There’s a lot of struggle in the world of search engines to define ‘relevance’. How do you determine, if someone writes the word ‘fly’, whether they’re talking about insects, zippers, airplanes, baseballs, or being cool and/or sexy? Google is famous for their Latent Semantic Indexing (LSI), which has become a huge keyword in the content creation industry and basically means “you gotta write about other related subjects in the same post to create the relevance.”
This has in turn created a huge flaw in the industry devoted to writing content, and that is that absolutely nobody ever actually speaks (or writes) about a single subset of content when they’re speaking naturally. While Google hasn’t actually come out and said is in particular, it is my firm belief that as part of the Panda content-cracking-down-on, they’ve implemented some sort of ‘relevance calculator’ that looks at the number of subjects that someone touches on in a few hundred words of text and tries to figure out whether that content is clearly overly-SEO’d by figuring out which content is actually too narrowly relevant to be good content.
Here’s a specific example. Let’s say I hire a writer to talk about his favorite song of the moment — Macklemore’s “Ten Thousand Hours“. He could pump out 350 generic words talking about the song and what it means to him, and not mentioning anything else about it. Or, he could skip the generics and look at some specific thing in the song — for example, the fact that the song mentions Malcom Gladwell, who publicized the ‘10,000 hour rule’ — notion that it takes about ten thousand hours to become a world-class expert at any one thing (about ten years of more-or-less daily practice.)
Now, Malcom Gladwell is definitely a completely separate subject from Macklemore, even if one appears in the other’s art. But there’s much more — the same song mentions graffiti artists Basquiat and Keith Haring, opening up a completely different area of relevance — graffiti art (something that, as a Detroit native, I have some appreciation for myself). It also brings up the No Child Left Behind program and points out at least one significant flaw in the theory behind standardized testing — which, just to come full circle, Malcom Gladwell also pokes holes in from time to time.
What’s the point here? The point is that CONTENT! can be distinguished from ‘more web-space filler’ because, among other things, great content draws connections between different areas of relevance. That act — connecting A, B, K, and 14 without doing it by copypasta — is how new, relevant, interesting thoughts are crafted, and it’s the kind of content that Google likes to see.
How does that work with LSI? It’s simple — LSI looks for other related words nearby to establish the context of a given ambiguous phrase. What it doesn’t do is use the context of a given phrase to establish the context of an entire page. Normal humans simply don’t talk or write about one single thing at a time. (Just look at all of the references packed into that three-minute song to see what I mean.) Google knows that — and they’re not going to penalize you for creating natural, human content that adds something useful (or at least new) to the Web. So stop writing content that’s overly-focused, and connect whatever you’ve been assigned to write about to something different and create a new relevance.