At this point in its existence, Google Book Search seems of most use to academics in a limited range of pre-twentieth century fields--in fact, of most use to eighteenth- and nineteenth-century specialists. On the one hand, our books are out of copyright; on the other, they aren't so decrepit as to be unscannable. (However, see the caveat below.) Anyone who wants to work with a 1769 edition of Pope's Iliad can certainly do so. Of course, anyone working with medieval manuscripts or a Gutenberg Bible is, so far, out of luck, as is anyone interested in books with the bad manners to be still copyrighted.
While having access to all sorts of weird and (occasionally) wonderful things is a plus, the real advantage comes with the search function. Being able to find keyword strings across rare texts in multiple and far-flung libraries opens up all sorts of possibilities. Here, for example, are the search results for Rose Allin, whose torture at the hands of Edmund Tyrrel inspired both poetry and fiction. Now, it's worth pointing out that the search doesn't pull up either Anna Eliza Bray's The Protestant or Emily Sarah Holt's The King's Daughters, both of which
plagiarize appropriate John Foxe's account. Nevertheless, it's interesting to find Rose Allin referenced in the High Church magazine The Monthly Packet, as the higher reaches of the CofE were often ambivalent about Foxe. Of course, there's a danger here, and that's in assuming that every writer on Rose Allin actually spelled it "Rose Allin." And so, a second search for Rose Allen pulls up several additional references (mixed in, unfortunately, with a lot of irrelevant material). At this point, then, the search function certainly makes it easy to start a historical or comparative project, but anyone who relies on it to finish one will be in trouble. Plus the usual problems with garbage in, garbage out abound. Right now, I suspect that the book search is much more useful for people who already know something about a topic than for those who are just starting--largely because the former know just how much is missing from the search results.
If Google Books is going to be a real scholarly resource, though, it must pay closer attention to the annoying little details. I'm repeating myself here, but this is a genuine problem. I frequently come across books with scanning errors like missing pages or repeated pages, as well as blurred text; wood engravings, needless to say, don't always survive the trip successfully. Although we can now download full volumes of early periodicals, which is a terrific resource, there are still far too many utterly useless periodical uploads with incorrect or missing data (including inaccurate titles!). Finally, snippet view as it currently exists needs to be taken behind the woodshed and summarily executed. One recent "snip" pulled up...an entirely blank image! There's no way to evaluate the use value of any given excerpt unless we can see it.