What does Entity Salience mean in SEO?
Entity Salience is understanding what a particular bit of text or content is about and it is heavily used by search engines like Google.
Search engines, when crawling content, are trying to identify the entity or entities on the page and from there, which one or ones are the most prominent (salient). The most salient entity on this page is the concept 'Entity Salience in SEO' but an entity could be a place, person, business, subject, etc.
Page's usually have multiple entities. I used Google in the first line and Google is an entity.
Now let's just say I drop the word 'banana' in here and I also use it in my title tag, url and main heading (h1 tag). These are all strong signals to a search engine like Google that the salient entity (most important subject) is bananas. I could mention the word again here and there throughout the text. How would Google then know that the main entity of this page wasn't 'bananas'.
They do it by looking for words that should
also be on the page. Here are the words I've already used that are connected with 'Entity Salience in SEO':
- search engines
- place, person, business, subject
If this content was really about bananas there would be more related words like 'fruit', 'yellow', 'skin', 'pulp', etc. scattered throughout the text. Don't get me wrong, by using the word banana and banana related words I have actually reduced the entity salience of 'Entity Salience for SEO' but I haven't done enough (given the length of this text) to convince any crawler that bananas are the main entity instead.
Salience is measured on a scale of 0 to 1. 0 means this page may contain reference to a particular entity but there is nothing at all to suggest it has any useful information about that entity.
1 means this page is almost completely and definitely about that entity and very little else.
A test in the Google Natural Language Tool
for the text of this article up to the first subheading:
- Entity Salience - salience score of 0.16
- SEO - salience score of 0.15
- Bananas - salience score of 0.00
So despite this being a very short article Google can still pick up the main subject area and almost completely discount my fruity attempts to steer it otherwise!
And remember the natural language tool is a very simple analysis of text. It does not see my title as a title or my subheadings and so does not weight these (but Google in full swing does!)
But why did this text only score 0.16 for 'entity salience'. Because if this really were an article about entity salience alone I would be talking about tokenization, lemmatization, word dependency and all sorts of fun academic stuff. However we don't need to go there unless we're talking about very advanced SEO.
This article should not rank when someone is searching for 'entity salience' - it is a vast field in itself. It should rank for 'entity salience in SEO', and it has a balanced score for both of those terms.
Why is entity salience used?
From the early days of the Internet search engines had to deal with spammers who tried to rank pages where they shouldn't and were often successful. You could search for 'pictures of kittens' and find half the top ten results were for gambling or adult related websites.
Spammers would so it simply by using 'pictures of kittens' in specific places such as the url and page title and a large number of times on the page itself. One way to achieve this would be to insert the phrase using white text on a white background so only search engines would see it.
To see through this search engines used a second major signal - who links to that page/site? If a lot of respectable
websites about kittens linked to it this was a strong confirmation of what the page's entity was.
The weakness of links as a ranking factor
The issue with links is that websites who already rank highly in the search results for a particular search term tend to gather more links going forward than anything else. This means there could be better content out there but its hard for it to get discovered. Catch 22.
The Entity Salience search engines will attempt to understand when 'better' content appears on the web and should rank higher than other websites even if it has a very weak or non-existent backlink profile.
The future of entity salience
As of 2019 the work of search engines in this area is fast gathering pace and we can expect it to change the way rankings work over the next few years.
While I've looked here at entity salience of one page or one piece of content search engines will also consider the entity salience of an entire website or of a section of a website.
Remember earlier the Google Natural Language Tool gave SEO in this text a salience of 0.04. However it quite clearly understood the salient entity was 'Entity Salience'.
At a site level it knows the salient entity of this website is SEO so it can now join the two things together. A website about SEO with an article on entity salience should rank highly for 'entity salience in SEO'.
How to use entity salience in SEO
If you want a page to rank for a particular term you have to make sure search engines see it as the salient entity.
First stop is the Google Natural Language Tool is a great place to check basic texts but obviously don't forget the other signals like page title, url, heading and subheadings, image alt tags, etc.
Secondly you should also study your text for areas which lack clarity when being machine read. For example:
Keyword density was important in SEO but many people argue it is now dead
What is dead? Keyword density or SEO. A crawler might struggle to know because it still can't understand the overall context of the wider world and include that knowledge the way humans can.
Thirdly look at the related words used by the websites that currently have high rankings. Are you using these in your content.
Fourthly look at what phrases your page is ranking for but not getting clicks by using the Google Search Console. Could you use these to expand your content and so make it more salient to these entities as well - because there is no reason content cannot rank well for multiple entities.