I disagree on that count. I used several android based and windows mobile based voice searches and they usually came up with pretty much the same answers. Voice recognition on the part od Kinect 1 was also significant.
I think the Cortana assistant which will backend Kinect 2 and Bing will eclipse all earlier efforts in the market.
That's the problem. You are using it only for searches. ^_^
When Cortana arrives, Google Now and Siri would have advanced even further and with more integration done.
The speech tech itself is not so interesting, although the ability to mix multiple languages in 1 sentence would be awesome.
The thing that's interesting with speech driven interaction is the integration of multiple services, and the representation of the object models. In the legacy AppleEvent model, the user can address the contents of the apps. e.g., delete the word “bogus“ from the second paragraph; and chain multiple apps together. But it's too heavyweight.
In iOS7, it looks like they added motion context (e.g., is the user driving, walking, stationary ?), and more prevalent use of data detectors (e.g., is this segment of text an address, or time, or ... ?).
Traditional telephony services, UI navigation commands, and car kits are a lot more primitive.