We all have noticed Google’s personalized search feature which, like any Google feature/product, will be in beta for years to come (mostly to justify poor performance). At first glance it may appear a useful functionality, intended to make navigating an ever-increasing amount of indexed pages easier. However, similarities with PageRank’s notion and purpose exist: it is a vote.
While PageRank revolves around the idea of positive feedback – by one site linking to another – personalized search instead relies on user punishment. Users may vote out any website they deem irrelevant to their search term; the restriction can apply to the particular search phrase at hand, or the entire domain no matter the search term. This could have two potential effects on SERPs:
- The search results are cleaned of spam sites. Given enough out-votes, a domain may be pushed down in the results for any search term or may even be thrown out of the index altogether.
- Search results improve, because Google will be able to fine tune the semantic meaning of a search term.
Great, Google solved the problem of diluting search quality. But wait, let’s not be so quick to jump to this conclusion, because supplementing relevancy with personalized search relies on two important components :
- User participation
Ultimately the personalized search feature’s impact depends on whether users will actually use it. This might be a problem if a user’s searches are vastly distributed across a large number of topics – if I search once for a hardware store that sells fuses I may not care to help make the search results cleaner. However, if I am a frequent traveler and my searches are often clustered around travel I might have a much higher incentive to participate in the voting system. But once a user is a “clustered searcher” he is also experienced, i.e. his searches are refined enough to get good results in the first place!
- Truthful voting
Obviously, if users would start to randomly exclude sites from SERPs regardless of the relevancy, personalized search would be useless for improving broader search relevancy. More importantly, this behavior could actually result in lower quality of search results. Webmasters found a way to get around the PageRank idea by exchanging or selling links; personalized search leaves room for similar loopholes: imagine collective out-votes for particular sites sparked by coordination similar to Google-bombing. Or imagine a trading site where users get paid to exclude sites from their SERPs.
Google’s personalized search is nothing but a way to extract user feedback about a site; potentially it could impact SEO beyond just technical SEO, linking strategies and writing good content. But I have once written about a similar feedback mechanism fail miserably on eBay & the opportunistic behavior it can elicit; will Google’s mechanism work better? Perhaps I should embrace the new feature and help David be heard by excluding competing sites on Google for the search term “search engine marketing blog”…
Tags: search engine optimization, personalization, algorithmic search, collaborative filtering