OK, so not so long ago Rand put out a post saying that Google will count only one link to a URL from any given page. This is a phenomenon that was originally noticed by Michael VanDeMar but many an eyebrow was risen at both posts. That particular facial hair elevation trend has seemingly culminated by David Eaves’ post which supposedly debunks the theory. However, some problems were found in the way David has conducted his research (starting with the fact that it was done in google.co.uk which seems to behave like Yahoo did 10 years ago). Since we love to solve dilemmas around here and since argument from personal incredulity is not terribly popular in my neighborhood, I decided to conduct a separate test which will cover most of the questions raised in and around the above mentioned posts. (BTW the initial discussion is covered at SERoundtable, as usual)
So, in order to test the theory, i picked two sites that sometimes double as my furry lab animals and set them up so that site A links to site B with two links using different anchor texts. The phrases appeared only on a site A, they were not to be found anywhere on site B and if that site was to rank for any of those two phrases, it would be only due to the anchor text of the links pointing from the site A.
The testing process went like this:
- Setup the sites like described above and wait for Google indexing. After the links are indexed, check the rankings of the site B for the two phrases.Result: Site B ranked for the first phrase and not for the second phrase. The below image describes the situation:
- Switch between the position of links to site B. Now the previously second phrase appears above the previously first phrase on site A. Wait for Google to index the change and check the locations of B for the new phrases.Result: Site B dissapeared from the SERPs for the new second phrase (previously first) and appears for the new first phrase (previously second), like in the image below:
- Revert the situation to the initial state: Switch back between the phrases, wait for indexing and check the rankingsResult: Site B reappeared for the initially first phrase and dissapeared again for the initially second phrase, like in the image below:
OK, so to me this is a pretty solid data supporting Michael’s and Rand’s claims. The fact that the SERPs reacted to my changes in the order of links back and forth is kind of hard to debunk.
However, one question that kept popping up is whether nofollowing the first link will cause Google to index the second link even though it is placed lower in the code. So i performed an identical test to the one described above, just with the first link nofollowed throughout the test. The results are pretty interesting – Google did not change its behavior due to the nofollow attribute applied on the top link.
Even when i switched between the first and the second link (preserving the state where the top link is nofollowed), Google still counted only the top link and not the bottom one.
Besides the fact that nofollowing the top link will not get you around the Google ignoring the second link, it is interesting to see that Google will still count the top link even though it is nofollowed. Maybe it was due to the fact that the nofollowed link was the only link on the web using that phrase as an anchor text ? Who knows… sounds like a theory worth testing.
As for the general settings of the experiment, the keywords I have used belong to the “semi-promiscuous” kind, meaning that they are made up of words that mean something but as a combination are not commonly used. The phrase #1 had about 40 results and the phrase #2 had about 270 results in Google SERPS. It took about two to three weeks for Google to index each of the changes, meaning that each set of results persisted for that time. It also means that the testing for this post took about a month and a half. Hence the posting frequency
As i write in every post describing an experiment, the conclusions should be taken with a grain of salt, Google is a black box that cannot be 100% scientifically tested (I am not talking about you Matt), i was not wearing my lucky underwear when performing the tests, etc.
PS. In the course of the described experiments, i did come up with a method that can help circumvent the described situation and force Google count all the links from a single page to a target domain. It was even independently confirmed by an SEO who knows what he is talking about (unlike me) and is being successfully used on some of our sites. However, I was asked not to blog about it, so if you want to know more and I know who you are, either personally or from Sphinn/Twitter/Plurk/etc., contact me through the form and I’ll drop you a hint