Hit by may update

tazarbm said:

this makes sense, but then how can we explain the downfall in search quality that we've been experiencing for the last several years.

I've personally not used google (extensively, because otherwise I do use it once in a blue moon when the other SEs don't provide good results) for years, and part of that (despite of my non-dying hate for this company) is because of the quality of their main product (well, main product as perceived by its users, otherwise google are probably considered something different than search as their main product).

So, it's not just my never-dying hate towards this company, it's also the quality of their main product, which means that me (and most definitely many other users too) are not satisfied by the google product anymore as lots of us have gone to other SEs, which means that things are not quite right despite of the numerous updates which should improve the product, right? Not make it worse...

Click to expand...

That's another discussion and that's waaaaay too broad to try to pin down to embeddings.

The whole search landscape has changed. The whole world has changed! post-covid, post-AI. There's so many confounding variables. We couldn't possibly attempt to pin this down on the theory of embeddings for ranking.

tazarbm said:

when you say "by seeding" I don't suppose you mean that google are manually labeling sites as being trustworthy, right? This would be ludicrous, but for me - as a non-native English speaker - that's what the word "seeding" means... which is probably where my confusion stems from...

Click to expand...

It's absolutely manual.

It's been like this for a long time.

But think about how it works. *seeding*

If I have a room with 1000 people, and all those people are going to talk and share info and recommend each other, who do I know who to trust?

I could manually rank them all.

Or

I could seed 10 people as trusted.

The people those people recommend, I can trust.

The ones those ones recommend I can also trust.

And so on. The further away I am from the seeds, the less trust.

This is how it works.

tazarbm said:

no idea what those embeddings are, it's the 2nd time I hear you mention them, but I've never heard this term before, so not sure how to interpret it (but don't explain if you don't want to, it's ok!)

Also, from reading your entire post it looks like this centroid thing is the real actor in this whole google ranking thing, everything revolves around that centroid thingy from what I understand

Anyway, you mentioned a few technical words and I'm already mentally exhausted, so not going to reply to the rest of your post, but thanks for taking the time to explain, I'm sure that less tired and more techy people will be able to understand what you've said here

Click to expand...

It's not an easy one to understand over a couple of posts unfortunately.

It's a machine learning term.

The answer is an embedding is "a representation of something in high dimensional vector space."

splishsplash said:

The further away I am from the seeds, the less trust.

Click to expand...

I was about to say this ... And I think that this justifies (at least partly) why newer sites manage to break into the top 10 (this, combined with the targeting of long tails and other minor - but seemingly important - site / page optimization tricks)

splishsplash said:

This is how it works.

Click to expand...

thanks! Honestly!

And hopefully I don't forget this next time when I reply to posts because boy, am I a forgetful f**k!

splishsplash said:

It's not an easy one to understand over a couple of posts unfortunately.

It's a machine learning term.

Click to expand...

all good, man! I still don't believe in machine learning (integrated circuits can't learn shit on their own, they do what they're programmed to do, ie. acquire more information and then probably hierarchize and execute data based on the updated information, but that's not learning by themselves, that's programming, plain and simple), but it's ok!

splishsplash said:

The answer is an embedding is "a representation of something in high dimensional vector space."

Click to expand...

see? This is the shit that makes me go crazy every time I think about getting back into SEO. This technical jargon is just not for me, I've failed all of my math classes, I usually excel at the humanist part of life (righteousness, kindness, empathy, that sorta crap), technology is just not my thing and period!

To be honest I kind of was pushed into SEO now that I think about my past, but that's a story for another day...

I go eat, thanks again for the wonderful explanations, I am sure that others will find this info useful for their SEO projects, for me it's just headaches

Peace!

masterweb2 said:

Hello

Two of my websites, which are on completely different topics, have seen their traffic drop drastically after Google's May update — down to one-third of what it was.
View attachment 526979
The site for which I shared the GSC screenshot is about two years old and has 1,500 pieces of Non english content, which were written and Resaerched with AI but manually edited.
It gets updated daily.I've been very careful with backlinks it has around 150 real, manual backlinks. ( for example i have 5 wikipedia backlink) I also worked on EEAT and have 7 real authors with social media profiles linked.
Internal linking is done very carefully, manually, and is precisely relevant to the topics without using exact-match keywords.
All social media accounts exist and have content (though I haven't updated them for 3 months).
Currently, traffic has dropped very strangely and dramatically, and I don't know exactly what to do.
Now that the update is over and the results are clear, should I just wait?
What changes would you suggest? what i missed?
thank you

Click to expand...

