18, May 2026
7-minute read
On May 15, 2026, Google published something the SEO community has been waiting for: an official, consolidated guide on how to optimize websites for its generative AI search features, including AI Overviews and AI Mode. The document, titled Optimizing your website for generative AI features on Google Search, was announced by John Mueller of Google’s Search Relations team via the Google Search Central Blog.
The guide doesn’t introduce a radical new playbook. In fact, its most significant contribution is what it dismantles: years of speculative advice, proprietary acronyms, and vendor-invented tactics that Google now explicitly says are unnecessary.
Here is a thorough breakdown of everything in the guide, and what it means for your website strategy going forward.
Why Google published this now
Google’s AI features have been rolling out at pace, AI Overviews became widely available in 2024, and AI Mode (a fully AI-generated search experience) entered broader availability in 2025 and 2026. As these features matured, a cottage industry of “AEO” (Answer Engine Optimization) and “GEO” (Generative Engine Optimization) consultants emerged, each selling proprietary methods to rank in AI results.
Google’s guide is, in part, a direct response to this noise. The company has been quietly pushing back on AEO/GEO terminology at conferences for months. This document makes the official position crystal clear: there is no separate discipline for AI search optimization. The foundations that have always driven good SEO drive visibility in AI features too.
Key context: Google’s generative AI features are built on top of its existing Search index and ranking systems, not a separate index. Content that performs well in traditional search is the content that gets surfaced in AI responses.
How Google’s AI features actually work
To understand why the guide’s recommendations, make sense, it helps to understand the two core technical mechanisms Google describes:
1. Retrieval-Augmented Generation (RAG)
RAG, also known as grounding, is the process by which Google’s AI models pull content from its Search index to generate responses. Rather than relying on what the model “knows” from training, it retrieves fresh, relevant content from indexed web pages and uses that to form its answer. This is why crawlability and indexability remain foundational, if Google can’t access your pages, your content can’t be grounded in AI responses.
2. Query fan-out
When a user submits a query, Google’s AI doesn’t just process that single question. It generates a set of concurrent, related sub-queries to gather a broader range of information before synthesising a response. This means your content doesn’t need to match a query word-for-word, it needs to be genuinely useful and comprehensive across a topic area.
The content strategy that wins
The guide also emphasises organising content in a way that helps readers navigate clear structure, logical flow, and headings that reflect what users are actually looking for. This isn’t new advice, but the context matters: AI systems that extract and summarise content respond better to well-structured pages.
Crucially, Google warns against overdoing it. Padding articles with redundant information to hit a word count or forcing keywords into content where they don’t naturally belong, works against you. The benchmark is whether a real person would find the content genuinely useful and worth reading.
Technical SEO is still non-negotiable
Given that RAG depends on Google being able to retrieve your content, the technical requirements haven’t changed, they’ve become more important:
Pages must be crawlable and indexable. Robots.txt exclusions, no index tags, or broken canonical structures will keep your content out of AI responses entirely.
Semantic HTML helps AI systems understand content structure and hierarchy. Use heading tags correctly and meaningfully.
JavaScript SEO best practices apply. If your content is rendered via JS and Googlebot can’t execute it, that content doesn’t exist from Google’s perspective.
Page experience signals — Core Web Vitals, mobile-friendliness, and HTTPS remain part of the ranking foundation AI features are built on.
Minimise duplicate content. Thin or duplicated pages dilute your site’s overall quality signals.
Local & e-commerce signals
For businesses operating in local search or e-commerce, the guide provides specific direction:
For local businesses: Maintain an up-to-date Google Business Profile. AI Overviews and AI Mode increasingly surface local results, and the Business Profile is how Google understands your location, hours, services, and reputation.
For retailers: Use Google Merchant Center feeds. Product data submitted via Merchant Center feeds into shopping-related AI responses and surfaces your inventory in ways that organic pages alone cannot replicate.
Both of these are existing best practices, the guide simply reinforces that they carry direct weight in AI-generated responses, not just traditional search results.
Myths Google just officially busted
Perhaps the most immediately actionable section of the guide is where Google explicitly names tactics that do not help with AI search optimisation. If you or your agency have been investing time in any of the following, you can stop:
✕ Creating llms.txt filesGoogle does not use llms.txt files for its search or AI features. Adding one provides no benefit for AI Overviews or AI Mode.
✕ Content chunkingRestructuring content into small, discrete chunks in an attempt to make it more digestible for AI is unnecessary. Write for human readers.
✕ AI-specific rewritesRewriting content specifically to “sound better to AI” is not a recognised optimisation signal. Good human-readable content is what’s needed.
✕ Special schema markup for AIThere is no schema.org markup that specifically targets generative AI features. Structured data remains valuable for rich results, but only in that context.
Worth noting: The guide also addresses the AEO/GEO terminology debate directly. Google’s position is unambiguous, these are marketing terms, not distinct disciplines. The company considers all of this to fall under the umbrella of SEO done well.
AI agents: the emerging frontier
One section of the guide ventures into less settled territory: AI agents. Google acknowledges that AI agents, systems that can take actions on behalf of users, not just answer questions, are an emerging and fast-evolving space within Search.
The guide provides only initial guidance here, acknowledging that norms and best practices are still forming. What is clear is that Google is thinking about how agent-based queries will surface content differently from conversational AI queries, and that website owners should expect further guidance as this space develops.
For now, the same foundational principles apply: well-structured, authoritative, crawlable content is what agents will rely on to take action and retrieve information.
The bottom line
Google’s AI optimisation guide is less a revelation and more a consolidation, bringing together advice that has been scattered across blog posts, conference talks, and documentation into one clear reference document. Its value is in the clarity it provides, not in introducing new tactics.
The single most important takeaway: the websites that will perform best in AI search are the websites that would have performed best in traditional search ones with unique, trustworthy, well-structured content built on a solid technical foundation.
The arms race of AI-specific optimisation tactics was always a distraction. Google has now said so officially. The path forward is straightforward, if not always easy: build something genuinely worth citing.
“The websites that perform best in AI search are the ones that would have performed best in traditional search always.”
Source: Google Search Central