The introduction is about how Google Search is constantly testing, tying in a recent quote from Danny Sullivan plus a couple of Google Cloud articles relevant to how Search works.
“Of the proposed changes this past year, many never went live, because unless we can show a change actually makes things better for people, we don’t launch it.”
In 2022, Google Search also ran 13,280 live traffic experiments involving real users.
“We enable the feature in question to just a small percentage of people, usually starting at 0.1%. We then compare the experiment group to a control group that did not have the feature enabled. We look at a very long list of metrics, such as what people click on, how many queries were done, whether queries were abandoned, how long it took for people to click on a result, and so on. We use these results to measure whether engagement with the new feature is positive, to ensure that the changes we make are increasing the relevance and usefulness of our results for everyone.” [Highlights added.]
What are these live experiments?
They’re the kind of things we share on social and Barry Schwartz reports in SERoundtable.
Google announced SGE in May 2023, yet by January 2024 (8 months later), Google was still describing SGE as a “testbed for bold new ideas.”
As we can infer from the How Search works page (hopefully we’ll get 2023 data soon!), Google won’t push live a change unless it believes it will “actually make things better for people.”
As Roger Montti points out in a recent SEJ article about SGE:
“One of Google’s guiding principles for the use of AI is to only use it once the technology is proven to be successful and is implemented in a way that can be trusted to be responsible and those are two things that generative AI is not capable of today.”
He goes on to explain about Google’s use of AI in the background for years, including since 2015 with RankBrain.
On that front, I came across a pair of articles in Discover this week from Google Cloud that provide some information relevant to how Google Search works:
They specifically speak to the use of RAG (retrieval augmented generation), or information retrieval plus LLMs, in enterprise products like Vertex AI Search, “a fully managed platform for developers to build Google-quality search experiences for websites, structured and unstructured data.”
As SEOs, we often have to read between the lines, whether it’s in patents, Google Research papers, or, in this case, Google Cloud enterprise products.
Whether or not the technologies described in those articles are used exactly the same way in Google Search today, what matters is grasping their concepts for contextual knowledge.
Here are a few notable excerpts from both parts:
“Another common misconception is that semantic search is a relatively new innovation that has gained popularity with the rise of LLMs. While semantic search may be among the current hot topics, it’s actually the result of years of research and development.Google has been at the forefront of semantic search development for nearly a decade, starting with a strategic decision to invest in developing its family of custom, in-house AI processors back in 2013 — the Tensor Processing Unit (TPU). …”
“Beyond our expertise in semantic search, we have also spent over 25 years advancing keyword-based (or token-based) search technology. Vertex AI Search builds on this by providing a hybrid search engine that simultaneously performs both keyword and semantic searches for each query. The results are then merged and re-ranked based on their respective scores, combining the best aspects of both search approaches to fill in the gaps left by each. …”
“Approximate nearest neighbors (ANN) algorithms like ScaNN are proficient at quickly retrieving results in semantic search, but they are not as strong at scoring and ordering them precisely. Anyone who has ever used Google Search knows how important it is for the top results to be the most relevant to our queries and carefully sorted by relevance. …”
“That’s why modern semantic search engines use a two-stage retrieval approach to generate results. First, they use an ANN retriever to do a quick first pass and bring up results, and then apply a re-ranking model to fine-tune the results and make sure the most relevant ones are at the top — all in milliseconds. …”
“At Google, we have been utilizing Knowledge Graph in Google Search since 2012, helping add more context to search queries by providing information about things, people, or places that Google already knows about. Google Search leverages Knowledge Graph to tap into its existing intelligence and understanding of the web to find and return results related to a user’s search query, such as landmarks, celebrities, cities, geographical features, movies, and more.”
Source: Google
– Google Cloud, Your RAGs powered by Google Search technology, Parts 1 & 2
There are a few takeaways from all of this so far:
Google Search runs a lot of tests.
Google has a long history of using AI and semantic search technologies.
Updates to Google Search are for users’ benefit (for “everyone”).
