matures. For example, at my day job they argue semantics endlessly. Tailoring a websites' grammar to the user could increase effectiveness. This is an area I consider fascinating and wouldn't mind chatting more about it off line if you like.