You cant recover. The site has been added to a BLOCKLIST. It's over
Ditch the domain

Nikolaos said:

He said he humanies AI content. His content should be fine.

Click to expand...

I'm thinking about the possibility that if the content is detected as AI by Google, even if it has been manually edited.

Steptoe said:

External signals - have brand signals and engagement suffered recently? I notice you say you haven't updated social in a few months
Content - is your newer content drifting from what you were successful with before?

Click to expand...

Yes, I haven't updated social media for about 3 months, but before that, it was updated daily.
Regarding the articles ,yes, the older ones definitely performed much better. Back when I used to find related questions or related topics, after publishing, I would easily rank #1 for some of those questions.
I still have some pages that are still #1, but many have dropped significantly.
Here's a very interesting point: when I check the sites that are now ranking #1, they have Spammy backlink profiles or non related websites guest posts , comment backlinks, 100% follow links, exactly like the ones we used to build with GSA. Meanwhile, as you can see, only 19% of my backlinks are follow and manualy created.

I've been in SEO for almost 10 years, but honestly, I feel like I don't know anything anymore. I'll try my best to improve the situation by publishing manual analytical articles and updating social media. But if nothing changes, I will definitely implement the old black hat methods. Why not? It's a case study.

From what you've described, it doesn't sound like you've ignored the usual SEO fundamentals. A two-year-old site, consistent publishing, real backlinks, author profiles, and careful internal linking should normally provide a solid foundation.

masterweb2 said:

Here's a very interesting point: when I check the sites that are now ranking #1, they have Spammy backlink profiles or non related websites guest posts , comment backlinks, 100% follow links, exactly like the ones we used to build with GSA. Meanwhile, as you can see, only 19% of my backlinks are follow and manualy created.

Click to expand...

if you cant beat them join them

masterweb2 said:

I'm thinking about the possibility that if the content is detected as AI by Google, even if it has been manually edited.

Yes, I haven't updated social media for about 3 months, but before that, it was updated daily.
Regarding the articles ,yes, the older ones definitely performed much better. Back when I used to find related questions or related topics, after publishing, I would easily rank #1 for some of those questions.
I still have some pages that are still #1, but many have dropped significantly.
Here's a very interesting point: when I check the sites that are now ranking #1, they have Spammy backlink profiles or non related websites guest posts , comment backlinks, 100% follow links, exactly like the ones we used to build with GSA. Meanwhile, as you can see, only 19% of my backlinks are follow and manualy created.
View attachment 527165
I've been in SEO for almost 10 years, but honestly, I feel like I don't know anything anymore. I'll try my best to improve the situation by publishing manual analytical articles and updating social media. But if nothing changes, I will definitely implement the old black hat methods. Why not? It's a case study.

Click to expand...

It's not about your backlinks. You're focusing on the wrong thing.

Google are very rarely penalizing sites for certain types of backlinks in 2026. It's not a good tactic and it makes it quite easy to neg your competitors if they focus on that.

Literally, what's stopping me just building 100's of semi-legit looking, but spammy backlinks over a 3 year period to my competitors? How can anyone tell if it's them, or if it's neg seo.

Actual neg seo is not the retarded "buy viagra" ultra-spam. That's what kids do who are upset they can't rank.

Even on link spam updates, you won't get penalized for bad links. What sometimes happens is they discount links they perviously counted.

Generally, with links, they just determine "Is this a legit link? Yes, pass juice, no, pass nothing" -- The bar isn't high. Link building is the norm.

The front of the war is at user satisfaction.

Think about it like this:

If you have a site that's satisfying users. They get the information they want. They find what they need. "Goal complete". Dopamine achieved - at the simplest most raw level.

Imagine that site has spammy'ish backlinks.

Then you have a complete garbage site. Ultra clean backlinks. Users hate it.

Should they penalize the quality site, that users like, and rank the ones they don't? Makes no sense.

The war on links was in the days when they had no choice. Links were the differentiator.

