For the past two years, the SEO industry has spent a great deal of time trying to determine what, if anything, fundamentally changed when Google introduced generative AI into Search.

Do you need to create an llms.txt file? Should you break your content into tiny chunks? Is there a special way to write for AI systems? Does traditional SEO still matter or is SEO dead as so many articles proclaimed? Or, do we now need to think in terms of Answer Engine Optimization (AEO) or Generative Engine Optimization (GEO)?

Google’s newly published guidance on optimizing your website for generative AI features in Google Search provides the clearest answers we’ve received directly from Google to date.

Welp, according to Google, it looks like not much has changed.

The guidance from Google says if you want to perform well in AI Overviews and AI Mode, continue doing the things that have always mattered in search. That means to create genuinely useful content, have a technically sound website, and focus on helping your audience rather than chasing shortcuts.

Mostof that sounds quite a lot like EEAT, right?

What stood out to me most here was Google’s inclusion of query fan-out as a core part of how these systems retrieve information. Google has referenced query fan-out before, particularly in its announcements around AI Mode. What is notable here is that the concept is now included directly in Search Central’s guidance for website owners. That is important because it formally reinforces that Google’s AI systems are not limited to the search results for the exact query entered by the user.

My interest in this topic actually began with Google’s patent for AI Overviews. In my May 2024 article, How Does Google AI Overview Work? Insights From the Patent, I highlighted that the system expands beyond the original query to retrieve documents from related, recent, and implied queries.

That idea was central to a follow-up article I wrote in July 2024, Google AI Overviews: Do Ranking Studies Tell the Whole Story?, where I argued that many studies at the time and SEOs themselves were overlooking how Google expands to related queries when building AI Overviews. Authoritas and I published the first study on query fan-out in AI overviews before it came to be known as “query fan-out.” We called the function “related queries” as that was what it was referred as in the Google AI overview patent.

Google’s latest documentation strongly supports that earlier analysis.

A Note on Google’s Guidance and What It Doesn’t Tell Us

Before getting too deep into Google’s recommendations, it is worth keeping one important point in mind.

These are Google’s guidelines.

That does not mean they are inaccurate or misleading. In fact, I believe much of the guidance is directionally correct and aligns closely with what many SEOs have observed through testing, research, and patent analysis.

At the same time, Google has a vested interest in protecting the integrity of its search systems. If Google were to provide a detailed blueprint of exactly how AI Overviews and AI Mode retrieve, evaluate, and rank content, it would make those systems far easier to manipulate. For that reason, Google’s public guidance is best viewed as a high-level framework rather than a complete technical explanation of how these systems work.

This is not unique to AI search. Google has always shared broad best practices while keeping the finer details of its ranking systems close to the vest. That is why independent research, testing, and patent analysis remain so valuable. Google tells us what it wants website owners to focus on, while studies and real-world observations help us better understand how those principles play out in practice.

In this case, Google’s latest guidance is particularly useful because many of the concepts it discusses, such as Retrieval-Augmented Generation (RAG), query fan-out, and the importance of non-commodity content, align closely with what earlier research and patent analysis suggested.

One interesting aspect of Google’s guidance is what it does not mention. There is no explicit discussion of backlinks, EEAT, internal linking, reviews, or brand mentions in third-party publications, even though many of these signals have long been considered important for building authority and trust. That is a useful reminder that Google’s documentation is intended to provide broad strategic direction rather than a complete list of ranking signals or a comprehensive description of how Google evaluates sources.

SEO Is Still SEO

One of the most important lines in Google’s new guidance is also one of the simplest:

“Optimizing for generative AI search is optimizing for the search experience, and thus still SEO.”

This is an important statement because it positions Google as railing back against the idea being pushed that AI search requires an entirely new discipline. Yes, Google AI is using large language models to synthesize responses. Yes, the interface looks different. But the systems behind these responses still rely heavily on Google Search.

Pages still need to be crawlable and indexable. Content still needs to be useful and trustworthy. Strong technical SEO and a positive page experience still matter. For anyone wondering whether SEO fundamentals remain relevant, Google’s answer is unequivocally yes.

Google also specifically notes that information from Google Business Profiles and Merchant Center can help local businesses and ecommerce sites appear in AI-generated responses.

Query Fan-Out and Why It Matters

Google defines query fan-out as a process where the model generates a set of concurrent, related queries to gather additional information. For example, if someone searches for “how to fix a lawn that’s full of weeds,” Google may also search for related questions such as “best herbicides for lawns,” “remove weeds without chemicals,” and “how to prevent weeds in lawn.”

This process allows Google to pull information from a wider set of search results rather than relying only on pages ranking for the original query. That helps explain one of the most common observations in the early AI Overview research: cited pages often rank poorly, or not at all, for the query being analyzed.

The diagram above is based on my original review of Google’s patent. While simplified, it illustrates a key point that is now reflected in Google’s documentation: the system can use Retrieval-Augmented Generation (RAG) to pull information not only from documents that directly match the user’s query, but also from related, recent, and implied queries before generating and linking to the final summary.

Why Some Cited Pages Don’t Rank for the Query

When early AI Overview studies showed that many cited URLs ranked outside the top ten, some people viewed that as evidence that traditional rankings were becoming less important. I saw it differently due to my deep dive on the Google patent.

