Tweet
The answer you'll find everywhere is that topical authority is something you gain when you have enough content about a specific niche.
Ok great, but why does that help us? Why does google rank sites that have more content about a niche?
I have a great analogy for you.
Imagine you are a software engineer, and you go for a job interview.
What will the interviewer do?
He will probably ask you a whole bunch of questions he already knows the answer to.
The more correct answers you give him, the more you show him that you know the topic.
That's it. This is topical authority.
You are simply showing google you know the answers.
Now, this is slightly different when it comes to SEO and AI(Which is what google is now)
Google isn't actually "checking" your answers. It's checking your questions. This is exactly why PAA sites are so powerful.
Google already has all the connections between sub-topics and questions within a niche. It's already mapped out niches based on existing websites.
By covering as many questions and sub-topics as possible, you are showing google you understand the niche. You aren't proving to it you know the answers. They don't really have the computing power to be doing fact checking on a large scale. But you are showing to them you are covering all areas within the topical graph that they know about.
I can explain to you how they actually use AI to determine your topical authority score relative to particular niches.
They use embeddings and vectors.
An embedding is a high dimensional vector space which can represent very complicated entities. You can then reduce that high dimensional vector down to a low dimension vector in order to compare them very quickly(or with the human eye).
Google maintains a large vector database of topics.
Let's take an example.
You want to rank for the keyword "how to train my rabbit"
Google has this keyword represented on a vector database.
So if you have an article on your site called "how to groom your rabbit", then it can measure the distance between the 2 keywords.
It'll do this for all the articles on your site, as it's a very fast and easy computational operation.
This, mathematically, is what topical authority is. Analogously it's the interviewer description above.
They can calculate the vector distance for every article on your site, to a keyword, and give you a score. That score is your topical authority.
On-page uses similar concepts. They're measuring distances in vector space to work out how close your topics are to the main user intent.
Here are some patents that talk about their usage of embeddings and vectors.
https://patents.google.com/patent/US9740680B1/en
This one is just talking about word vectors.
But here is a very interesting one..
https://patents.google.com/patent/WO2021252076A1/en
Generating a graph data structure that identifies relationships among topics expressed in web documents
And here is a highlighted part of the document where they talk about measuring the distance between the vectors.
There you have it. Topical authority explained in practical terms, and an explanation of the AI that makes it work. Embeddings aren't "AI", but they are used by the AI and play a big role in machine learning.
Ok great, but why does that help us? Why does google rank sites that have more content about a niche?
I have a great analogy for you.
Imagine you are a software engineer, and you go for a job interview.
What will the interviewer do?
He will probably ask you a whole bunch of questions he already knows the answer to.
The more correct answers you give him, the more you show him that you know the topic.
That's it. This is topical authority.
You are simply showing google you know the answers.
Now, this is slightly different when it comes to SEO and AI(Which is what google is now)
Google isn't actually "checking" your answers. It's checking your questions. This is exactly why PAA sites are so powerful.
Google already has all the connections between sub-topics and questions within a niche. It's already mapped out niches based on existing websites.
By covering as many questions and sub-topics as possible, you are showing google you understand the niche. You aren't proving to it you know the answers. They don't really have the computing power to be doing fact checking on a large scale. But you are showing to them you are covering all areas within the topical graph that they know about.
I can explain to you how they actually use AI to determine your topical authority score relative to particular niches.
They use embeddings and vectors.
An embedding is a high dimensional vector space which can represent very complicated entities. You can then reduce that high dimensional vector down to a low dimension vector in order to compare them very quickly(or with the human eye).
Google maintains a large vector database of topics.
Let's take an example.
You want to rank for the keyword "how to train my rabbit"
Google has this keyword represented on a vector database.
So if you have an article on your site called "how to groom your rabbit", then it can measure the distance between the 2 keywords.
It'll do this for all the articles on your site, as it's a very fast and easy computational operation.
This, mathematically, is what topical authority is. Analogously it's the interviewer description above.
They can calculate the vector distance for every article on your site, to a keyword, and give you a score. That score is your topical authority.
On-page uses similar concepts. They're measuring distances in vector space to work out how close your topics are to the main user intent.
Here are some patents that talk about their usage of embeddings and vectors.
https://patents.google.com/patent/US9740680B1/en
This one is just talking about word vectors.
But here is a very interesting one..
https://patents.google.com/patent/WO2021252076A1/en
Generating a graph data structure that identifies relationships among topics expressed in web documents
And here is a highlighted part of the document where they talk about measuring the distance between the vectors.
There you have it. Topical authority explained in practical terms, and an explanation of the AI that makes it work. Embeddings aren't "AI", but they are used by the AI and play a big role in machine learning.
