Google's Use Of Machine Learning In Ranking & Spam Algorithms Is Complicated

This week we reported that Google's Penguin algorithm doesn't use machine learning. This was based on what Gary Illyes told us on Twitter. But then someone asked John Mueller of Google answered the question differently.

Here is Gary's tweet:

@jenstar nope

— Gary Illyes (@methode) October 10, 2016

Here is John's tweet:

@rjonesx @rjonesx @methode Uhhmm. We do use machine learning in a lot of our systems, so in general, it's a good assumption to make.

— John Mueller (@JohnMu) October 12, 2016

So based on Gary, the answer is no. Based on John, we should "assume" yes.

So then Gary came in to defend John and Google saying John didn't say Penguin outright uses machine learning:

@jroakes John didn't say we used ml in old penguins. To the best of my knowledge we didn't @rjonesx @JohnMu @jenstar

— Gary Illyes (@methode) October 12, 2016

Then John came in to explain more:

@jroakes @jroakes @methode @rjonesx @jenstar We analyze a lot of data, ML or not. Don't get hung up on the name :).

— John Mueller (@JohnMu) October 13, 2016

So then I asked the ranking boss at Google, Paul Haahr, and he said on Twitter "it's complicated."

.@rustybrick Very ambiguous question and not a great one for twitter. The most I could say here would be “It's complicated.”

— Paul Haahr (@haahr) October 11, 2016

Well, it is complicated to say if most of Google ranking is based on machine learning or not. Maybe I will try to get a better answer from Google in the near future on this topic.

Forum discussion at Twitter.