I hear this so often it must be true. But then I also hear "slow down or you'll get a ticket and our insurance will double" every time my wife is in the car, so maybe I should double check on the relevance thing. And I mean specifically how does Google see pages, not how can you get qualified traffic?

Personally I don't think they care or try to understand if or how two pages are related to each other. How could it? That's just too big a question to answer with some type of program code. I write ASP.NET web applications all day long, and I'm not the type of genius that starts Google, but still, the question seems unfathomably big from where I sit.

I'm starting to get links from respected blogs. Ones I've never heard of, and I'm learning about parts of the web I didn't know about, so blogging is a good experience overall. But now that this is happening, I want to put some thought into what I write, who might relate to me, and where new links will come from.

Let's use an example to clear things up. I'm writing a post right now on what might be my favorite topic: evolution. Only, so far I haven't used the actual word evolution. I talk about Darwin a lot, genetics, survival of the fittest, and all kinds of concepts that people have built up a rich enough background knowledge to know are related. Darwin sort of became the poster child for evolution, and with all the news and fighting over intelligent design, maybe Google has a lookup table that says Darwin = Evolution. But what if I didn't use either one of those two words? What if I just talked about the concepts, about genetic mutations, sexual selection, and statistics?

I'm hoping to get links from web sites about science to this piece I'm working on. Really that's a hope, and whatever I get is a step in the right direction. But people keep acting like how related two sites are has a lot of bearing on how Google treats those links. I want to know if that's true? And if it is, are they smart enough to understand in our example, that even though two pages don't use the same word a whole bunch of times, they're discussing the same concept?

This is sorta academic, I'm going to write the post however it sounds good no matter what the Google algo is. The point is I think evolution is a good question, and the last post I wrote on the subject, people kept complaining was too short. I'd love to investigate in more detail, hopefully get people to comment on what they think and if I'm on the right track, and if people want to see more of that, it's a match made in heaven. Otoh, now that I'm starting to understand a little, I want to know more.

related/relevance is for humans not search engines, and even then it's subjective.
One mans persons meat and all that.

Related in respect of SEs is merely a way of potentially "second guessing" what the user was looking for and returning those results.

Why would Google care? Or other search engines, for that matter? If anything, it seems like being noticed outside your industry is probably a better indication of quality than when two sites about the exact same subject mention each other.

I'm sure Google has some idea about whether things are related. They seem able to spot madlib spam, and "personal search" is the next big thing. I have no idea how that works, but I'd be really surprised if Google wasn't trying to leverage their user data.

It seems like you're wanting a different question answered. The links you're getting lately from other sites, "respected" ones, are probably coming in because you're being controversial and topical. Your most recent post about the sky is falling because Scooter Libby won't do any jail time is a great example of link bait. It sounds like you wrote it from the heart, it's too impassioned to be a SEO trick, but it appeals to people who follow the news with eagle eyes. And to anyone who follows politics like a sport.

The topics you seem to be choosing lately are more likely to attract readers and attention than song lyrics.

I disagree and think Google does care and can tell more about how two pages are related to each other than it might seem at first.

With all the pages Google has indexed they can easily see correlation between words like 'evolution' and 'Darwin' and 'genetics' etc. How do the know? Because the pages they have in their index that use the word 'evolution' also use those other words often. If they encounter a page that uses all the words except for 'evolution' I think they'd still be able to conclude the page in question had something to do with evolution.

Assuming there are links to your page from other pages I think Google can look at those other pages and decide what they think those other pages are about. Let's say all the pages linking to yours mention the word 'evolution.' I think in that case Google would again conclude your page has something to do with evolution.

Clearly if the anchor text (uh oh, did I just say anchor text. time to ban me) used to link to your site mentions 'evolution' they could conclude your page is also about 'evolution'

Say you get a link from a page that uses the word 'evolution' in the you know what text. That page also includes 'evolution' in the page title and on the page itself. You don't mention the word 'evolution' anywhere, but all those other words you use are also included on the page linking to you. I think Google would still determine the pages are related.

I'm speculating about everything here, but I think they can determine in many ways how similar two pages are in their topic and how related they might be.

Chris is absolutely right in the sense than many things that are related are very subjective. I don't think Google or any current algorithm can determine those kind of relationships.

But by observing common words used and linking patterns I think all search engines can determine a certain amount of relatedness.

The reason I think they would care is in how they view links. Let's say you have a problem with the plumbing in your house and you want o fix things yourself. You ask a few people how to go about the work. One of the people isn't necessarily the brightest bulb in the box, but he happens to be a plumber. The other person you ask is a Nobel Prize, but has never held a wrench in his life. Who's opinion will you value more in regards to your plumbing question.

