Expert SEO Strategy: LSI Revisited

There was a short period of time after the Panda update hit that expert SEO people were actually declaring the “death” of keywords…as if the actual foundational mechanic that Google relies on for basically everything the search engine does at all ever would suddenly just go away. Of course keywords didn’t go away — but for whatever reason, around the same time, the SEO community at large did stop paying attention to the phenomenon called Latent Semantic Indexing, or LSI. And that was a mistake.

LSI Defined

Latent Semantic Indexing works like this:

• Start with a huge pile of documents (in Google’s case, it’s “all indexed websites,” which makes it the biggest Latent Semantic Index in the world.)
• Pick a word.
• Note the number of times that every other word appears in the same document as the chosen word.
• Put together a list of words that appear very, very often together in documents.
• Link together those most-common-connections to create a map that shows, by linking each word with dozens to hundreds of other words, what ‘meaning junctures’ a given word connects to.

The LSIndex that Google maintains is precisely what allows it to know that the word “fly” has at least five separate ‘junctures’ of meaning (insect, zipper, move through the air, kind of event in baseball, super-cool).

Application to SEO

If you’re an expert SEO guy, what the existence of the LSIndex means to you is that it’s not enough to just have a keyword repeated 1.4 times per 100 words or whatever standard you wish to go by. If you don’t have the LSI words that go alongside the main keyword and show Google what you’re actually talking about, you might get your page centered around the keyword ‘wicket stumps’ ranked for searches about Ewoks rather than searches about the game of cricket.

For most content creators, LSI wording happens quite naturally as a normal part of creating human-readable, engaging content about a subject — but if you’re not a native English speaker, or you’re just whipping words together for the purpose of posting something to meet a deadline, you might just end up missing the LSI train and creating a page that Google has no idea how to interpret. Don’t!