Terry Van Horne AKA Webmaster T

Exactly and that's why I believe gap analysis is risky it can suggest content that doesn't really add value to a site. IMO, if content doesn't provide a pathway to meeting a site "goal" then it's a waste of time and $'s.

IMO ranking for "free X" when you don't have a free plan just pisses users off

Agreed! AI links are about building trust & authority not ranking. Link building based on volume and not quality topical authority sites will not be effective. SEOs should be using outreach to niche authorities for link building.

ChatGPT actually has lost share of search as it is below levels of last year. I don't see a potential growth that others do!

IMO, users don't really want agents doing everything for them... it's all hype w/no factual evidence to prove to back it. I believe growth will happen offline not in search

Great post! One of the things I've been looking at is using .md files for AIsearch that prefer machine readable data.

IMO, if your concerned about "chunking" .md beats degrading user experience. Managing 400 token or less "contextual windows" is the real problem not chunking

Maybe they wanted to sell more tools so they could charge even more for their ridiculously overpriced toolsets!

If you're using visibility in LLMs as a KPI metric I question your judgement in the choice due to LLM visibility being random at best.

AFAIK occasionally things change on the web! They don't check them now quite possibly because such a low % of SEOs use them thinking they're some kind of GEO BS.

Say you used an accept header to just serve .md files to AIbots and included a link to the LLMs.txt where you had all your .md listed?

Ummm then I just remove the header! That's the simplicity in managing it. ;-)

I wouldn't count on any of them doing much about spam... more worried about prompt injections and STS which affects the LLM training. Most SE spam doesn't manipulate them enough to make worrying about it worth the costs

They aren't served to SE only to AI user agents using accept header.

Like authors and businesses Google wants as much info as they can get for Publishers to establish trust and perhaps to add Publisher icons etc. in SERPs

Google Zero is a huge exaggeration of what Google can do, however, an argument can be made that Google will do as much as they can to keep users on the site so SEOs must be mindful of that and act accordingly by evolving their tactics as Google adjusts in an effort to fulfill their goals

ChatGPT Sources Mostly From Wikipedia While Google AI Overviews Sources Mostly From Reddit A study from Profound of OpenAI's ChatGPT, Google AI Overviews and Perplexity shows that while ChatGPT mostly sources its information from Wikipedia, Google AI Overviews and Perplexity mostly source t...

Listicles replaced by Wikipedia makes sense given they have a significant share of ChatGPT sources
www.seroundtable.com/chatgpt-goog...

I thought you don't need schema?😉

Any thoughts on the old technical SEO debate "content delivery" vs cloaking?

Delivering markdown that's exactly the same content one part of the cloning risk is removed when the content the user sees is the same. However, if by happen stance SE check for .md:

.md= dupes or unwanted manipulation

I've decided that AEO is like a machine readable layer that is across all disciplines of SEO including technical SEO, on-page, Semantic SEO. 3 of those things in the list are for data extraction. Front loading of KW and important points was a tactic early on when Google was very theme related

I'm sure that's true... as client they're only responsibility is to pay the bill and ask the right questions.

SEOs who think clients should know anything haven't thought it through! If they know that much about SEO they don't need to hire an SEO 😇

Ummm not as bad as I thought at first... turns out it's a minor injection as it only affects the "user" clicking it. An injection that affected the whole userbase of the LLM would be a big deal. This IMO, is at best annoying to AI search and stupid SEO cuz certainly a huge hit on trust if detected

For the most part I've treated GEO as nonsense. In particular LLMs.txt cuz NO SUPPORT from AI platforms and markdown for several reasons but mostly it possibly being CLOAKING.

I'm diving deeper on both but NEVER will I use GEO term again! Its an occupied NAMESPACE in SEO jargon (see location pages)

I'd say IF an .md file makes your page more visible in AIsearch you've done a sub-par job of chunking. AI uses HTML elements that don't translate the same in markdown

Typical of Altman... the energy used in training humans is paid wholly by the beneficiary. Every user of the energy and resources AI uses are subsidizing their MODELS trained ENTIRELY on STOLEN CONTENT!

This quote from the post is so true!

"They do not reflect lived experience. They are competent summaries of what already exists."

IMO, the above is often the difference between being cited in AI w/a link or just being part of an answer w/no benefit!

Not to point out the obvious but Google has rolled updates to major updates starting with Panda.
IMO,These "phantom updates" (I think you started that term) are common. One of my oldest sites has disappeared and re-appeared a few times lately... why I don't obsess over volatility! it is what it is☺️

Unfortunately it's EAT or be EATEN scenario. Way more respect for Google than the "johnny CUM LATELYs" who stole content and have given nothing back. Google is a huge part of why the web is what it is today. Sure they made a lot of money but LLMs are largely their invention & they Open Sourced it!

IMO, Years of Experience working in multiple niches and types of websites. The longer you've done SEO the more CHANGE you've seen in how SE work especially if you aren't an exploiter but a believer in "feeding the machine" what it wants & how it wants it.

Counting occurrence was as much for old SEO since initial relevance is based on BM25 (pre-processing by SE) or some phrase based ranking algo. I am looking into expanding my knowledge on embeddings but I believe I cover embedding-based similarity because I use all entities discovered in the corpus

Most SEO KPIs are dead or losing value in the AI age... the skills required haven't changed much they just keep evolving!

Not included in the article (likely on purpose) was whether the injection only affected the user who pushed the button or whether it affected all users of the LLM. I'm thinking since it's in "memory" it's a single user.

Anyone tested this?

The Natural Language & Knowledge Graph APIs are good sources that I used in the beginning, however, ranking relevance is based on the QUERY CORPUS so using python or OpenAI to pull and count occurrences of all entities in say the top 10 results is more efficient way to do entity analysis

That's one of the things that it's hard to wrap your head around. If that's how it works the challenge is keeping content fresh w/o publishing the same thing over and over which will at some point kill your organic...

New content strategies will be needed to "feed the machine".

Yeah and he was going to have Greenland "one way or another" too and we know how that's gone!

Delusional thinking on his part!

IMO, not just limited to technical audits. Too often keyword research isn't properly prioritized beyond the query numbers. However, free widget may have huge numbers but if they don't have a free widget then it should have very low priority.

I prioritized all tasks in the audit