Dispelling the Myths about the Myths about PageRank

Author: John Hughes

Date Published: 17 September 2007, 12:00

I just finished reading somebody's so-called expert opinion about PageRank in a high-readership SEO newsletter, which shall remain nameless. Their article was about PageRank Misconceptions, and included an entire section attempting to rebut the myths about PageRank.

The article was admirably honest and showed a real understanding of the facts and assumptions on which PageRank is based, but missed its most fundamental nuance.

This article, like so many I have read before talks about PageRank score like it is some kind of barrel of water – links in to a page leak water in, links out of the page leak water out. This model is simply wrong. It is appealing, I guess, in that it allows a reader to visualise a web page with PageRank flowing in and out, but it doesn't at all model the probability maths on which PageRank actually relies.

Imagine for a moment, that it is the links that have some kind of measurable score. The score for a Link is the PageRank of the page it is on, divided by the number of links on that page, and then reduced by a fixed "dampening" amount.

To put it simply, the Page Rank of your web page (ignore the 0 to 10 score on the Google Toolbar, that is a logarithmic simplification) is the sum of all the links pointing to it. Adding more links to your own page has no impact at all on the scores of the links pointing to it, and so no impact on the PageRank of your page itself.

However, there is a longer and more subtle effect on the PageRank of your own web page, as a number of the links on your page will be to other pages on your own website. If you add a link to a third party website, you will have increased the number of links on the page by one. In doing so, each link on the page is worth slightly less, as there are more to share the PageRank among when calculating the scores of the links themselves. This means that you pass slightly less PageRank to other pages of your own website than you did previously, reducing their PageRank. This in turn will affect the other pages that they link to, and (assuming there is a chain of links bringing us back to the original page) will affect the original page we were examining.

So, in summary, the important concepts to consider in PageRank terms are the scores of the links between pages, and the effects of PageRank on link-chains, and how these affect the PageRank score of web pages.

It may sound like I've just tried to refute any argument given in the original article, but in fact, other than its mathematical understanding, it was reasonably well written, it was factual, and it made sensible arguments.

Myths abouts PageRank

Here is a summary of what it got right:

It is a myth to simply say that the more links you get to a webpage the better. Correct - it is the quality (or score if you think of my mathematical description) of the links to your page which are important. A few high-scoring links will massively outweigh several hundreds or thousands of low-scoring ones.

It is a myth that PageRank is a score for web sites. Also correct. As I have said for many years, PageRank is a score for an individual web page. When people think of PageRank for a web site, they must normally be thinking of the Page Rank of the home page of a site. I say must be thinking of, as this is the only rational explanation of how someone might describe a website as having a PageRank score.

It is a myth that only links from third-party websites count. Again, this is right – PageRank makes no distinction between web pages, whether they are on a different domain of the same domain.

More than just PageRank

The point of the original article was that PageRank on its own is not the key to successful SEO. One important point that the original article did not discuss however, is that there is a lot more to links than just PageRank.

For example, although PageRank looks at individual pages, with no attempt to classify them into collections (or web sites), Google does classify web pages into collections, based on their semantic relevance. The Anchor Text of a link helps this classification, and consequently, the text of a link could have a significant bearing on the ranking of both pages at either end of the link, particularly of the page being linked to.

A consequence of this is that sections of a Web Site can be themed into separate classifications – for example, a financial services site could have a pensions section and a mortgages section, and careful structuring of the navigation could classify both section independently. Such classification is entirely independent of PageRank, but can have at least as significant an impact on a web site's search engine positions as any attempt to influence PageRank.