This is just a reminder that Google uses machine learning to understand if content is helpful or not within the Google helpful content system. This is not new, we’ve known this for over a year, but it seems many of us keep forgetting.
Google told me back when it was about to release the helpful content system that it aggregates a variety of signals about the page and site to determine the ranking of a page. Google would not say if links are part of those signals or not, but it does not seem so.
Google uses machine learning to identify such content – content designed to rank well in search and not be helpful to users. Google told me they validated these algorithms with quality raters and that using this system improves its search quality, just as Google validates any type of ranking improvement prior to launch, Google told me.
Danny Sullivan, the Google Search Liaison, repeated this Friday night on X, saying, “That brings things back to the helpful content system that yes, using machine learning to understand about content from words on the content and signals beyond it to know if it seems helpful.”
He added, “But the key thing is that it’s saying helpfulness often can’t be determined from “the words or images alone” which makes sense. I mean, if someone just wrote “Hey, this is helpful content!” anyone — not just a search engine — would look for ways to know it really was. That’s why the post went on to talk about how we use signals that “align with what humans might interpret as high quality or reliable.”
Here are those tweets if you want to dive into it, click through:
If you’re trying to understand how our helpful content system works, this is our page that explains it:https://t.co/b1MYLCMGC9
The blog post you’re referring to is from two years before we introduced this system, and it’s not about the system specifically. In particular, it’s…
— Google SearchLiaison (@searchliaison) October 27, 2023
