#FreeAdvice is a monthly feature where we take your questions about search/analytics/etc and answer them here. The number of answers will vary based on length, complexity, our energy levels.
Q. I’ve been hearing about entities lately and have no idea what it means. Help!
The thing about entities is they’ve been around forever. We just haven’t always done a good job associating them with search, or at least not using that name. The word came into the SEO sphere back in 2013 when the Hummingbird update brought knowledge panels and structured data into our world. But it’s kinda just been a thing in the background.
A lot of SEOs have pushed adding Schema markup to sites, but largely for the purpose or rich data in SERPs (e.g. star ratings for a product). That’s great and all, but totally missing the bigger point.
The real idea of entities (in the SEO world) goes back to the idea of latent semantic indexing, which has its roots in pre-WWW times, but has been long connected with Google going back to its early days. The very very basic concept is it’s a system of comparing documents based on related concepts, word frequency and group them based on common themes.
An example I often give is for a topic like Stephen King, the author, search engines have to be able to separate pages about him from anyone else named Stephen, from kings in the monarchical sense and from anyone else named Stephen King (plus other variations). How does that happen? You start with pages you know are about Stephen King, author (from America, who wrote It, the Shining, etc.) and see what other terms are similar. They probably mention It or other books you know he wrote. They probably mention Castle Rock, Maine. They probably mention his wife and kids.
That becomes your baseline to examine other pages that *might* be about the same Stephen King. SEOs took this as adding semantically related keywords to copy in order to make sure you rank for the thing you care about.
So while we weren’t calling Stephen King an entity back then, that’s the idea.
Now it’s easier, kinda.
This has gone on way too long already and I’m falling asleep just thinking of the idea of proofreading it all, so we won’t get into how entity associations are made or any of the technical stuff. Go look up Dixon Jones (person, not company) and you’ll learn more than you ever wanted. So let’s get to the point!
An easy way to think of entities is to look at a Wikipedia article. The piece that’s always in the upper-right is quite similar to Google’s knowledge panel (not by accident) and lays out how these things work in fairly simple terms. The example below has some terminology that comes from Schema.org, but should be pretty easy to follow. The thing (entity) is Stephen King. All the additional attributes help machines know which Stephen King we’re talking about.
The brilliance, for SEO, is some of those attributes are entities themselves. The interconnections between those entities is how Google gets to make sense of an impossibly large internet. Basically, if you’re only talking about keywords, you’re in a lot of trouble. Not everyone in the SEO world is beating the drum for just how critical this is, but they will before long (or they’ll be in a WHOLE LOT of trouble). Understanding and connecting entities is a crucial piece of how natural language processing will evolve in the coming years. Hopefully you’ve now got a good grasp of what it means. The concept isn’t too tough. Taking action is where it all gets tricky.