Google bug reveals favored web sites
A couple months ago I shared one of my Google secrets, since that secret no longer worked. ;-) Specifically, it was how to obtain a list of the most important web sites according to Google.
Now, surprisingly, this little trick appears to work again (it stopped working in 2003), thanks to a bug introduced into Google's algorithm. Two months ago, a search for http would have revealed results like HTTP - Hypertext Transfer Protocol Overview and Welcome! - The Apache HTTP Server Project. Today, these sites appear nowhere near the top of the results. Instead, the top results are occupied by a "who's who" list of highly important web sites — sites that don't include the word http anywhere in the text of the page.
As already noted by blogger Nathan Weinberg, this same phenonemon occurs when you search for www.
One thing I found curious is that http and www Google queries return different results. Now these results are NOT in order of PageRank score, at least not the PageRank scores as revealed by the Google Toolbar. You can verify this to be the case yourself simply by using SEO Chat's PageRank Search tool. Indeed, it's a well-known fact within the SEO community that the PageRank scores served up by the Google Toolbar servers are not the actual PageRanks used by Google in the ranking algorithm. PageRank debate aside, perhaps this list offers us a (now) rare glimpse at some of Google's Chosen Ones — the most important sites on the Internet according to Google.
What makes me say this is due to a bug in Google? For one thing, these results are NOT relevant to the search query. Secondly, I've uncovered another bug newly introduced into Google's algorithm, namely that the inurl: query operator does not work properly, and I think these two bugs might be related. For an example of this second bug in action, search Google for site:blogs.msdn.com scoble inurl:msnsearch and the top search result is currently blogs.msdn.com/mikehall/archive/2004/11/10/255417.aspx. Note there's no msnsearch in that URL!
I've compiled a list the top 1000 results for each of the two queries for your convenience. You'll see, they do vary quite dramatically:
Follow up:
501-1000 >>
3 comments
-
It's not a bug per se -- actually, Google just treats the word "http" like any other word. And a lot of people link to another site using "http://www.example.com", which in the eyes of Google is the same as the link text "http www example com", so all these words are "attached" to that page. So naturally the same "top 100" list happens when you enter "www" or "com" into Google. You just get very popular pages because the "http://www. ..." link type is so common across all fields and web sites it basically works as a "neutral" term.
Comment by Philipp Lenssen [Visitor]
· http://blog.outer-court.com —
01/09/05 @ 09:36
-
Ah, but why then do the SERPs for inanchor:http vary from the SERPs for http? If it weren't a bug, the SERPs for the 2 queries should be near identical. Unfortunately I don't have a screenshot of a http search from last year, but believe me it was markedly different from what it is now. Two months ago, I could search for http and get relevant results back in the top 10 — relevant to the hypertext transfer protocol. Currently that's no longer the case. Now I would need to search for intitle:http for relevant results.
Comment by Stephan Spencer [Visitor]
· http://www.stephanspencer.com —
01/10/05 @ 05:24
-
"but why then do the SERPs for inanchor:http vary from the SERPs for http? If it weren’t a bug, the SERPs for the 2 queries should be near identical."
They are in fact near identical.
I still can't see a bug. You may prefer last year's results, possibly they were different because now Google favors link text even stronger than before. In other words it would be incredibly hard to target the keywords "http" and "www" with any meaningful strategy, which would be a bad side-effect but not really a bug...Comment by Philipp Lenssen [Visitor]
· http://blog.outer-court.com —
01/11/05 @ 07:35
Comment feed for this post