for example, just type 'degree ' (make sure there's a space - you get results for "degree symbol" (or at least I do) - now all i have to do is rank for "degree symbol" and put links to my online degree affiliate links...
there are certainly keyword results that are a) in autocomplete, b) substrings of keyword results also in autocomplete, c) easier to rank for than the longer phrase.
I think topic-modeling (a la the Latent Dirichlet Allocation article posted about seomoz the other day) will make this difficult to do (at least without confusing the people you want to convert).
Actually there is a big section of the SEO world that focuses on the long tail keywords. This change is going to harm those folks, but up the reward for anyone that does rank highly for the frequently used terms. Good and bad, hopefully this will lead to less spammy junk, but also reduces the amount of real estate available for new entrants (or at least makes it more pricey).
In the live webcast they said that they use auto-complete to populate the instant results. Typing "online deg" auto-completes to "online degrees", so that's the query used to generate the instant results.
"online deg" probably isn't the best example, but there are still going to be cases where a prefix autocompletes to a different target than the final intended search, that SEOs for the final intended search will want to start ranking more highly for. Finding those cases will be part of the fun.
Ask it another way: do you think a service that shows what the current google autocompletions are for every prefix of a given term becomes more or less interesting to SEOs after this announcement?
I think what ruby above is saying is that with enough websites, urls and content, an SEO attack could turn 'online deg' into a valid keyword.
Google auto-completes brands and keywords that only have 2 or 3 website references (I know of a site in stealth and I just tried it, and it worked. Google has indexed their homepage and title, and as I typed it in it auto-completed their name, with only a single website to reference).
"an SEO attack could turn 'online deg' into a valid keyword."
If you have that much power, you probably have a better way to spend it. Also I would bet that suggest keywords come from user searches themeselves, not the web. That's why I suspect the impact of this is probably relatively minimal.
Call me crazy, but I think Google just might have considered the SEO impact before deploying this. If it's not literally on a checklist somewhere for deployment of search features, I'd be stunned.
there are certainly keyword results that are a) in autocomplete, b) substrings of keyword results also in autocomplete, c) easier to rank for than the longer phrase.