But..

Send me a pm with your site. I'll run my intent tool on some of your pages and see how far you actually are from what Google wants and give you my thoughts on where you've gone wrong and why you've been hit.

I don't normally do that, but, you said you've been in seo for 10 years, and it sounds like your site was probably pretty decent.

Most people that get hit have garbage sites, so it's a total waste of my time to look at them. It's like an obese man asking why he can't run a sub 3 hour marathon. I'm more interested in the 30 year old who's been training for 10 years in marathons, is in great shape, eats well and still can't get the sub-3 hour. That's interesting to look at. So you're welcome to pm me and I'll take a look.

splishsplash said:

The unfortunate reality is if you're hit with an update, you absolutely aren't going to be able to make some changes and recover in any short period of time.

The update isn't finished rolling out, so you may pop back, but generally, you're looking at fixing what caused you to be hit in the first place, then waiting for the next update.

EEAT isn't a thing. It's propaganda by Google and loved by SEO experts who have zero-clue and love to spout garbage that companies eat up because it sounds right.

I can't give you exact examples, but have a test site in an ultra hard finance keyword ranking #1. Not top 3. Number 1 for almost 2 years now beating stripe and various DR 85+'s

Fake phone number, address, not a real business, no real person, no trust signals, no EEAT.

It's all about your page embedding. If your embedding isn't close to the centroid, you don't rank.

Literally no one in the industry understands SEO anymore.

It started with the HCU. This was Google changing from an old keyword/text based model to embeddings.

Updates are this:

It's Google batch re-training events based on behavioural data.

Each query has a centroid embedding. The closer your page embedding is to that centroid, the better your chances of ranking.

Close sites = in the race.

They're then re-ranked based on power, topical authority, backlinks and engagement signals on the live serp.

That's SEO in 2026.

Click to expand...

This is gold holy smokes.

So updates are essentially Google changing the centroid based on new data.

If you were as close as possible to the centroid before the update the new one might be ever so slightly different so your rankings tank.

Interesting.

The logical step would then be to align to the new centroid.

Until it changes again.

Rinse and repeat?

As for EEAT I was under the impression that Google is looking for real entities so making branded social media pages reinforces that you actually are a real brand and not just a spam site.

I mean why would Google introduce social channels into search console: https://developers.google.com/search/blog/2025/12/social-channels-search-console

Vynith said:

This is gold holy smokes.

So updates are essentially Google changing the centroid based on new data.

If you were as close as possible to the centroid before the update the new one might be ever so slightly different so your rankings tank.

Interesting.

The logical step would then be to align to the new centroid.

Until it changes again.

Rinse and repeat?

As for EEAT I was under the impression that Google is looking for real entities so making branded social media pages reinforces that you actually are a real brand and not just a spam site.

I mean why would Google introduce social channels into search console: https://developers.google.com/search/blog/2025/12/social-channels-search-console

Click to expand...

Yeah. They are updating for every query.

If you have 2 queries like

best toasters and top toasters, they each have a centroid, but they're both incredibly close, so you'll rank for both of them with the same type of page.

Your goal is to find the mid-spot for the main keyword group.

But there are ways to actually build a page blueprint for what users really want.

Google are doing this with simpler models, but for an individual keyword and associated group we have access to rich data to analyse. They need to do it at scale based on the signals they have.

We don't need to actually embed and replicate what they do. We just need to predict what the idea page blueprint looks like.

Ie, it's a page blueprint. It's not "what keywords to use and how many times and in what headings" like tools like surfer do. That's 2018 style seo.

Yeah everyone thinks they look for real entities, but they're not. 100% no. I just can't show you examples because it's private sites and only someone behind the sites would know it's fake. But fake ones are ranking for major finance keywords. Fake everything. Beating DR 90's

Why do they have social channels? Why does anything do anything.

You can't just conclude google are verifying signals on your page just because they have a social signals on the console.

They're just giving you data.

And there's no reason that traffic they can measure coming in from social channels aren't used as ranking signals.

None of that is in dispute.

I would put this in the same bucket as backlinks, not on-site ranking signals. They are separate.

If your page is close to the centroid, you're still competing with others who are close, and there's where links and potentially social activity come in to boost. Popularity signals. Links are only popularity signals.

splishsplash said:

Yeah. They are updating for every query.

