When considering the growth of machine learning in relation to SEO, depending on your type of SEO, we may be facing a dire situation.
From a logical and historical point of view, SEOs may be more effective by relying on signal protocols and their fluctuations than by creativity.
I used to think about the "build great content, they'll come" approach, and can even think of the kind of SEO who executes that approach, and maybe less worried now. They should be. . . To some extent. Not yet:
To answer this question before we dive into this change:
What does machine learning have to do with SEO?
We're not going to learn just about machine learning here, nor discuss its impact on us and what future SEO strategies will look like.
All we really need to know from a very long distance is that he adds to Google's incredible data accumulation, interpretation, and speed of reaction. We'll review it at the end of the article, but if you really want to know what machine learning is, there's a free tutorial from Stanford that I signed up for at Cursera.
How does machine learning affect links and link building?
One of the simplest examples of machine learning is adding Google's capabilities to the connection.
Looking at a small example, machine learning comes into play in a key aspect of link evaluation: filtering out spam.
Previously, Google's engineers created lists of low-quality sites, manually blocked their link flow, based on the characteristics of bad links they've seen before, or set the link algorithm-heavy devaluation function well and hoped that it didn't contain too many mistakes. newspaper.
With machine learning, the world opens up.
Yes, there is still a major starting point - a list of bad domains and a hypothetical bad signal. But where the real power can be added using machine learning.
Instead of simply relying on a rigid set of criteria, machines can learn by observing patterns. Watch for sites with suspicious signals (in their linking-out or linking-in), and configure the machine. Then, once it's determined to be broken, it starts a reverse-engineering pattern for faster detection in the future.
What types of sites do spam links?
What types of links are spam sites?
Is there a link growth pattern?
Do pages that sell paid links also tend to link to other specific sites (they do), and if so, which ones?
The system can add this data to the metrics it applies.
This involves how machines can simulate the tip of the iceberg of what humans do and amplify it.
Wondering how Google announces how they demote sites with spammy links instead of manually penalizing them? This is because the machine has learned and applied de-weighting at an astonishing rate, with fewer false positives.
In addition, machines understand the quality and relevance of page content, and that individuality and quality are equal complements. A machine might ask, "Does the link have a high weight on your site?" and then go further, "Is the link purchased or has other issues?" Find and analyze the context in that page or site Other linked data.
This is a very limited example of applying machine learning to links.
We can never cover them, but we have to keep in mind that spammy links will be highly detected, and high-quality links will be given higher weight.
That means focusing more on quality, relevance, and legitimacy, unless you think you can fool Google into not being crawled by machines.
How does machine learning affect SEO content?
Although we use the linked example above, some other areas of SEO will be influenced more by machine learning than content.
To illustrate this, we need only look at Google's translation work. For 10 years, they used phrase-based machine translation - mainly matching known phrases and results for translation. We get results that need to be translated, but are crude.
In September 2016, they turned to a telemarketing list machine learning system (Google Neural Machine Translation System), which within 24 hours of its launch was much more capable of translation than it was a decade ago.
Basically, machine learning can understand language more effectively than human editors in 24 hours, and can do it 3,650 times faster than humans, even with the assistance of machines.
For SEO, that means the Holy Grail of digital marketing is upon us - our only job is to get the best content out. And if it's really the best, Google understands that.
That's not to say the machine is flawless and doesn't work for SEO. In fact, I think it will play a bigger role, but not in the use of keywords, but in making sure that the user's needs are met.
Wil Reynolds gave what I think is one of the best summaries by suggesting that we ask ourselves:
"What would happen if Google improved and only showed the best answers?" This is the question we need to ask. More interestingly, "best" is subjective.
I personally hate videos explaining how to get things done - just give me the necessary checklist and pictures.