Google’s latest documentation update on “Optimizing your website for generative AI features on Google Search” [1] was a fairly decisive blow to many of the tricks and hacks purported by AEO and GEO wizards.
Google basically spelled out what does and doesn’t matter for AI-based generative search.
Most seasoned SEOs have been saying this for a while “Good SEO is good AEO & GEO” however its nice when Google so concisely states it
Q. “Is SEO still relevant for generative AI search?”
- “In short, yes!”
In 3 words Google settled the debate.
Google even went a step further to hopefully and finally put to bed the AEO vs GEO vs SEO debate.
“optimizing for generative AI search is optimizing for the search experience, and thus still SEO”
You can’t get much clearer than Google’s documentation explicitly stating it.
There are some AEO & GEO gurus & AI bros out there wishing this document had never been published.
The Basic Principles use on AI Based Search
Another enforcing statement backing up what I have been saying is “Apply foundational SEO best practices to generative AI search”
This stands to reason, which is my acid test for anything in SEO.
Much of AI search draws from existing search indexed, although the base level LLM is fundamentally different, most AI searches are augment by a grounding technique referred to as RAG – Retrieval Augement Generation – said another way – the AI search has to go and do actual web searches, pull back a bunch of data and use it to help contextualize the answer.
This gives rise to a further technique known as query fan out.
Google addresses both RAG and Query Fan Out in this documentation.
Its really hard to get away from the fact that most AI search results rely on the traditional web search mechanisms, and its really hard to get away from the fact that required good foundational SEO best practices.
Fortunately – thats what we do so we got you covered on that front!
AI / LLM Based search is a different surface to optimize for, but the foundational best practices of organic search hold true and for good reason.
Now, this documentation is pretty clear and does put to bed much of the nonsense spouted by many fake gurus, however one area of the documentation does get a bit more vague and does cover some existing ground.
“Create valuable, non-commodity content for your audience”
Well …. Yeah ….. Duh!
This is Google’s fancy way of saying don’t spam our index with AI slop.
The old method of creating high quality content for people that is unique and valuable still holds true – who knew? (Psssst, we did, as did many experienced SEOs).
Myth Busing AEO & GEO techniques
Probably my favorite part of this whole documentation is the section titled;
“Mythbusting generative AI search: what you don’t need to do”
I am going to tackle each of these mythbusting points, however I do take issue with one of them and I will explain why.
- LLMs.txt
- Chunking
- Writing just for AI
- Seaking inauthentic mentions
- Overfocusing on structured data
LLMs.txt
The LLMs.txt file, an anthropic proposed standard, is probably one of the most controversial and misrepresented items in the SEO industry. I love standards, and if there was a method to structure and format data that moved the needle, believe me, I would be using it to move the needle, since I am in the needle-moving business.
However, when I looked at this, I was like, yeah, you are breaking down the content in a similar way to how the LLMs break it down to use it. Supplying a text file with a bunch of summarized info and references is unlikely to do much for you. I have been asked by some clients to produce this, however I have advised that it was not worth investing time into, and it turns out Google agrees.
Having a good solid understanding of your site and being able to format it into an LLMs.txt is not a bad thing and if you have the time to spare – have at it!
Also, I put this into the category of can’t hurt but also likely doesn’t help.
So, if you feel so inclined, go for it.
Chunking
Chunking is a method that LLM models use when ingesting data. Basically they strip out anything not significant and needed, a lot of stop words, white space and ohter none useful elements, then they chunk it together in a block that they can easily handle and process.
Formatting your text into chunks likely never did anything, other than probably not read as well, therefore convert worse and get artificially moved down in search.
This is one of the AI guru recommendations that didn’t stand up to a rational test.
If you actively did this, probably time to rethink your strategy.
Writing just for AI
This is just the modern spin on write for humans not robots or search engines. AI being the new shiny version of robots.
Seriously though – if you are not writing for humans – what are you doing?
SERIOUSLY??
I do believe we will get to a point where there will be an agentic web, possibly a parallel web, and it could quite possibly grow much bigger and faster than the actual web is today.
But people are who we are optimizing for – so keep writing for people!
Now there was some very bad advice about how to rewrite your content or create a parallel version of your content for AI only.
Please don’t do that.
Seaking inauthentic mentions
This is a thinly veiled message to not create listicle content that puts you at the top of the list.
This did work – but seriously who thought this would work long term?
This has already been decimated and brings us back to the old school early days of SEO where it was a cat and mouse game of trying to trick search engines (I am talking early 2000s here).
If you are trying to trick Google, they will find you, and they will deindex you.
Overfocusing on structured data
OK, I have a little bit of an issue with this one.
I guess I get it don’t over focus but ….. Google, you just said use good foundational SEO best practices and get your site in good technical shape.
Well, what good technical set-up and foundational SEO best practices don’t include schema?
Schema is cheap, largely programmatic and structured.
Whether it is used by LLMs (big topic – more on that in another post) is not the point here.
So I do take issue with this a little but.
Don’t overdo it, but Schema to me very much falls into the category of it can’t hurt and it can only help.
Conclusion
Google has produced a pretty conclusive and decisive document on what is and more importantly, what isn’t important for generative AI based search.
They even created a myth-busting section that should silence many of the AEO & GEO gurus that were touting techniques and strategies that were, as politely as I can be, not quite right.
The summary really is as it has always been;
- Write for people.
- Provide something of unique value and perspective (hopefully that is what I have done here? Let me know)
- Follow sound best practice SEO techniques.
- Don’t try and game the system.
References:
- Optimizing your website for generative AI features on Google Search – Google – May 2026
