Heh! I call that “leveraging content”.

I’ve been doing that with my personal sites for years. :)

Matt, the link counts in Google Webmaster Tools don’t survive close scrutiny. I often find links in the Google search results that are not reported in GWT.

I have no explanation for that.

I’ll second that…

And now we’re getting down to some really important fundamentals! Although I think we could go a step further and define what ‘value’ is exactly. In my mind there is keyword relevance, and authority/PageRank. (Essentially what a page should rank for, and how high it should rank)

There could even be a link between the two whereby a high authority page is allowed to rank for a wider set of keywords than a comparable low authority page.

I can see that I need to set up a lot more tests.

Matt, I would guess the people who started all this hoopla wanted to know what it means when an old PageRank document says that “links (to the same destination) on a page are normalized” (or something like that).

It is generally assumed that (in the PageRank calculations) subsequent links to any destination (within one document) are ignored.

However, since we cannot track and measure PageRank, many people look at the passing of link anchor text as a means of determining if links pass value.

If Google (or any search engine) allows a second link to pass anchor text, does that mean the link is passing other types of value?

While the question might seem academic (perhaps even ludicrous) to most people, there are some folks in the industry who obsess over PageRank. Maybe it would provide them with an angle for channeling more PageRank if they could determine that secondary links pass PageRank.

On the other hand, even if it could be shown that any page which passes value through its links will pass anchor text through more than one link, that would not be sufficient to prove that you were getting 2 links’ worth of PageRank.

We can summarize the options this way:

1) All links (to the same destination) pass equal PageRank and anchor text (said not to be so in numerous PageRank documents)
2) All links (to the same destination) pass only one link’s worth of PageRank and anchor text — subsequent anchor text is discarded (most tests indicate this is probably what happens)
3) All links (to the same destination) pass only link’s worth of PageRank but all anchor text (at least some tests indicate this CAN happen)

The implications for link-focused SEO are profound if it can be shown that you obtain value from 2 or more links to the same destination on one page.

I suspect the search engineers understand these implications and they are probably not inclined to make link-based SEO any easier.

I believe that SOME Web sites (Microsoft calls them “stingy linkers”) probably have the (revocable) privilege of passing value to a destination through 2 or more links. I also believe most Web sites probably never earn that privilege or they may lose it quickly.

In my mind, these tests are really probing the Google trust algorithms, not the link weighting algorithms.

But I would be unable to devise any truly useful tests — at least not any I would be willing to share. The problem with sharing algorithm probing tests is that people tend to share them with their friends, and if there is one thing most people in the SEO community have, it’s friends.

Even *I* have friends in the SEO world. :)

It just doesn’t make sense. It also doesn’t make sense when I observe the effects of adding keyword rich text links to the main textual content of a page, even if they are 2nd and 3rd links. For me, this has always been a belt and braces SEO move and I have seen increases in rankings and traffic after I’ve done this.

My personal conclusions based on numerous observations over the years are…

1. Google is intuitive when it comes to calculating the weight of a link (i.e. it knows when a link is hidden away in the footer, when its part of a menu, when it appears within the text or when it’s a logo link etc).

2. We know Google weights off site links differently depending on how they appear in the code so this should be the case for internal links as well.

3. Sites with more “trust” are treated differently in this situation.

4. I’ve been playing around with a theory about how Google handles non keyword rich text links for a while as I’ve spotted a trend where generic pages (accessibility, terms, privacy etc) end up in the supplemental index regardless of how high their internal link profile is.

Plenty of sites link to these pages more than any other and often high up in the code, but these pages often get slung in the supplemental regardless. Does Google look for “buzz words” that indicate the quality of the page?

A question…

Could this simply be a case of Google caching and calculating the first link on the page first but ignoring the second one for the simple case of speed? Maybe if this test was left in place for a length of time the second link would also be come in to play?

p.s. My core SEO skill has always been English so i’ve made any glaring scientific errors, my apologies.

Matt

I make no apology for ‘just making claims’ – there is a caveat to the testing validity on the About page. However, the goal of this website is to find out what’s real and what’s regurgitated dogma so constructive challenges are always welcome.

The test was carried out on the root space of this domain before this blog was created, therefore it was overwritten and the screenshots/code excerpts used instead. The test resources no longer exist, but it’s my intention to use folders in the future to allow people to ‘look under the hood’ at their own convenience.

As per the comment in the HTML, the test was started on 19th March 2009. I waited until the end of July for indexation.

The parent page, /index.html, had just a handful of links from the domain’s previous life as a graphic designer’s portfolio. It was around a PR2 before I let it rot for a couple of years. At the point of this test was a PR0. The domain is about 6 years old.

You make a very good point, this microtest looks at internal link anchor text only. (I’ve amended the post to reflect this.) I’d be interested to see some external link behaviour and may set up some more microtests or full tests in the future.

I’ve got no ego about my work. If you can see a better way to test an algorithm factor, do email me and we can throw it back and forth until we’re both happy with the methodology!

Here’s my beef in a nutshell: no two sites are the same. I would argue (and probably most people in our industry agree) that a well-established, trusted site has more link leeway than newer sites.

SEOs who try to test the algorithm don’t factor those kinds of differences into their tests. Hence, the conclusions people draw after making these really limited tests tend to be more sweeping generalizations than anything else.

It’s important that people who are still learning the basics in SEO don’t just take my claims or anyone else’s claims at face value. All claims should be questioned and investigated, and if the claims cannot be validated through data or replication, they should be discounted.

It’s good that people want to keep the discussion going (good, in my opinion, because search engine algorithms change all the time). But I feel the SEO community has not yet become comfortable with distinguishing between A-level and B-level sites.

Not to take anything away from your site, as your own test seems to have only looked at passing internal link anchor text (and I believe the original claim concerned links between different sites — certainly the impromptu test I ran was using two different sites).

And that will be my last word on your rather interesting post. :)

Your report is no more valid than Rand’s. Without showing us the sites and letting us look under the hood, you’re just making claims.