If you have 2 queries like

best toasters and top toasters, they each have a centroid, but they're both incredibly close, so you'll rank for both of them with the same type of page.

Your goal is to find the mid-spot for the main keyword group.

But there are ways to actually build a page blueprint for what users really want.

Google are doing this with simpler models, but for an individual keyword and associated group we have access to rich data to analyse. They need to do it at scale based on the signals they have.

We don't need to actually embed and replicate what they do. We just need to predict what the idea page blueprint looks like.

Ie, it's a page blueprint. It's not "what keywords to use and how many times and in what headings" like tools like surfer do. That's 2018 style seo.

Yeah everyone thinks they look for real entities, but they're not. 100% no. I just can't show you examples because it's private sites and only someone behind the sites would know it's fake. But fake ones are ranking for major finance keywords. Fake everything. Beating DR 90's

Why do they have social channels? Why does anything do anything.

You can't just conclude google are verifying signals on your page just because they have a social signals on the console.

They're just giving you data.

And there's no reason that traffic they can measure coming in from social channels aren't used as ranking signals.

None of that is in dispute.

I would put this in the same bucket as backlinks, not on-site ranking signals. They are separate.

If your page is close to the centroid, you're still competing with others who are close, and there's where links and potentially social activity come in to boost. Popularity signals. Links are only popularity signals.

Click to expand...

Damn the eeat propaganda got me good!

The page blueprint idea is gold.

I'll be thinking of that now when doing SEO!

The popularity signals explain why parasite SEO using Reddit and other juicy sites works so well.

They already have massive links and authority so it becomes a game of matching the centroid to the search term/intent.

I've literally never read anything about this on the main SEO sites I follow so thank you so much, sir.

I see why you have executive status here.

Vynith said:

Damn the eeat propaganda got me good!

The page blueprint idea is gold.

I'll be thinking of that now when doing SEO!

The popularity signals explain why parasite SEO using Reddit and other juicy sites works so well.

They already have massive links and authority so it becomes a game of matching the centroid to the search term/intent.

I've literally never read anything about this on the main SEO sites I follow so thank you so much, sir.

I see why you have executive status here.

Click to expand...

I'll be writing a killer guide with my full reverse engineering of google theory for people to use to create page blueprints that rank.

No one in the entire SEO industry is at this level. They're still years behind.

Over the past three days, I published three articles, one per day. I tried to move away from the style of my previous articles, completely remove any obvious AI-like footprint, and change the writing style as well.
I also chose more secondary/long-tail topics to see how they would perform in rankings. One of them, which had low to medium competition, ranked 8th. Another one appeared on page 3 of Google, and the third one has been indexed but has not gained any visibility yet. I should mention that they were only indexed today.
After checking GSC over the past few days, I can still see queries where my site ranks in the first position (low-to-medium competition, mostly low). However, my traffic has dropped to exactly one-sixth of what it used to be ( 4.8k to 800 clicks)
Now I want to know where I should start in order to identify the cause. I have already done a malware audit, especially considering the recent cPanel hacks, and there was no issue. I am also currently running a technical audit.
That said, I doubt the problem is technical. The only possibility that comes to mind is scaled content abuse in google's eye. But I still do not know where to start in order to find out what the actual problem is. Should I go back and revise/edit the pages that used to rank well based on the queries they were getting Or should I focus on reducing the AI-like footprint.

splishsplash said:

I'll be writing a killer guide with my full reverse engineering of google theory for people to use to create page blueprints that rank.

No one in the entire SEO industry is at this level. They're still years behind.

Click to expand...

Thank you, @splishsplash for sharing your insights and knowledge. I always appreciate it when you take us along into the fascinating world of SEO.
I'm also looking forward to the SEO killer guide you are going to write.

jorry said:

Thank you, @splishsplash for sharing your insights and knowledge. I always appreciate it when you take us along into the fascinating world of SEO.
I'm also looking forward to the SEO killer guide you are going to write.

Click to expand...

It is coming. I'm still working on my SaaS so I don't have time to spare right now, but once I've built what i need to build I'll be writing it.

You need to clone yourself and send me an v1 tom clone.
Trade it for my mother in law (not an clone)

But thanks again this is all gold!