Google’s latest Search Off the Record podcast discussed whether ‘SEO is on a dying path’ because of AI Search. Their assessment sought to explain that SEO remains unchanged by the introduction of AI Search, revealing a divide between their ‘nothing has changed’ outlook for SEO and the actual experiences of digital marketers and publishers.
Google Speculates If AI Is On A Dying Path
At a certain point in the podcast they started talking about AI after John Mueller introduced the topic of the impact of AI on SEO.
John asked:
“So do you think AI will replace SEO? Is SEO on a dying path?”
Gary Illyes expressed skepticism, asserting that SEOs have been predicting the decline of SEO for decades.
Gary expressed optimism that SEO is not dead, observing:
“I mean, SEO has been dying since 2001, so I’m not scared for it. Like, I’m not. Yeah. No. I’m pretty sure that, in 2025,the first article that comes out is going to be about how SEO is dying again.”
He’s right. Google began putting the screws to the popular SEO tactics of the day around 2004, gaining momentum in 2005 with things like statistical analysis.
It was a shock to SEOs when reciprocal links stopped working. Some refused to believe Google could suppress those tactics, speculating instead about a ‘Sandbox’ that arbitrarily kept sites from ranking. The point is, speculation has always been the fallback for SEOs who can’t explain what’s happening, fueling the decades-long fear that SEO is dying.
What the Googlers avoided discussing are the thousands of large and small publishers that have been wiped out over the last year.
More on that below.
RAG Is How SEOs Can Approach SEO For AI Search
Google’s Lizzi Sassman then asked how SEO is relevant in 2025 and after some off-topic banter John Mueller raised the topic of RAG, Retrieval Augmented Generation. RAG is a technique that helps answers generated by a large language model (LLM) up to date and grounded in facts. The system retrieves information from an external source like a search index and/or a knowledge graph and the large language model subsequently generates the answer, retrieval augmented generation. The Chatbot interface then provides the answer in natural language.
When Gary Illyes confessed he didn’t know how to explain it, Googler Martin Splitt stepped in with an analogy of documents (representing the search index or knowledge base), search and retrieval of information from those documents, and an output of the information from “out of the bag”).
Martin offered this simplified analogy of RAG:
“Probably nowadays it’s much better and you can just show that, like here, you upload these five documents, and then based on those five documents, you get something out of the bag.”
Lizzi Sassman commented:
“Ah, okay. So this question is about how the thing knows its information and where it goes and gets the information.”
John Mueller picked up this thread of the discussion and started weaving a bigger concept of how RAG is what ties SEO practices to AI Search Engines, saying that there is still a crawling, indexing and ranking part to an AI search engine. He’s right, even an AI search engine like Perplexity AI uses an updated version of Google’s old PageRank algorithm.
Mueller explained: