Okay, I have really looked forward to sit down and write this article because this is a subject I was really curious to find out.
You properly have seen many SEOs talk about how updating your old content can help with your rankings right? – But what is the logic to that and how might the machine understand text and content when is edited?
That is what we are here to find out.
What are the SEOs saying about updating old content?
To answer the above question why not start with the overall question and make a search for that on Google and see what we will find out:
Let’s start with the featured snippet from our friend Neil Patel.
Okay so in the first post from Neil, he has 14 reasons why it is good to update your content and how you can go about it, here are them all:
1. Improve your click-through rate
2. Fix grammar and spelling mistakes
3. Show Google that your content is fresh
4. Improve your content’s accuracy
5. Improve the freshness of your entire site
6. Show value by removing broken links
7. Link to newer, better resources
8. Include multimedia for better rankings
9. Improve how often Google indexes your site
10. Optimize for the right keywords
11. Share your content and promote it again
12. Fix any issues with recent Google updates
13. Repurpose your content for more backlinks
14. Leverage the value of your old content
So from first sight they all sound like they make sense, right?
However it is not all I would just blindly trust and go and edit my content based on. Mostly because I know that a machine looks a text a bit differently than we humans and in the same sense doesn’t have the same amount of conceptual understanding of text.
Therefore those points I would like to test out would be point 4 and 10.
The reason for this is that those two are directly regarding changing the text and the possibly the context of the page.
His main argument for editing the content is to trigger Google’s freshness signals but could there be a bigger risk in terms of misleading the context of the content to a point where a page’s topical relevance will get diluted and thereby loose some ranks?
Lets find out.
But first lets us take a look at the next article from the above search. This time let’s not take a Neil Patel page (to diverse it a bit) and take the #3 result namely the article from Make Traffic Happen.
Okay, first of all, it must be said that they are a bit biased as they obviously offer courses based on this matter, but let’s take a look anyway.
The funny thing when both scrolling trough Neil Paten and Make Traffic Happens article is that no where are there mentioned that updating your content could have a negative impact on your rankings.
Like, if you change the context completely you could see a direct decrease in your ranking, I mean this warning should at least be there somewhere, right?
Never mind let’s continue.
How does the machine look at content?
So now that we have get a feeling of what others have to say when it comes to updating blog content, lets us take a look at how a machine sees it.
To do so we will use Tabtimize text analysis engine that normally powers our contextual link prospecting tool and the tale of Peter Rabbit as reference text.
If you care to read the story, here it is:
What I will do is to add that text onto a normal blog page and get it analyzed by the Tabtimize text analysis engine that will get back with an overall topic (what we call URL topic), some niche-specific topics (what we call Content topics) and the associated scores. This is very similar to how a search engine like Google will look at the content.
We can see the machine have categorized the URL topic as Children’s Literature with a context score of 99 out of 100 and have the following content topics: Beatrix Potter (97/100 context score), Peter Rabbit (55/100 context score), The tale of Peter Rabbit (45 context score).
Now let’s change the context a bit and try to add a new section from another tale. This time from the Little Red Riding Hood.
The section that has been added is the marked text in the screenshot below:
This time the URL topic was still categorized as Children’s Literature but with a score of 98 and have the following content topics: Beatrix Potter (95 score), Little Red Riding Hood (70 score), Peter Rabbit (54 score).
So a small reduction in the overall topic score but a big change in the content topic hierarchy with Little Red Riding Hood being more dominant than Peter Rabbit.
Let’s try to edit some the text again. This time lets rewrite and add a new section while keeping the Little Red Riding Hood section.
The first marked part has been rewritten and the second marked part is a need added section from Show White.
Let’s see how the machine understand and scores the text now.
Okay the machine still has the URL topic as Children’s Literature but downed the score to 94 and the content topics: Little Red Riding Hood (97 score), Beatrix Potter (85 score), Peter Rabbit (66 score).
So even though we tried to edit and add a lot of different sections to the text, the machine still understood the main context regarding the tale of Peter Rabbit – Nice job machine! 😁
Real-life examples of blog updates and their impact on the context
Now that we know how Google’s machines might look at and scoring text, lets us take a look at some real-life examples of blog updates, their context scores changes, and the post’s ranking change.
Why not start with an old acquaintance, namely Neil Patel.
So I found an article from him that I think could be very relevant to our little study.
What we will try to do here is take the current version and compare it with the first version that we can find.
The way we will try to find the first version is to use The Wayback machine, which is an online web archive that takes snapshots of websites at any given time.
We will first see if the page has been updated at all. We will do this by first comparing the word count for the two pages, if there is a difference, then we run the two different pages through an online plagiarism checker to make sure the post content has been updated.
We then run the two versions through the text analysis engine to compare context scores, where we finally compare rankings for the most important keywords. Based on the above, we will then conclude if the blog post has benefited from getting its content updated (yes, there could be 100 other reasons why the page’s rankings have changed, but this is better than guessing).
Let’s get started!
Study of the first real life example
Here we got the current version (Ver. 1) of the blog post from Neil Patel:
Here we have the first visible version (Ver. 2) from February 4, 2017:
To make it easy to compare the text, we have used one of Tabtimize’s machine learning models, which automatically sorts out all the noise from the page and gives us only the “main content”.
The latest version of the post (Ver. 1):
The oldest version of the post (Ver. 2):
Examen if the post has been updated
Now that we have the main textual content let us get the word count for each of the versions.
To get the word count we are again using Tabtimize to provide us with the word count. This time it is taking straight out of their backend, so no screenshots, but the word counts are:
Ver. 1 = 4.297 words
Ver. 2 = 4.541 words
Now let test to if the wordings have changed as well:
Okay, so around 11% have the content changed in terms of wordings. We can already from the above screenshot see that at least one new paragraph has been added and from the word count we can see that there have been fewer words on the newest version.
Textual analysis of the two versions
We are now sure that some editing has been made to the post, so the interesting part is to see how it has affected the machine’s understanding of the text.
Again the numbers have been taking directly from the backend so no screenshot ☹ – But the context scores are:
Overall topics (URL topic): Business & Industrial (73/100 context score) – Web Services (65/100 context score).
Niche topics (Content topics): Search engine optimization (98/100 context score) – Blog (79/100 context score) – Internet terminology (54/100 context score).
Okay, so the machine is not very secure on the overall topic, which is quite normal with text that is 4.000+ words long. But it is sure that it is about SEO and Blogs.
Lets see how the machine understood the old post
Overall topics (URL topic): Business & Industrial (69/100 context score) – Web Services (53/100 context score).
Niche topics (Content topics): Blog (96/100 context score) – Search engine optimization (88/100 context score) – Internal link (72/100 context score).
Heck, while we are at it, lets us just put some extra data on the table.
Ver. 1: blog post (82/100 context score) – Old blog posts (69/100 context score) – new blog post (58/100 context score) – blog post graveyard (58/100 context score) – good posts (57/100 context score)
Ver. 2: Old blog posts (81/100 context score) – Bring Old Blog Posts (67/100 context score) – new post (58/100 context score) – blog post graveyard (58/100 context score) – good posts (56/100 context score)
# of images:
Ver. 1: 23 images
Ver. 2: 34 images
All the above indicates that the blog has been shortened a bit and the overall context score has been slightly improved with this change. However, the last question is, how has that impacted the rankings if at all?
The updated posts impact on rankings
TO BE CONTINUED…