When Danny Sullivan (Google Search Liaison) said this week that:
“Virtually everything someone asks about what Google wants, the touchstone is “Is this what your reader / audience wants?” Because that is what Google wants.”
“That’s the big picture. We want our systems to show content that people think will be satisfying. When they land on a page, they go “Yes! This is exactly what I want!” (And yes, our systems aren’t perfect. Yes, there’s unsatisfying content that can surface. It’s all something we’re working very hard to improve on).” [Highlights added.]
We can understand the impulse for that, as well as appreciate the myriad of ways that Google Search’s hard work “to improve” may be getting applied.
We can embrace this ourselves as SEOs.
Though I’m an independent consultant today, I spent 8+ years at agencies, and one consistent theme was a preference for new content rather than updating existing content.
However, newer websites or brands aside, I almost always make existing content optimization 50% or more of a content strategy.
On my website, I even take this to a crazy level. It’s not unusual for me to publish a blog post and then make 100+ revisions in days after.
But as I wrote about recently in a blog post (which I’ve already updated a ton), the question isn’t “Is this change good for SEO?” but rather “Is this change good for users, and how can we help Google’s ranking systems understand and reward us for it?”
One way to figure that out is to test, test, and test again.
Of course, it helps to keep track of when and where a change was made, so you can know if you’re moving in the right direction. 😉
Sometimes tests also have unintended consequences, as well.
Yet, from all of that testing, we can piece together themes.
One of the themes I’ve learned for my blog, for example, is to first write a post, then restructure it, usually by moving most of the introduction down to a conclusion.
Stay curious, buckle up for a full week’s recap, and enjoy the vibes:
Summary of the week’s SEO news and content
Danny Sullivan as Google Search Liaison on X gave suggestions for evaluating a traffic drop; also talked about misconceptions from the last helpful content update’s impact plus the system’s timing
OpenAI unveiled Sora AI model to create realistic video from text; also talk of a “web search product”
Google Search Console made some design changes; breaking a favorite Chrome extension of mine, but it’s back 🙂
European searchers may notice“more visual and entity focused results” on Google Search, given the legal environment
LOTs of talk about E-E-A-T, including new and updated posts
Enjoy!
Missed last week? Don’t worry, I got you! Read Part 44 to catch up.
Thank you for supporting Hamsterdam and helping make SEO accessible to all!
Also, feel free to support content you find valuable with a like, follow, or friendly comment.
Now, let’s step inside the white flags …
Top posts
These are highlights of news items, tips, or other content for the week. Great for when you just want a quick recap!
I would suggest:
1) Go to Search Console, select Search Results from Performance 2) Date range, Compare last six months to previous period 3) Sort the Queries report by Clicks Difference 4) Look at some of the top queries where clicks dropped the most 5) See if you're still in…
today we are starting red-teaming and offering access to a limited number of creators.@_tim_brooks@billpeeb@model_mechanic are really incredible; amazing work by them and the team.
These are newsworthy events in the SEO world, search engine updates, or SERP tests to be aware of from the last week.
This week we made some *design* changes to Search Console to improve accessibility and standardize its look and feel. Changes were across the board, did you notice? Check the filters in the Performance reports! pic.twitter.com/1Mlr1yPfsM
Google Announces New Search Experiences in Europe: Rich results, aggregator units, and refinement chips: "Users in EEA countries searching for queries such as "hotels near me" may notice more visual and entity focused results. Learn more: https://t.co/vg1O4ag33fpic.twitter.com/xlkYxLvcwm
Google now showing 'Forums' as a default menu item in the search bar? Not sure we need even more forum results in SERPs, but there you have it. We had a 'Web Results' test recently and now Forums – only a matter of time until 'Chat' appears. pic.twitter.com/9X9s5Z8Dhp
Can't go into specifics, but one of our accounts got some negative press that went into detail about what their product does — and it has been the best week they've ever seen for traffic and qualified leads.
Lesson learned:
Ignore the haters – any press really is good press.