You might generally trust the Nobel Prize winner, but in this particular instance the plumber by the nature of being related to your question is probably the better source.

I think it works the same way with how Google would see links. A vote from a page or site that they've determined is about the same topic as your page or site would be seen as more valuable than another page or site on a different topic. (all else being equal of course)

Clearly if the anchor text (uh oh, did I just say anchor text. time to ban me) used to link to your site mentions 'evolution' they could conclude your page is also about 'evolution'

Three days in link exchange. You get to choose whether to teach blind people who don't believe some people can see, or to amuse us with your antics.

That's a lot of great info, or at least speculation, about how things might work. I've read articles about vaporware search engines that did things like send your query to Wikipedia and read the top 10 results, find the most heavily used words, and suggest other searchs. So if you say "Civil War" it might list battles, the people involved, things like that.

Google might, at some point, translate your query at some point, and sort the results by whether it finds the exact phases, or synonyms, or even just common associations. If someone searches for "Lamarkism," I'm guessing you haven't used the word and none or very little of your anchor text will ... a friend of mine is a PhD in genetics and likes explaining her work. Anyway, Google probably knows about the guy's bio, and also that he thought a giraffe stretched his neck really far to get an apple, and he had children with long necks, about 50 years before your friend Darwin. I wouldn't be surprised if someone searched for the man's name with "ism" at the end, they probably associate it with ideas involving the man, rather than his biography. You might come up for a search like that, but I'm speculating wildly.

Why are you not using the word evolution in a page about evolution? Is this some kind of experiment?

Actually Google does play some importance in "related" terms and keywords, but I'm not going to go further on that.

I don't think Google bother checking if sites are related because it could actually do more harm than good. If you think about a pro-evolution site and a pro-creation site they probably mention evolution, Darwin, Natural selection about the same amount of times. But the pro-creation site isn't going to be what someone wants if they search "evolution".

A machine and and algo are going to be terrible at telling if something is "related" and at best can only determine what pages use the same words.

****...poor Steven. He just got sent to Hell for 72 hours. ;D

Seriously, my opinion on the relationship is very similar to Steven's. An algorithm can be made to determine if two sites have a related theme, and it would be relevant to the overall quality of SERPs for one simple reason: PageRank Protritanopia. All the idiots running around trying to gather links from everywhere that they can possibly get links, exchanging links, participating in link schemes a la Digital Pointless, etc. and so on.

Determining the "theme" of a site based on the words it uses in relation to the "theme" of a site linking to it (and vice versa) will give some idea as to whether or not two sites are related. It will never be 100% accurate, due to the wide variety of dialects, slang words, and other English twists and turns that people use. But if it can be made to figure out if two sites are related in say 75-80% of cases, then it could be used in an algorithm.

It is used to at least some degree, and the proof is in the semi-obscure and rather crude IMO related: search operator.

http://www.google.com/search?sourcei...3agoogle%2ecom

Originally Posted by stOx

I don't think Google bother checking if sites are related because it could actually do more harm than good. If you think about a pro-evolution site and a pro-creation site they probably mention evolution, Darwin, Natural selection about the same amount of times. But the pro-creation site isn't going to be what someone wants if they search "evolution".

A machine and and algo are going to be terrible at telling if something is "related" and at best can only determine what pages use the same words.

That's kind of what I thought. An interesting case is that the Church of the Flying Spaghetti Monster ranks #5 in searches for Intelligent Design. It's not what IDers want, but it's what they asked for.

A lot of good info going on in this thread. I thought "related sites" was link exchange speak, but it sounds like there might be something to it. And I agree it can never be 100 % accurate, but I guess it can do more than I thought.

Originally Posted by ForrestCroce

Why are you not using the word evolution in a page about evolution? Is this some kind of experiment?

Actually I wound up putting "Evolution 2.0" in the title. I just hadn't used the word yet at the point when I asked, and wanted to know what would happen, how this stuff is processed.

Evolution is the name of the concept, and you can avoid using the E word by talking about "mutation over time," "genetic adaptation," "natural selection" and stuff like that. When you start repeating these phrases you go with the E word just so the copy doesn't sound so redundant when you read it. But I didn't want to open with "evolution" when we've all heard it so much, thought using other words to describe the same thing would make my piece stand out a little more.

I just hit publish, though, if anyone wants to check it out. Folks are in for a surprise, I did the unthinkable. I actually defended something Bush said. I know, three years in link exchange, then I'm going to hell.

I think I'm free from link exchange hell after John's admission. I knew mentioning that phrase would get me in trouble though I hope it gave you a laugh.

Matt I would say those two sites are related. They don't agree with each other at all, but they are related. A page promoting a democratic candidate for the 2008 election and one promoting a republican for the same election probably don't see eye to eye on much, but both are about politics and a specific election.