Google Search Powered by Artificial Intelligence, RankBrain

[engine]

Here we go, Google RankBrain.

Not the best of terms, but now you have it.
Could this be contributing to Zombies?

For the past few months, a “very large fraction” of the millions of queries a second that people type into the company’s search engine have been interpreted by an artificial intelligence system, nicknamed RankBrain, said Greg Corrado, a senior research scientist with the company, outlining for the first time the emerging role of AI in search.

RankBrain uses artificial intelligence to embed vast amounts of written language into mathematical entities -- called vectors -- that the computer can understand. If RankBrain sees a word or phrase it isn’t familiar with, the machine can make a guess as to what words or phrases might have a similar meaning and filter the result accordingly, making it more effective at handling never-before-seen search queries.

Google Search Powered by Artificial Intelligence, RankBrain [bloomberg.com]

[ogletree]

I have been seeing this and I'm sick of it. I don't know how smart this algo is. I think all it does is take a zillion searches and boil it down to a much smaller number of searches. The only problem is some of us know what we are searching for and don't want Google to rewrite our search. What I see most the time is that I do a very specific search that used to bring up good results but now brings up the more mainstream result completely unrelated to what I searched for. It is like Google used to play all kinds of music but only plays top 40 now.

Is there a way to get old Google and not let them do this?

[IanKelley]

By making a phrase a "numeric entity" it becomes easier to determine what other phrases it's similar to.

I imagine after that it starts to get complicated pretty quickly. Sub phrases, relationships to other parts of the algo, history, etc..

[masterjoe]

ogletree I've noticed this too. Especially when I am looking for very specific things like model numbers of certain products, or trying to find reviews on them, and especially when it comes to doing my own research of trying different combinations of searches to find related content about the same thing. It tends to change far more dramatically than before. Absolute waste of time and money when the results were fine before. It could have still used some tweaking to get ridiculous "authority" content (thin articles spread out over 10 pages) off the first page, which was one of my only complaints. Now my only complaint is that this AI stuff is useless at deciphering what were actually looking for.

[ken_b]

Does the "rank" in rankbrain indicate the fake brain is pretty rank?

If so I think I saw some of it's work last night when a search for "delete" (something) brought up two pages of listing about "recovering" (something). Pretty frustrating.

.

[Valdo909]

ogletree I've noticed this too.

And why Google would not have discovered this in their tests?

[heisje]

Perfect way to totally mess up the "long tail".

.

[ogletree]

The reason that Google does this is that they really don't care about the power user any more. They are all about profit. The average search on Google is very basic. Most people find what they are looking for. They also have to do this because the average searcher does not know how to search. Google has dumbed-down search to the lowest common denominator. I'm guessing that the vast majority of searches that did not provide results in the past can now be fixed by this system because most of the time Google can guess what the average person is searching for.

[dipper]

@ogletree - "Is there a way to get old Google .. "

Yes. It's called Bing.com .. ;)

[heisje]

The average search on Google is very basic. Most people find what they are looking for.

This is the thinking behind Bing and its limited index size. It has not brought any spectacular success to Bing.

.

[aristotle]

I thought they were working on something that has a self-learning capability, using some form of feedback to gradually improve the relevance of the results. But the quoted passage seems to indicate otherwise.

[jrap]

I get a _lot_ of legit 'long tail' searches, and have seen a significant drop in traffic since the 20th.

[martinibuster]

using some form of feedback to gradually improve the relevance of the results.

That's the likely way the machine trains itself, using click logs.

[IanKelley]

No doubt there's much more to it but the basic idea seems straightforward.

If both concepts and phrases are interpreted as objects with numeric properties then their relationships to one another can be quantified without the algo needing to "understand" them (which is still a long way off).

The more time and data involved, the more refined the relationships become.

And newly created objects will often have numeric relationships with existing objects out of the box.

[timemachined]

AI = explains why the results are so bad sometimes, no substitute for human intelligence and thought - not yet anyway and when it is we will be long exterminated as a race. They have turned down the thesaurus algorithm though and offered more exact match and independent plural searches - which is quite annoying as was enjoying the similar match search but not so on returning results that had little to do with phrase I was searching i.e. the above.

We're all running businesses and they're playing with AI which hasn't the intelligence to think like a human yet. Which is often a thought process that doesn't have any obvious mathematical equation at all!

martinibuster
That's the likely way the machine trains itself, using click logs.

As you may well know, google already ranks on click throughs and final destination. Which is ironic, ok for them to steal all our data and the user's data to rank, but we can't have the keywords...

[fathom]

An interesting note, RankBrain DOES NOT sound like a totalian algorithm:

The system helps Mountain View, California-based Google deal with the 15 percent of queries a day it gets which its systems have never seen before, he said. For example, it’s adept at dealing with ambiguous queries, like, “What’s the title of the consumer at the highest level of a food chain?” And RankBrain’s usage of AI means it works differently than the other technologies in the search engine.

Sounds more like a way to provide more relevant pages to the newest of queries e.g. longtail.

[heisje]

Sounds more like a way to provide more relevant pages to the newest of queries e.g. longtail.

If so, it is failing miserably, as e-commerce long-tail conversions have tanked.

[martinibuster]

google already ranks on click throughs and final destination.

AFAIK, CTR is used for quality control, machine training and testing purposes. Even as far back as around 2003/2004, the only claim GoogleGuy made for CTR was for quality control. If you have a link to an authoritative source that shows it's an actual ranking factor then I would welcome reading it.