Impacted the HCU? There seems to be a bit of confusion about if the helpful content update can impact Discover visibility. YES, IT CAN. Google even included that in their documentation about the HCU (see below). And here's a good example… Site flatlined in Discover after being… pic.twitter.com/tUdeZvBenG
Thank you. I appreciate the clarity here that I did not say two weeks.
I try to be as careful and precise with responses as I can, because what I say on behalf of Google Search carries a lot of weight. I (along with anyone at Google Search) really don't want people to…
When you start to think through the AI path we're on today, it's quite hard to envisage a future where the balance between publishers and gatekeepers is maintained. Google has been the gatekeeper with the world's most profitable business model. Disruption is coming, however.
Know thy audience Always do the research Craft multiple headlines Tell stories that resonate Create content that adds value Build the content you wish existed Distribute content on multiple channels
I have done a few before. Might do as time allows. The difficulty is that people can misinterpret as criticism (perhaps even as harsh criticism) rather than it really meant as "Hey, maybe this is helpful feedback" — especially as posts can be so impersonal sounding.
This recent blog post contains a lot of interesting information relevant to how Google Search works.
It covers semantic search, RAG, and hybrid search, also mentioning RankBrain and neural matching and a vector search infrastructure called ScaNN.https://t.co/pIb6sO1fJo
Google has confirmed that the driving directions anomaly was a bug. "We recently identified a bug that caused inaccurate numbers in the Direction metric for the timeframe Jan 17th – Feb 9th. The data has now been corrected, so if you downloaded the data for external use, please… https://t.co/rFvqWLgyEf
It's more if you accidentally 404'd something and fixed it. You obviously don't have to fix 404s that you want to be 404s. Also, this is more about tracking for you ("I fixed this, tell me when you see it fixed too").
Google's @searchliaison talks about timing for recovery after a Google helpful content update: – classifier is always running – recovery can happen between updates – but it takes months to prove your content is helpful more details at https://t.co/IUAFivHNJtpic.twitter.com/3nYyakYKBE
Here’s a recap of AI news, SEO tool updates, new tools, along with tips for reporting on data.
Annoyed by SEMrush and Ahref's prices?
Well, I got to the "meat and potatoes" of keyword research, and made an inexpensive tool to streamline your keyword research. Get 5 free reports, and get unlimited reports for $20/Monthhttps://t.co/jX04BnP8qM
We’ve just introduced trend change detection, a new type of insight that surfaces subtle but long-lasting and important changes in your data. It works like anomaly detection since they both detect a change in data. The main difference is that anomaly detection highlights sudden… pic.twitter.com/RYqzikf8r1
Today we're introducing Gemini 1.5, our next-generation AI model. It shows dramatically enhanced performance, including long-context understanding across modalities, which opens up new possibilities for people to create and build with AI → https://t.co/TjDy8GHIQS#GeminiAIpic.twitter.com/043FGirXB0
LOL, these cartoons had me laughing out loud. His latest is the future of search, but he provides several others like the future of ecommerce, advertising, internet of things, and personal privacy. Funny stuff 🙂 Here are two, check the post for the others:… pic.twitter.com/uPxAFADjcg
To whoever is in charge of the Thorntons website: if you're wondering why your Google traffic has dropped, it's because you've accidentally told them to stop crawling the site.
Does anyone truly understand how AI works? – ai.explored (TikTok)
Great job making it to the end. You rock!
Please show your support for creators: If you liked any of the content shared above, show that person support by liking their post or following them. 🙂
Newsletter, anyone?
Hungry for more? Check out the past weeks of Hamsterdam.
Or sign up for the newsletter, which gets delivered every Sunday!
I’m an independent strategist and consultant based in Orlando, Florida, focusing on holistic SEO strategies for brands. Currently, I’m taking new clients for 2024!
Feel free to visit my about page for more information about my background.
Let’s connect!
Hit me up anytime via text or call at 813-557-9745 or on social or email:
Leave a Reply