Marrisa Mayer has a new title at Google:  head of consumer products.  Which is a little vague, if you ask me, but most major company job titles are that way.  At the LeWeb 2010 conference in Paris, she sat down on stage for an interview with Techcrunch founder Michael Arrington.  And he got some pretty interesting stuff out of her.  

The big reveal is that Google is actively working on a kind of predictive search that pulls in your search and browsing history, as well as your current location, to deliver search results before you’ve even searched for anything.  She calls it “contextual discovery.”  They’re already testing and experimenting with both the feature and its possible design and layout, and Mayer says we can expect to see the it rolled out within the next year or two.

I get the sense in reading the transcript of the interview that this will be an add-on to the standard search experience… something that you can opt out of if you want.  Which makes sense.  Contextual discovery can never hope to be 100% accurate.  For example, this morning I talked with a friend who had recently seen the movie Winter’s Bone and thought I would enjoy it.  So when I went online a few minutes later, I went to Google and typed in “Winter’s Bone” in order to learn more about the film.  Now… there is simply no way that my search behavior, browsing history, and location information would ever lead Google to predict that search need, because it stemmed from an offline, real-world conversation.

But with that being said, some of the possible uses tossed out in the Techcrunch article sound terribly useful and a little awesome.  Say you’re headed into a restaurant you’ve never patronized before.  You might be able to pull out your smart phone and have the menu already waiting for you because Google can tell where you are.  And let’s just say that in addition to that menu, there’s an added layer of information regarding the restaurant and its food based on what your social networking friends have shared about their experiences–Ted says get the chicken, but Mary thinks the burger is best.

The money quote from Mayer is this:

“The idea is to push information to people. It’s location in context. Inside the browser and a toolbar, can we look at where people have been going on the web — then we deliver it.”

Man, that sounds pretty cool if you ask me.  I’m generally all for Google tinkering and improving their service.  But I think the mobile version of this contextual discovery is going to have a lot more practical applications than a browser version–simply because of its ability to use location data to more accurately guess what kind of information you’re after.  But there are clear limitations across the board, mostly because computers can’t literally read human minds… at least not yet.

The big question is this:  How is this going to impact SEO?  How can I predict the keywords my prospective customers are going to use if they never get the chance to even type them?  Furthermore, how will this impact keyword research down the road; if fewer and fewer people need to search to find something, then there’s going to be a drop in useful historical keyword data.  That could be a challenge.

But of course, SEO will still be a necessary element of your marketing plan.  Because predictive search is never going to completely replace text-based search.  And even if it does, Google’s index will still rank the sites it knows about using their algorithm, which will still rely on on-page and meta tag information to properly understand each site.  Human beings are complex and spontaneous creatures, and we don’t typically like being told what we’re thinking.  Despite a few really cool potential applications for contextual discovery, I don’t see any way for it to ever be consistently effective for general search queries.  At least… not until computers can actually read our minds.  At that point… SEO might start to fade away, but I have this feeling that we’re still a good ways off from that kind of technology.