[timemachined]

I think there's an article on MOZ or another that alludes to tests being made on CTR and a ranking factor. I'll look later. From my point of view there is weight behind CTR and final destination for staying power. It makes sense if you have ten sites vying for first position and choosing which is best to rank on top of other algos. GA tells google everything even if you don't wish it too, which is why some don't have it installed at all.

[goodroi]

Be careful not to chase after symptoms and ignore the fundamental cause.

I don't care about RankBrain. It is horrible! It is great! Whatever. None of us can change Google so save your breath and energy. Rankbrain, machine learning and AI don't really change the long term SEO game (If you are looking to do some high risk churn & burn SEO it is a different story). It makes it more complicated but the game hasn't really changed. You can use an abacus, a pocket calculator or a spreadsheet - at the end of the day Google is still counting up quality signals to determine rankings.

The stronger you make your website, the more traffic you will gain from non-Google sources. When you legitimately boost traffic to your website you will see it leads to a boost in Google rankings regardless of what system Google is currently using to count quality signals.

Think about what the users want that isn't on 100 other websites and give it to them faster, better & easier than any other site. Listen to what your audience is saying via contact forms, industry forums, product reviews on other websites and then speak to your target market using the same vocabulary. Don't think about keyword stuffing but think about using relevant keywords that match your audience's vocabulary. Doing a better job will lead to referrals via email traffic, twitter mentions, facebook, blog posts etc. Engage and interact with your audience to grow even more mentions and loyal users referring others to your site. Those referrals will lead to backlinks and usage signals. Good content, good usability, good traffic sources make for a profitable website and has a nice side effect of often attracting good google rankings.

[fathom]

Agree goodroi!

SEO refinement is the finishing coat not the foundation.

[martinibuster]

It makes sense if you have ten sites vying for first position and choosing which is best to rank on top of other algos.

Of course it makes sense. There have been numerous studies by academics and scientists, both in Google and outside of Google. However the context has always been in the service of training an algorithm or quality testing an algorithm. The training part comes into play when investigating whether certain changes influence CTR and whether or not there's a correlation with user satisfaction. If user satisfaction works then the algorithm will learn from that and keep the change (a change that is incorporated algo-wide, within a particular niche topic or a user intent type scenario). Testing an algorithm is similarly about gauging user satisfaction with an algorithm. In both cases the changes are incorporated on a scaled basis, not granular or in real-time.

I've not seen a research paper or patent describing a real-time independent ranking adjustment based on CTR click log data.

[timemachined]

Search on

moz ctr rank

and also article

Queries & Clicks May Influence Google's Results

I think that's the one I originally read, of course it's all conjecture. With personalisation search going on (turned off), and the rest, one question I find myself asking these days is this.

Why when I view certain pages and search terms via google as a logged in admin, where my views don't count on internal stats, why do I then see traffic on or increase on the pages I've viewed during the day or days later? The above is the only answer to that which I can think of.

Seriously, I told someone else who is actually in seo and they found it puzzling. It went as far as me saying, "I clicked on my own links via google, then a day later not only did I get traffic but did a sale" It's easier to distinguish a pattern on a low traffic site or lesser searched term and article. I haven't tested it but it's a bit weird and keeps happening, though not always a sale at the end.

[mcneely]

Thinking in the simplest terms with regard to the intelligence of Google machining and now it's so-called AI, we might do well to remember that it's overall combined intelligence doesn't even get to that of the common house bug or insect.

It's all still a machine, programmed by humans, to employ the use of mechanical mathematics. Nothing really too impressive or smart about it at all. It's a menagerie of trips and triggers put on by the guy behind the curtain .. It's a new distraction that's meant to take away from the fact that Google really couldn't index the entire internet after all, much less present what they did index in a practical and useful way.

If the technology and expertise really existed to put on a genuine human kind of search experience, complete with all of the presumed real time human intelligence, then I'm pretty sure that two very obvious things would happen. #1 None of you would want it -- #2 Google would go out of business faster than you could say Search Engine.

[ogletree]

I agree with mcneely. Google makes money when people don't find what they are looking for. It is not in their best interest to be the best search engine. I would imagine that bad results increase PPC click through rates.

[blend27]

I get an exact phrase from AutoComplete, prediction works 99%. When the long(3 words) tail is selected I get 90% of organic links on first page results pointing to Amazon. I get the Artificial part of it, but not the intelligence part. This one was a very healthy niche few years ago.

Why, oh why? Also, Is Wu from the Tang Clan?

"Google Turning Its Lucrative Web Search Over to AI Machines". No it does not. Someone named Jack decided to write a PR article for GOOG.

[bumpski]

The stronger you make your website, the more traffic you will gain from non-Google sources. When you legitimately boost traffic to your website you will see it leads to a boost in Google rankings regardless of what system Google is currently using to count quality signals.

So the best Google SEO is NOT Google SEO, therefore, it would seem, there is no need for the Google SEO News and Discussion Forum! SEO must be in the name for some reason.

[keyplyr]

I am not a big fan of artificial intelligence. I still remember what happened with Clarke's H.A.L.

[seoskunk]

I am not a big fan of artificial intelligence. I still remember what happened with Clarke's H.A.L.

If you take the next letter in the alphabet after each initial H.A.L becomes I.B.M , Kubrick did enjoy his jokes, I wonder what he would of made of Google?

[goodoldweb]

This may come handy:

All about rank brain
[searchengineland.com...]