Google publishes a guide to preparing for productive AI search

Google published an official guide on May 15, 2026, which specifically addresses how to configure the productive AI features in Google Search.

This article, titled “Optimizing your website for generative AI features in Google Search,” was published by John Mueller via the Google Search Central Blog and is now housed under the new “Generative AI fundamentals” navigation section in the Search Central articles.

This is Google’s clearest, on-the-record statement yet about what they say works – and what doesn’t – with features like AI Overviews and AI Mode.

Including the guide

Many of the guidelines include positions that Google has shared at conferences, blog posts, and interviews over the past year. It is organized into five main categories:

Is SEO still important for productive AI searches?
Use basic SEO best practices for productive AI searches
Fictional AI search: what you don’t need to do
Check out the agent’s experience
Next steps: what to focus on

What’s new is that it’s now text, which serves as a reference point for advertisers who have been asking for clarification.

Google’s key message: AI search visibility is still SEO

The guide’s position is clear: SEO is still important for productive AI searches. Google clearly states that AI Overviews and AI Mode do not work on completely different systems.

“SEO best practices continue to work because our generative AI features in Google Search are based on our core Search terms and quality systems,” according to the new guide.

Google specifies that its productive AI features use AI techniques such as generation-augmented retrieval and query followers “to highlight content in our Search index.”

In other words, if your content isn’t technically sound and isn’t high enough quality to rank in traditional search, it won’t work for AI-generated results either.

Myth section: What Google says to stop doing

Google now says that SEOs can “ignore” the following tactics of Google Search and its artificial intelligence features:

llms.txt files: The Google search engine may find these files, but they are treated like any other text file. There is no specific treatment or targeting method that is preferred. (Note that other browsers may use such files.)
Reducing content: There’s no need to break content into smaller pieces for AI programs, Google says. Google claims that its systems can understand pages with multiple topics and extract the relevant passage without the author having to pre-categorize the article.
AI-specific rewriting: AI features can understand synonyms and common definitions, according to Google. Rewriting content to capture all long-tail keywords isn’t necessary, the guide says.
Special schema or Markdown versions of pages: It’s not necessary for Google’s AI search installation, says the new guide.

Google also realized that looking for fake “spokes” to influence what’s said about your products and services is probably not going to be helpful because its productive AI features rely on the same systems and safeguards as the main ranking systems.

“Our basic programs focus on high-quality content while other programs block spam; our productive AI features rely on both,” according to the guide.

An agent’s experience stage: A signal for what’s next

Google refers to agent-friendly best practices and emerging standards such as Universal Commerce Protocol (UCP) and WebMCP, which allow AI agents to take action on users directly from search results.

The section is framed as optional and forward-looking.

“If this is something related to your business and you have some extra time, check out the available agent information and review the best practices guide for an agent-friendly website…” the guide says.

Read more: Agent Search: How AI agents will determine what species are available

Why this is important for marketers

It is important to note that this guide only applies to the Google ecosystem. ChatGPT, Claude, and other AI engines can play by different rules. And Google may not reveal all the details about what works and what doesn’t.

However, this GEO guide is important for marketers for two reasons:

First, the infrastructure around AI search is getting stronger, and when the infrastructure gets stronger, accountability follows. Search marketers are increasingly professionals who are well positioned to lead this career.

Second, the guide legitimizes discipline while limiting what that discipline actually entails. Google does AEO and GEO as extensions of SEO – not separate channels that require separate expertise.

How Semrush supports AI search visibility

Google Guide isn’t introducing a new playbook. But it confirms one that works already. The question is whether your current practice is truly sustainable. Semrush gives you data to find.

Use these tools:

Keyword Tool Magic: Find relevant keywords and topics that create a basis for visibility within AI-generated answers.

The Keyword Magic Tool shows a list of keyword ideas and metrics like intent, volume, and difficulty.

AI visualization toolkit: See which of your URLs are being cited in AI responses, which commands are causing those citations, and how citation frequency is trending over time across platforms.

Visibility Review Report showing overall score and metrics such as citations, citations, pages cited, and distribution by LLM.

Developing enterprise AI: For business teams — track AI mentions and quotes, analyze sentiment, benchmark against competitors, find AI content gaps, and track foot changes throughout the AI system change.