Mattybuns

SEO Strategy. SEM. Digital Marketing

[Cyrus SEO @CyrusShepard]

It's a strange time to do SEO

Many "best practices" no longer work like they once did, and new tactics around AI optimization can be hard to implement or have questionable longevity/ROI

At the end of the day, it's important to remember that these systems are designed to reflect the real world (as imperfect as they are)

So in times of uncertainty, your best tactic is to go back to marketing basics: make noise, solve real user problems, create assets, and create positive real-world attention

Do *real shit* or whatever SEO tactic you employ may ultimately be rendered obsolete

[Lily Ray 😏 @lilyraynyc]

This is so bad - I asked @perplexity_ai about recent SEO/AI news and it responded with a information about a non-existent, made-up algorithm update

Perplexity cited two AI-generated articles on digital marketing agency domains that made up information about an algorithm update that never happened, and used a fancy SEO terminology to make it sound official, or something.

The snake is eating its own tail.

[Chris Long @chris_nectiv]

SEO Tip: F5Bot spies on Reddit so you don’t have to.

Choose your queries and it will send you automated emails about new threads + posts:

[DEJAN @dejanseo]

LLMs do not crawl the web, do not maintain indexes, and do not enforce ranking algorithms at internet scale. They operate as presentation and reasoning layers on top of the classic information retrieval (IR) pipeline.

The recent paper Why Language Models Hallucinate (Kalai, Nachum, Vempala, Zhang, 2025) shows why this distinction matters: LLMs inevitably hallucinate due to statistical limits and evaluation incentives. Without grounding in real retrieval systems, they cannot provide reliable search.

https://t.co/Ppkd2Y8bLx

[Cyrus SEO @CyrusShepard]

Wow - court rules Google must share data with competitors. Looks like they'll be forced to share:

1) Basic Search Index: URLs/first seen/last crawled/spam score/etc.
2) "Glue": uses click data to build SERPs
3) RankEmbed: deep learning to understand semantic meaning

Bing, DuckDuckGo, and OpenAI will have a field day with this

[Chris Long @chris_nectiv]

OK so huge news today.

I'm launching Nectiv an SEO/GEO agency that helps companies thrive in the AI search era:

[Barry Schwartz @rustybrick]

Finding long conversational query data in Google Search Console https://t.co/1oPLv2lX5r via @metehan777 @VijayChauhanSEO @da__kon

[Mattybuns @bergerm57]

@mattmerrick16 niche

[Mattybuns @bergerm57]

@ItsAustinAI Sent

[Mattybuns @bergerm57]

@Zephyr_hg Flow

[Mattybuns @bergerm57]

@mattdiggityseo YES

[Mattybuns @bergerm57]

@adcrookes Reddit

[Mattybuns @bergerm57]

@Adam_DelDuca go

[Chris Long @chris_nectiv]

Huge SEO news: A new Google exploit was discovered that revealed 2,000+ classifiers for click probability, site_quality, consensus + more:

[Ori Silver @OriSilver]

Google's elephant in the room is the fact they cannot differentiate between types of content, there is no discerning factor to allow them to know that out of 10,000 pages that say almost the same thing one of those is the actual best (expertise and authentic etc.)

The system itself is the reason so...

Google systems work on detecting what good looks like, the system does not intrinsically know what is quality and what not, it is just good at interpreting user behavior as a response to results (sort of...)

If you know how to detect satisfaction and do it well, you don't need to know if a quiz, an article or a module appears for a query, all you need to do is get really good at detecting user behavioral patterns and act accordingly.

You can then use prediction models to shape and refine results even further. In the world of pre LLM and AI tools, this method worked pretty well.

You cannot satisfy all the users all the time but you can satisfy most of the users most of the time.

And they did, mostly. Slowly by combining anti competitive practices (not my words the DOJ's verdict) they were able to become the "best search engine".

In an age of LLM though where low quality, low effort, mass scaled content is the norm, the method of knowing "what good looks like" fails, because everything looks, feels and says the same. So what is the discerning factor?

In my humble opinion, they dialed up the user satisfaction weight in order to partly solve the problem. If a brand is trusted by the users and also gets a certain quality score, they get to appear and everyone else that do not match the threshold, vanish.

This can explain why all of a sudden, the only websites showing to the most part, replacing all the smaller publishers are big names with huge money behind them.

This is a panda'esque change to their system, a deeper layers of filtration which allows to eliminate all the "look a like" pages claiming they are experts.

Google loves brands is an old trope but it is the soft belly of the system, there is no way to "solve" it, because reputation and user trust is number one and even though Forbes is all for profit, users might want to read that rather than https://t.co/A6fEExoxab and in many cases users got so fatigued by the plague of endless affiliate articles they go to Reddit, to read real experiences.

[Mattybuns @bergerm57]

@Zephyr_hg GPT

[Mattybuns @bergerm57]

@godofprompt GPT

[Mattybuns @bergerm57]

@LB_BurtonSr @MKBHD Initially thought you were referring to his note after removing the clip, simply mentioning it "added nothing to the video", which I did think was a bad look. Seeing his real acknowledgment, I'm more in line with your comment. Well said

[Mattybuns @bergerm57]

@CyrusShepard I know there was a lot of speculation that Lina Khan would get canned if Kamala had won too given many of her doners were pushing for it similar to Trump's higher net worth donors.

Unfortunately, Lina Khan's job security wasn't looking great no matter who won.

[Mattybuns @bergerm57]

@GeorgeBounats list