There were a lot of memorable moments in the 1939 classic, "The Wizard of Oz". My favorite is when Dorothy first lands in Oz and the movie suddenly shifts from black and white to color. However, there was another moment that has proven to be more significant on many levels; when we find out the wizard is really just an old fragile man hiding behind a grand facade.
Recently, I've begun to wonder if much like that famed wizard, Google's algorithm is more smoke and mirrors than hard substance.
While tinkering around with Google's various search commands during the Thanksgiving "break" (no such thing for SEO) I came across a startling pattern. It seemed as if every natural search engine result I pulled up was a mirror image of its corresponding allinanchor result.
Most enterprising webmasters are fully aware of the importance of
Google's allinanchor command as it relates to understanding search
engine rankings, but just in case you need a primer, here goes:
The
allinanchor command essentially shows what urls have the most inbound
links containing anchor text for a specific search term. To test it
out, simply visit Google and type in "allinanchor:" followed by your
chosen search term.
Allinanchor tutorials aside, the strikingly identical result sets stopped me in my tracks. Why were the allinanchor results virtually identical to natural search queries for the exact same search terms? At first, I figured that allinanchor was simply broken. After all, notable SEO voices like randfish have noticed issues with the command over the years. It's also not a good sign when Google's own help page on advanced search operators does not list either inanchor or allinanchor.
However, a closer look at the command revealed that some of the telltale signs of malfunction where missing. First and foremost, the result sets for regular search and allinanchor are not completely identical. Secondly, the amount of results for each is not identical. Lastly, there seem to be plenty of well-respected SEO folks that still refer to allinanchor and believe that it is still a valid command within Google.
And if it is, then Google's algorithm is much simpler than I previously believed.
I tested a variety of search terms, ranging from single words to three-word phrases and across various genres including news, products, celebrities, and SEO related topics.
The results are striking to say the least:
Keyword: NFL
Results: In Google, top 10 rankings and top 10 for allinanchor are perfectly matched
*allinanchor
trumped factors such as indexed pages, page backlink totals, domain
backlink totals, keyword density (no surprise) and domain age, which
are considered to be fundamental ranking indicators.
*Yahoo results had six top 10 matches, and only two were exact
Keyword: Madonna
Results: top 10 rankings and top 10 for allinanchor are perfectly matched
*allinanchor trumped all other aforementioned key factors
*Yahoo results has four top 10 matches, and only three were exact
Keyword: ipod
Results: top 10 rankings and top 10 for allinanchor are perfectly matched
*allinanchor trumped all other aforementioned key factors
*Yahoo results had five top 10 matches, and only one was exact
Keyword: SEO
Results: top 10 rankings and top 10 for allinanchor are perfectly matched
*allinanchor trumped all other aforementioned key factors
*Yahoo results had five top 10 matches, and only one was exact
Keyword: world news
Results: top 10 rankings and top 10 for allinanchor are perfectly matched
*allinanchor trumped all other aforementioned key factors
*Yahoo results has three matches, and zero were exact
Keyword: Britney Spears
Results: top 10 rankings and top 10 for allinanchor are perfectly matched (note: Google News was No. 1 so that's thrown out)
*allinanchor trumped all other aforementioned key factors
*Yahoo results had five top 10 matches, and only two were exact
Keyword: Nintendo Wii
Results: top 10 rankings and top 10 for allinanchor are perfectly matched (note: Google News was No. 1 so that's thrown out)
*allinanchor trumped all other aforementioned key factors
*Yahoo results had four top 10 matches, and only one was exact
Keyword: paid links
Results: top 9 of top 10
rankings correspond with allinanchor, and 4 were perfectly matched (one
was not in top 10 and the other five were one off)
*allinanchor trumped all other aforementioned key factors
*Yahoo results had one top 10 match, and it was not exact
Keyword: 2007 presidential campaign
Results: top 9 of top 10 rankings correspond with allinanchor, and 9 were perfectly matched
*allinanchor trumped all other aforementioned key factors
*Yahoo results had four top 10 matches, and only one was exact (no. 1 spot)
Keyword: Samuel L. Jackson
Results: top 10 rankings and top 10 for allinanchor are perfectly matched
*allinanchor trumped all other aforementioned key factors
*Yahoo results had seven top 10 matches, and only one was exact
Keyword: iomega zip drive
Results: top 8 of top 10 rankings correspond with allinanchor, and 8 were perfectly matched
*allinanchor trumped all other aforementioned key factors
*Yahoo results had four top 10 matches, and three were exact
Keyword: social media optimization
Results: top 10 rankings and top 10 for allinanchor are perfectly matched
*allinanchor trumped all other aforementioned key factors
*Yahoo results had four top 10 matches, and two were exact
Note: I want to take a moment to mention the folks over at webuildpages.com for providing a tremendously useful tool that allowed me to search for all of the aforementioned factors all at once. They've got other great tools over there as well.
So what did I take away from this little exercise in data mining? It's simple. If Google's allinanchor command still works, then it means that the amount of anchor text containing a specific search term is by far the most influential factor in determining Google search engine results, trumping virtually all other SEO factors. In layman's terms, anchor text is king.
Scary, huh? Although, to be fair, even if Google's algo is this simplistic, it still somehow provides what I believe to be by far the most relevant and useful results of any engine.
Sadly, I'm not totally convinced that the allinanchor command is working, so I encourage all who read this to chime in with anecdotes and evidence that either refutes or confirms its validity. I also encourage you to find examples of search engine results where allinanchor does not necessarily predict ranking order.
After all, the power and insight of the community almost always trumps that of the individual, and I would really like to get to the bottom of this potential epiphany.
I thought I had commented on this before, but I guess not.
Excellent findings! The lack of much differences between allinanchor results and organic results is big.
And G hiding and fooling with the operator makes it look even more fishy.
Posted by: Marc | March 31, 2008 at 02:59 PM