In my 2024 article, I suggested that these studies were taking too narrow a view by looking only at rankings for the exact query. If Google was expanding its search to related queries, a cited page might rank highly for one of those related searches even if it did not rank for the original query. That’s what Authoritas and I discovered in our related queries research.

That interpretation aligns closely with what Google is now describing publicly. This does not mean rankings are irrelevant. It means that the relevant ranking set may be broader than the single keyword you might be tracking.

How This Connects to the Patent

Long before Google added query fan-out to its public documentation, the concept was already reflected in the patent underlying AI Overviews.

In my article How to Optimize for AI Overviews: Patent & Research Insights, I examined the retrieval and document selection process described in the patent.

The patent suggested that when the initial search results were not sufficiently diverse or high quality, the system could expand to related queries, recent queries, and implied queries. It also described how documents might be evaluated based on factors such as ranking, relevance, diversity, and trustworthiness. Those ideas later informed my article on related queries and helped explain why AI Overviews frequently cited pages that did not rank for the original query.

Google’s new guidance does not reference the patent directly, but its discussion of Retrieval-Augmented Generation (RAG) and query fan-out aligns closely with the broader framework described there.

Google’s Emphasis on Non-Commodity Content

The section of Google’s guidance that I found most compelling was its discussion of non-commodity content. Google contrasts generic content such as “7 Tips for First-Time Homebuyers” with more specific, experience-based content like “Why We Waived the Inspection & Saved Money: A Look Inside the Sewer Line.” That’s a pretty significant distinction.

Commodity content summarizes information that is widely available. Non-commodity content reflects direct experience, unique observations, and original insights that are harder to replicate. Again, this aligns heavily with Google’s EEAT guidelines.

This is the type of content that provides the most value to readers and gives Google a reason to surface your perspective over countless similar summaries.

What Google Says You Can Ignore

Google also addresses several tactics that have gained traction over the past year. According to Google, you do not need to create special AI files such as llms.txt, artificially chunk your content, rewrite your pages for AI systems, or obsess over every possible keyword variation. That does not mean these tactics are harmful in and of themselves, but Google is clearly signaling that they are not a requirement to succeed in AI search experiences.

Google also reiterates an important point from its spam policies.

“In the context of Google Search, spam refers to techniques used to deceive users or manipulate our Search systems into featuring content prominently, such as attempting to manipulate Search systems into ranking content highly or attempting to manipulate generative AI responses in Google Search.”

Using AI to generate large volumes of pages primarily to manipulate search rankings can violate Google’s scaled content abuse policy. The issue is not the use of AI itself. AI can be incredibly helpful for research, outlining, and drafting content. The problem arises when businesses publish large amounts of generic content that add little original value for users.

In other words, Google is encouraging website owners to focus on creating genuinely useful content rather than chasing technical shortcuts or mass-producing low-value pages.

So, What Does This Mean for SEO and Content Strategy?

If there is one practical takeaway from Google’s guidance, it is that the best long-term strategy remains largely unchanged. Focus on creating content rooted in real expertise and experience. Build pages that address the broader topic and the related questions your audience is likely to have. Ensure your website is technically accessible to search engines. Support your content with strong visuals when appropriate.

In short, focus on helping your audience rather than trying to reverse engineer shortcuts.

Final Thoughts

Google’s new guidance does not reveal a secret formula for AI visibility. What it does provide is a clearer explanation of how Google wants website owners to think about AI search. And, surprise, surprise…SEO still matters. Original, experience-driven content still matters. Technical fundamentals still matter. And Google’s AI systems look beyond the exact query to a broader set of related searches when assembling responses.

For me, the most interesting aspect of Google’s new guidance is not that Google mentioned query fan-out here. It is that Google has now incorporated the concept directly into its official Search Central documentation for website owners. That formalizes something many of us have observed through patent analysis, testing, and research: Google’s AI systems are capable of expanding beyond the original query to gather information from a broader set of related searches.

At the same time, it is important to remember that this is still Google’s public guidance. Google has every incentive to share the broad principles it wants site owners to focus on while keeping the finer details of its systems private. If Google fully disclosed exactly how AI Overviews and AI Mode retrieve, evaluate, and rank content, those systems would be much easier to manipulate. That is why Google’s documentation should be viewed as an important directional guide rather than a complete technical blueprint.

When Google’s guidance aligns with insights from patents, independent research, and real-world testing, it increases our confidence that we are focusing on the right things. Ultimately, Google’s latest guidance reinforces a simple point.

The best way to succeed in AI search is the same as it has always been in traditional search: create genuinely useful content that offers unique value to real people.

References

• Creating helpful, reliable, people-first content by Google
• Generative summaries for search results Google Patent
• Google AI Overviews: Do Ranking Studies Tell the Whole Story? by Rich Sanger
• Google AI Overview Study: Link Selection Based on Related Queries by Rich Sanger and Authoritas
• How Does Google AI Overview Work? Insights From the Patent by Rich Sanger
• How to Optimize for AI Overviews: Patent & Research Insights by Rich Sanger
• Optimizing your website for generative AI features on Google Search by Google