“Irrelevant” backlinks still work

I think what a lot of people call “relevant backlinks” are overrated. There are still pages that rank for very competitive terms for which the backlinks are all just links in the footer on every page of a big site on a diversity of topics. Relevancy will surely become more important as algorithms get smarter but for now I don’t think it’s as important for SEO right now as lot of people say it is.

Everything being the same(like amount of effort in acquiring a link), I’d take a link that is given in a natural flow of text that is closely and unequivocally related to the content at hand, but those links are harder to get from others and I’m going to try to get whatever links I can that are not part of some artificial scheme that could get my site in trouble. That means forum signatures, social networking sites, blogrolls, footer links on every page of a site, whatever.

Here’s a good post on the subject. http://seo-theory.com/wordpress/2007/06/18/the-relevant-link-myth/

The post is almost a year old but I think it’s still relevant because whether he was right or wrong then, enough time hasn’t passed for that to have changed.

Link relevancy seems to be largely determined by anchor text

Michael Martinez’s contention is that the anchor text is what dictates relvancy, so that any link going to a page with relevant anchor text is relevant by virtue of that text; it doesn’t matter what the rest of the page is about. Well it’s hard to know exactly how Google measures the relevancy of a page in relation the page it is linking to but I find it difficult to believe that Google’s algorithm — even given all the data it has collected to work with and glean info from — is smart enough to determine relevancy of a page in relation to a page it links to with a lot of accuracy unless it relies mostly on anchor text and text in close proximity. Let me rephrase that, an algorithm can definitely determine the relevancy between two pages, based on probability models and such, but not in a way that could take the human factor fully into account and in a way that is necessarily in the best interest of the user.

Relevancy algorithms probably are still lacking in their usefulness in the real world

Sure you could probably quite easily determine the relevancy between two pages with an algorithm or fancy formula, and I’m sure Google does this to some extent(I just don’t know how important it is), but an algorithm will never be able to determine a relevant relationship nearly as well as a human. So why force an assessment of context that is likely to be flawed outside the academic world of mathematical models and penalize a page that is likely to be useful to a human who is much more capable of discerning context?

Take portal sites for example; Many portal site pages provide highly relevant links to many different types of pages, they just offer different sections on the page for different topics. Michael Martinez in the post I linked to above used the example of CNN, saying, to paraphrase, ‘if CNN linked to your site about car insurance from their home page would you email them and tell them to remove the link because their site is not about car insurance’?

I just took a quick glance at CNN’s home page. From a quick scanning I see references to Iraq, Bill Clinton, Martin Luther King, the rising price of gas and stock market closing numbers. So how would an algorithm determine the theme of that page in a way that could determine the usefulness that a link to another page has? (How does an algorithm know that this page or at least this area of the page isn’t really in any way related to the five subjects I mentioned two sentences ago?) If there was a section on car insurance on the home page of CNN would it be irrelevant? I don’t think so because as a user, when I go to CNN I want a small amount of information from a variety of topics so I, being the reasonably intelligent human that I am, can decide what resource I would like to consume at that point in time. Algorithmically, I just don’t think we are there yet so apply a heavy penalty for being non-relevant when the algorithms are not smart enough to completely discern the context.

Looking for irrelevancy as opposed to relevancy

What the algorithms probably are able to do at this point is find links that have certain characteristics that make it likely to NOT be relevant. This means detection of artifical linking schemes(link exchanges) and maybe even where the links are placed on a page. For example all popular blogging platforms have fingerprints that identify them, if links showing up in a certain area of the page(in relation to the code of the page) and that area has shown a high probably for paid links, then a probability model could be applied to find it’s liklihood of not being relevant.

Summary

I guess to summarize the opinion I have on the subject is don’t worry too much about just getting “relevant” links(get them if you can for sure) but do worry about not getting links that could be easily seen as NOT relevant, in computer or human terms.

I’d like to see some controlled experiment done to try to get insight into how Google does measure relevancy of backlinks, but I haven’t found one yet. If you know of such an experiment please leave a comment and let me know.