the footer does have stuff that I always considered to be EEAT signals, though (the Refunds page for example I've heard almost a decade ago that it's an important page for ecommerce sites and that it raises the trust score of the site).
Also, the "Trust" title itself of the 3rd footer section, there has to be something about that too. I mean, Lab reports, reviews, Ingredient standards... IMO all of these sections contribute to the increase of the trust scores of this website (therefore, to the improving of the EEAT score), and whether or not the content of these pages is true it's a different story. But the fact that this website had to create those pages tells me that EEAT signals are still a thing... I'm not sure that a similar site (same niche, same DR and approximate number and quality of backlinks, same or similar content quality and optimization, etc) would rank as well as this site were those footer credentials not there...
While I do believe you that you can rank with fake phone number, staff page, etc (maybe even with fake location too), I will still also continue believing that (at least in some cases) EEAT is needed. There has to be some exceptions for this, just like there are many exceptions with many things google-related
There isn't a trust metric.
Google doesn't work like that in 2026. They're so far beyond using metrics for the core ranking now.
That was old school SEO, in the days of "use the main keyword 3 times in the h2's" and such. These are actual 'ranking factors'.
Now, it's machine learning. It's embeddings.
There isn't 1 mathematical algorithmic rule that looks for certain footers, certain words, certain structures, fb pages, yt pages etc.
With each update they're adjusting the models with new data.
They train them to detect the kind of pages and sites that satisfy searchers. That's their primary goal. Satisfied searchers = they come back to google and search more. If people stop being satisfied, they literally will stop using google. You don't use things that don't satisfy the need you need satisfied. (Sounds a bit kinky when you keep saying satisfy!)
There are definitely trusted domains and a seeding mechanism, and trust does flow from there. But there's not trust signals on a site/page basis looking at things on the site/page. Trust can only be gained reliably through links and by seeding initial trust sites.
Each search query has an embedding for an ideal page. This is called the centroid.
The pages that are close to that compete to rank.
So here's the way to understand this:
In the past, Google could only really measure 'relevance' directly by looking at things like keyword usage, h2's and things that you can measure.
They could have in the past looked for certain indicators for trust, and maybe they did.
But today, everything is rolled into embeddings.
The embedding captures not only RELEVANCE, but also trust!
Here's a question:
You're a human.
As a human, can I show you 5 websites and ask you which of them look trustworthy and which don't, without you physically checking if the phone numbers are real?
If you look at a man, and a woman, can you tell me which is the man?
How?
Do you know? Not really..
You can try to quantify it, but you just SEE it.
You can SEE based on all the sites you've seen which look trustworthy and which don't. You can SEE if someone is a man or a woman. (Usually, but this is 2026!)
Embeddings are the same.
They capture relevance, meaning, tone, content, facts, style, trust. They capture EVERYTHING. They are massive tensors representing the page in high dimensional vector space.
Based on real data, Google learns the ideal embedding.
It can then create an embedding for your page, and it does it for your site too to determine other things btw.
It can then simply measure the distance in vector space from your page, to the centroid.
If your page contains high relevance signals, but low trust signals, you fail.
Maybe last week it ranked fine, but they changed the centroid. New data.. New learnings.
Suddenly, your page goes from X distance from the centroid to Y and you drop.
Maybe your site as a whole for the niche goes from X distance to the centroid for the niche to Y, and you're no longer considered a good match for this niche.
They are using multiple different types of centroids. Centroids for pages and your site.
They have a metric called siteRadius and siteFocusScore -- They use the word 'radius' - That's geometry.
If they weren't using embeddings, they wouldn't use that word. It wouldn't make sense. You can't measure the radius with keyword counts.
But they're Google. They are the inventors of the transformer model. They have countless amazing models. It would make no sense for them not to be extensively using the best possible tool available to rank pages, which is machine learning.
