Google Outlines Extensive Developments In Search and the Impacts on Ranking

[engine]

Every SEO should read this document and understand how Google has developed search, and how AI is impacting search.

Here are a couple of snippets.

BERT is now used in almost every query in English, helping you get higher quality results for your questions. We’re also sharing several new advancements to search ranking, made possible through our latest research in AI

Very specific searches can be the hardest to get right, since sometimes the single sentence that answers your question might be buried deep in a web page. We’ve recently made a breakthrough in ranking and are now able to not just index web pages, but individual passages from the pages. By better understanding the relevancy of specific passages, not just the overall page, we can find that needle-in-a-haystack information you’re looking for. This technology will improve 7 percent of search queries across all languages as we roll it out globally.

[blog.google...]

[JesterMagic]

Thanks for the post. Some interesting info on sub-topics etc...

IMO though Google search is worse than it was 10 years ago. I find Google generalizes my query to much now and makes assumptions about keywords in my search string which are incorrect.

Maybe they have made strides and part of the problem is spam has increased dramatically and become much more detailed making it harder for them to find the good pages.

The one noticeable good thing with search now I have to admit is Google does understand youtube videos better (must be the closed captions). Of course this is a monopoly understanding their own content which is not fair to other video sites.

I like how the document praises organic search but doesn't mention how great all the extra ads (that are usually so generic it is laughable) that are shown to users along with all the useless Google widgets designed to keep you on Google. The reason Google in the post does not mention the full layout of the search page is because the layout is what is best for Google, not the user.

[graeme_p]

@jesterMagic I feel them same about over generalisation. I think its worse for people who know how to make a query specific to what they want, but better for people who do not. There are far more of them than use (not on WW!).

[Wilburforce]

IMO though Google search is worse than it was 10 years ago.

My opinion too. DDG obviously doesn't use BERT, and is often (usually) better for my own searches.

I'm not sure that this necessarily represents a fault in Google's approach, however: what Google offered to early web users was a tool that enabled an intelligent choice of search terms to query a large dataset rapidly and yield relevant results.

Since then, the web and the number of users have massively increased, and I think Google's changes probably reflect that democracy. It is no longer a few nerds like us (the bulk of users in the mid 1990s) querying the dataset, it is nearly everyone. Where most users were once concise and accurate in their choice of search terms, they are now verbose and vernacular. That Google has retained its place as the default choice of search tool (and that "google" has become an internationally recognised verb) probably reflects the fact that for most verbose vernacular searchers, the results are at least OK, and probably better for those users than the competition.

What this means for SEO is debatable, but some things are written on the wall in large letters.

My own view is that SEO has always really been an attempt to game the algorithm - however we might dress it up - and in seasonal terms the days when that worked are fast heading into winter. What we need for web success now is to focus almost exclusively on the user. Does your site look good, does it load fast, is it easy to navigate, is the product your user is looking for easy to find and - OK, keywords - clearly labelled? Are your descriptions accurate, are your information pages helpful, are your clients happy? How many times you use the word Luxury Products in your shop window isn't what will lure customers away from Harrods.

That, oddly enough, is what Google has been telling us for some years.

[iamlost]

Back in ye olden days SEs used ‘keywords’ and ‘key phrases’ which were pretty much along the lines of

+webmasterworld +engine [phone]

which typically returns engine’s webmasterworld posts from 2005 but not 2003 and 2004.

In this keyword universe word order was usually seen as in descending order: webmasterworld being first filter, engine being second. If the order was switched often the results varied greatly; engines the machines might push engine the WebmasterWorld member out of or down the results.

Then, 2010 on, keywords began to be replaced by entities (keywords are entities, entities are not necessarily keywords), which also come with associated relationships. With, from 2012, mobile voice search, deciphering NL (natural language) queries became increasingly critical. Enter RankBrain in 2015. And more recently BERT in 2019.

Perhaps the most fascinating bit of any automation is how seamless it is, how unnoticed, when it works. And how blatantly bad when it fails. There is no try in ML there is only succeed or fail. No shades of grey, no almost.

Which is why so many webdevs say Google query results are increasingly bad - when they are bad they are so very very bad. And so very very obvious.

The deep dive query results aka ‘page part’ indexing are simply a ‘gone public’ rendition of how G has long been serving answers, snippets et al. Put simply Google answers are overflowing their box and assimilating the entire query results page.

Etc. Et al. Ad nauseum.

[JesterMagic]

I agree with those above that Google's generalization of the search query probably works well for most users who are looking for straight forward answers but for those doing any sort of research it starts to fail.

In Search with mobile phones and Discover I think Google is there own worse enemy. They award new content (when the old may be perfectly fine). I hardly use discover any more (I read a lot of tech articles) as I am tired of scrolling through articles that rehash the same thing and don't even answer the question in the title, or if they do it basically boils down to the fact they don't know. There are a lot of news publications in the tech world I use to read and respect who have either been bought out and/or has flipped to the format of more articles is better even if we are not answering any real questions.

Because of this, I have been using Youtube more for news now. It is still relatively easy to find channels on Youtube in topics you are interested in where someone (or a small group of people) is passionate about the topic and actually does unbias reporting. Of course this is changing with the rise of larger sponsorship deals so we may end up seeing Youtube results looking increasingly more spam like

[aristotle]

If someone is just looking for one little tidbit of information, they probably wouldn't spend much time on your site anyway. Also, for this type of query, you can often find the answer you want in the snippets that google shows in its organic results. This isn't new -- it's been possible to do this for years.

[graeme_p]

Because of this, I have been using Youtube more for news now.

So instead of using Google to find news on other sites, you are getting all your news from a Google site.

[JesterMagic]

Yes of course I know Youtube is owned by Google. I was really referring just to tech news and how it is still fairly easy to find unbiased news on Youtube especially once you find a channel and subscribe to it.

Discover is a feed that molds itself based on your history and while it was ok at the beginning it is now IMO much worse because websites are falling over each other just to publish anything to get new content in Discover for clicks

So instead of using Google to find news on other sites, you are getting all your news from a Google site.

While all the content is controlled and hosted by Youtube (Google), the content itself is obviously made by others. Just like Discover since most (or all?) content returned uses AMP which is operated by Google.

I think what I miss is the old days of RSS readers where I just picked which websites to follow. It was great for getting news from a select sources (but not for finding new content)

For world and local news I tend to listen to the radio while in my office or car. If I want to learn anything more about a topic I would then just look it up on the internet.

[glakes]

BERT is now used in almost every query in English, helping you get higher quality results for your questions.

When Google released BERT, within a week converting traffic from Google dried up. We ended up letting staff go and didn't move nearly as much product as we anticipated. It was a dark time for us since BERT was released in our busy season. Things have certainly improved over time, so I hope this year won't be another repeat. Still, I find it difficult to believe all this tech really improves search when Google is flopping with basic stuff like indexing, canonicals, crowding useless pages (ie. Pinterest), etc.

[pims]

Yes, Google is continuously changing its algorithm for ranking and search factors, but still struggling with indexing issue. I really liked the recent google updates regarding the ranking of few lines from any post and quick answer snippet.

[chesterloke]

I remember few years ago when it was very difficult to find the information you were looking for, especially when previously a long query returned very minimal to no results and you had to broaden your search terms.

[saladtosser]

I wonder how this will impact information sites that rely on AdSense and AdSense similar advertising. I've clicked a growing number of google result links recently that when clicked (in chrome) immediately jump/scroll to the point of the page that tidbit is located, bypassing scrolling past the ads... Since this, I'm seeing more sites fixing ads to footers to ensure at least one ad is seen.