Why The Wikipedia/Google Search Results Study Is Flawed

Many SEOs have been chatting this week about a recent study by Intelligent Positioning (first reported by Search Engine Watch) that showed Wikipedia ranks on Google UK for 99 percent of searches.

Yikes, right?

Wikipedia — more specifically, Google’s apparent love of Wikipedia — has long been a sore spot in the SEO industry, so seeing a statistic like that is a big pile of salt in the wounds at this point.

But it’s really not a statistic to get worked up about because, in my opinion, the study itself was flawed.

Study Methodology

As Intelligent Position explains, the company used a couple random noun generators to come up with a list of 1,000 nouns — words like “ashtray” and “volcano,” “snowflake” and “melody.” It then did 1,000 unique searches on Google UK and charted if and where Wikipedia showed up in the first page of results.

The results? Wikipedia was on page one for 99 percent of those searches, was the top-ranked result for 56 percent and was in the first five results for 96 percent of those noun searches.

Over on eConsultancy today, Kevin Gibbons makes the point that this shouldn’t be too surprising because Wikipedia does a lot of things right where SEO is concerned: usually very rich/deep content, highly targeted web pages, strong domain authority, loads of inbound links and more.

I don’t argue at all with those points, but I’d add something that seems just as obvious to me: the study only used one-word nouns. Wikipedia is an online encyclopedia — in large part, it’s a repository of information about nouns.

With all of that solid SEO in its favor, chances are great that searches for things like “tortoise,” “asphalt” and “liquid” are going to have Wikipedia very high in Google’s results. The Wikipedia page about tortoises is nothing short of amazing. Ditto the pages about asphalt and liquid.

A Suggestion For Different Methodology

Most search queries are longer than one word nouns. Chitika recently pegged queries at between 4.07 and 4.81 words on average, depending on the search engine. A couple months ago, Hitwise reported that 27 percent of searches that produced clicks were one word — leaving 73 percent of searches not represented in this study.

What I’d love to see someone do is this: Do a thousand searches (or more) that represent actual search engine behavior. Make 27 percent of those random searches be a single word (like “tortoise” or “liquid”); make 24 percent be two words (like “buy laptop” or “ankle pain”); make 19 percent be three words (like “u2 song lyrics” or “funny Valentine’s cards”), and so forth up to seven or eight words.

And then, using a variety of search terms that mimics actual search behavior, show how often Wikipedia appears in the first page of results. I’m pretty sure it’ll still be very high, but it won’t be 99 percent of the results.