A big conversation has been taking place recently in the SEO community about whether we should still be treating keywords as the north star of SEO, or should instead focus on topics. Latent semantic indexing, or LSI, is at the heart of this conversation.
For marketers, LSI keywords can bridge the gap between the way we're creating content now and the way we need to do so in the future. They give us a way to talk about the changes we need to make to our SEO strategies.
But enough of the concept—here are a few definitions.
Latent semantic indexing (LSI) is a system search engines use to analyze the other words people use surrounding a given topic.
LSI keywords are words and phrases with a high degree of correlation to your target topic. Google's algorithm uses them to help determine content quality and relevance to the search term.
I've seen a number of posts refer to LSI keywords as synonyms of the target keyword, but this is misleading. It's more accurate to say that they are the words that are most commonly correlated with the search in high-value content. That includes synonyms, but a great deal more as well.
At this point, LSI keywords appear to be more important than keyword density in search engine algorithms, so let's dig in a little deeper.
What Latent Semantic Indexing (LSI) Does
Search engines use LSI to judge the quality of the content on a page by checking for words that should appear alongside a given search term or keyword. Like every other SEO development we see these days, LSI was designed to help searchers find what they're looking for, not just what they searched for.
For example, if I wanted tacos for dinner, I might search "where are the best tacos in Knoxville." Might? Who are we kidding, I would definitely search that.
When "best tacos" is the topic, Google might expect to see words like "salsa," "guacamole," and hopefully "handmade tortillas" along with it, because these words are used on high-ranking pages about tacos and in users' searches.
A personal blog post mentioning "the best tacos" only in passing won't include those words, while a page full of Yelp reviews for Oscar's Taco Shop will. Hence, I see Oscar's at the top of my search results and I know where I'm going for dinner tonight. You've seen LSI in action when you conduct a Google search and then see highlighted words that don't exactly match your search term.
So what do we know for sure? Google has confirmed that using more LSI keywords will typically cause your page to rank better. This makes sense because people are searching for any given topic in various ways, and a page with more LSI keywords not only gives Google more context and information, but will match the search terms of more users.
Finding and Using LSI Keywords
When it comes to tools that can produce a list of LSI keywords with semantic relationships to your search term, the pickings are slim.
LSIGraph LSI Keyword Generator is the best, easiest tool I've found, and by far the least sketchy-looking. Start there.
You can also look at the "Searches related to" section at the bottom of the first page of search results, though related searches may include searches people conducted to follow with secondary questions, to clarify a search term they phrased poorly, etc. If you want to dig a little more deeply, you can also look at the other terms used in the pages that are currently ranking for your target keyword.
Finally, you can use topic modeling to come up with semantically-related terms that will make your content appear more relevant to search engines.
This 10-minute video on topic modeling and semantic connectivity from Moz explains this brainstorming process:
Naturally, people search for the same topic in many different ways. A certain portion of your organic traffic (probably a minority) will come from your target keyword, while a significant portion of your traffic comes from LSI keywords, long-tail variations of your target keyword, and other search terms.
Remember when I said that keyword density isn't as important anymore? Well, that's not exactly true.
If you were to visualize the keyword density in a piece of content, you might picture a page with your target keyword highlighted throughout it. Looking at things that way, your keyword density seems low. And it has to be, or you risk getting penalized for keyword stuffing.
Using the LSI keyword model, though, you could imagine the keywords in your content as a web. Start with your target keywords, and picture lines running to important related terms (LSI keywords). These terms relate to each other, too, so you could draw lines among them, as well. This network of semantically-related terms is your true keyword density.
So, where do you need to use LSI keywords?
- Throughout the individual page that you want to rank for the topic (obvi)
- In headings
- As link anchor text when possible
- In the alt text, file names, and titles of images
- In the page title to support your target keyword, when possible
- Use closely-related LSI keywords close together (but naturally) on the page
For the topics that are most important to your business, you can also create additional content targeting the LSI and long-tail keyword variations, and then link to it using the variation as the anchor text.
You know those times where you want or need to use your target keyword again, but it's starting to sound redundant or spammy? That's a perfect time to use an LSI keyword.
To wrap things up, here's how I would summarize the takeaways:
Instead of target keywords, begin focusing on topics and semantically-related clusters of keywords for your content. When you're working on the usual elements of on-page SEO, select from this list of terms rather than forcing in your target keyword. Use your target keyword where it both sounds natural and produces the highest SEO value, like in the title.
As it turns out, taking advantage of LSI keywords isn't too difficult. You're probably doing much this already—after all, Google has been talking about semantic search for a while, and substituting a related term when things are sounding spammy is natural.
However, understanding LSI keywords allows you to do all of this more strategically, and that's pretty satisfying. Much like watching LaMorne Morris flipping pancakes.