Google Just Said AEO and GEO Are Still SEO. What Replaces Them?

On May 15, 2026 Google Search Central, with John Mueller, Gary Illyes, and Cherry Prommawin attached to the messaging at Search Central Live, published its first official guide titled "Optimizing your website for generative AI features on Google Search." The headline finding is that AEO and GEO are not separate practices from SEO. They are SEO under different names.

The guide also explicitly names tactics that site owners can ignore. That list includes llms.txt files, content chunking for AI, AI-specific rewriting, special AI schema markup, and so-called inauthentic mentions. Several vendors sell retainers built on those tactics. This post is what to do when your CEO asks whether the AEO budget still makes sense.

What Did Google's May 15 AI Search Guide Actually Say?

The guide states that from Google's perspective, optimizing for generative AI search is optimizing for the search experience, and is therefore still SEO. AI Overviews and AI Mode use the same Google index as classic results, grounded through retrieval-augmented generation against indexed pages, not against the LLM's training data. The fundamentals win.

The list of confirmed-irrelevant tactics is the more interesting part. Llms.txt does nothing for Google. Content chunking does nothing. Special schema for AI does nothing. Inauthentic mentions do nothing. That is direct guidance from the Search Central team, not interpretation.

Is AEO Actually Different From SEO According to Google?

No. Google's position, confirmed by Gary Illyes and Cherry Prommawin at Search Central Live, is that AEO and GEO do not require separate frameworks. AI Mode and AI Overviews use the same crawl, the same index, and the same ranking signals that classic Search uses, with one additional layer of generative summarization on top of the retrieved pages.

That conclusion does not mean AI search is identical to classic search. The citation surface and click-through behaviour are very different. But the optimization surface, what you change on your site, is the same surface SEO has always operated on.

Should You Remove Your Llms.txt File Right Now?

If you added an llms.txt file in 2024 or 2025 expecting AI engines to use it, you can remove it without losing anything. Google does not read it. OpenAI does not read it. Anthropic does not read it as a ranking signal. Llms.txt was a community proposal that no major engine implemented as a discovery or ranking input.

The honest counter-argument is that llms.txt does not harm SEO either. If yours is already in place and the maintenance cost is zero, removing it changes nothing observable. But do not build new ones for new sites in 2026.

What Is "Non-Commodity Content" and Why Does It Matter?

Non-commodity content is Google's phrase for content that adds something a model cannot generate by sampling its training data. First-party data, named expert opinions, original frameworks, primary research, customer case studies, and contemporaneous timelines are non-commodity. Generic definitional content is commodity. AI Mode does not need to cite commodity content because the model can already produce it.

For B2B SaaS blogs the practical filter is whether your post contains a fact, opinion, or framework that did not exist before you published it. If yes, you are non-commodity. If no, you are competing with the model's own output and you will lose.

Does Content Chunking Help Google AI Overviews or Not?

According to the May 15 guide, no. Google's indexing pipeline does the chunking itself based on the document structure it crawls. Manual chunking, the practice of breaking long content into artificially small sections to make each "AI-grokable," does not produce better citations. What works is clear semantic structure, scannable headings, and one core idea per paragraph.

The Answer Block pattern I covered in my guide to getting cited by ChatGPT, Perplexity, and Google AI still holds, because it is good SEO, not because it is special AI optimization.

What Is the Actual Difference Between AI Mode and AI Overviews?

AI Overviews is the generative summary that appears at the top of a Google SERP for certain queries. AI Mode is a separate tab inside Google Search dedicated to conversational queries that surfaces answers and follow-ups in a chat-style interface. Both run on Gemini, both ground in the Google index, and both cite source URLs.

The difference matters for measurement, not optimization. AI Overviews appear inside the regular SERP click flow. AI Mode lives in its own tab. According to a March 2026 Ahrefs analysis covering 4 million AI Overview URLs, only 38 percent of AIO citations now come from the top 10 organic results, down from 76 percent in July 2025. I covered the implications in my B2B SaaS AI Mode playbook.

How Does Retrieval-Augmented Generation Actually Ground Citations?

Google's RAG layer retrieves the top-ranked pages for the query, breaks them into passages, ranks the passages by relevance to the query and likely citation value, and feeds those passages to Gemini as grounded context. The model then composes the answer with citations attributed to the passages it used. The citation choice is RAG-driven, not Gemini-driven.

That mechanism explains why classic SEO still matters. If you rank, your page becomes a candidate passage. If you do not rank, you are invisible regardless of how well-written your content is. RAG citations follow ranking. Ranking follows fundamentals.

Should B2B SaaS Sites Add Special Schema for AI Search?

No. The May 15 guide states that AI-specific schema is not used by Google. Standard schema.org types like Article, FAQPage, Product, Organization, and BreadcrumbList continue to help, because they help classic SEO. There is no SoftwareApplication-specific AI schema, no "AI-friendly" markup variant, and no LLM-specific JSON-LD type.

The May 7, 2026 deprecation of FAQ rich results in the SERP also matters here. FAQ schema still helps Google understand structure, but does not produce the visible FAQ accordion in results any more. Do not strip it. Just stop expecting the visual.

How Does the May 7 FAQ Rich Result Deprecation Fit In?

Google retired the FAQ rich result rendering on May 7, 2026, two weeks before the AI Search guide. The FAQ snippet that used to appear in the SERP as an expandable accordion is gone for new pages and being phased out for existing pages. The underlying FAQ schema is not deprecated. The rendering is.

For Webflow sites with FAQ schema implemented via the Open Graph and Schema fields, no action is required. The schema continues to feed Google's understanding of the page. The visible accordion was a layer on top, and that layer is gone.

What Should You Tell Your CEO If They Ask About the AEO Budget?

Tell them AEO is a labelling problem, not a budget problem. The work that vendors call AEO (question-based H2s, answer blocks, named entities, citable stats, primary research) is also good SEO. The work that vendors call AEO but Google says is irrelevant (llms.txt, AI-specific schema, manual chunking) can be cut from the retainer.

The honest reframe is to ask the vendor to show you which deliverables changed because of their AEO offering versus their classic SEO offering. If the answer is "the same deliverables under a different name," you have your answer about the budget.

If you want a Phoenix Studio audit of your current SEO and AEO retainer to identify which deliverables Google's May 15 guidance just made redundant, drop me a line. Let's chat.