Decoding Google's Official AI Optimization Guide — What Actually Works for AI Overviews / AI Mode, and Which GEO/AEO Tactics Google Just Killed (May 2026)

On May 15, 2026, Google published a new Search Central document titled "Google's Guide to Optimizing for Generative AI Features on Google Search" (official guide), accompanied by the announcement blog A new resource for optimizing for generative AI in Google Search.

This column reads that guide and its related Search Central docs as primary sources, and lays out what Google now says you should do — and, more importantly, what Google says you should not bother doing — for AI Overviews and AI Mode. A surprising amount of the GEO/AEO/LLMO advice circulating in 2025–2026 is now explicitly disavowed by the official guide.

The summary panel that appears above traditional SERP results. It acts as a "starting point" for complex topics and shows up when Google decides the AI summary adds value on top of the normal results. It uses retrieval-augmented generation (RAG / grounding) backed by Google Search's core ranking systems.

Since John Mueller's May 2025 post, Google has consistently framed AI Overviews–driven clicks as "higher quality" with longer session times.

Announced at Google I/O 2025 and rolling out from the US, AI Mode is a standalone, chat-style search experience that replaces the entire SERP. A single query is decomposed into sub-queries (definitions, comparisons, counter-examples, expert takes) and fanned out in parallel (Query fan-out) before Gemini synthesizes the answer. The expanded Deep Search runs hundreds of searches and produces a cited, expert-grade report.

Refs: AI features and your website / blog.google AI Mode

The list reads as a refresh of mainstream SEO best practice — almost on purpose.

1. Make uncommoditized, original content — Google explicitly writes: "this is more likely than any other recommendation to influence your visibility in generative AI search" 2. Clear technical structure — indexable, crawlable, semantic HTML, JS-SEO best practice 3. Page experience — fast, comfortable across devices, clear separation of main and supporting content 4. High-quality images and video as supporting media 5. Cut duplicate content 6. Get local / e-commerce data right via Google Business Profile and Merchant Center (AI Overviews surfaces product and local results) 7. Helpful, reliable, people-first content principles 8. Crawl optimization

This is the most useful part of the new guide.

On llms.txt specifically, Mueller stated on Bluesky in June 2025 that "no AI system currently uses llms.txt", and Google has since removed an llms.txt file that had been accidentally placed in its own developer site (Search Engine Roundtable).

Translation: the whole concept of "adding new files or AI-specific schema *for AI*" is officially off the table.

Read the wording carefully:

- Structured data is not required for generative AI search - There is no special schema.org markup for AI search - But keep using structured data as part of your normal SEO (rich-results eligibility, etc.) - Accurate Local Business / Product / Merchant Listing data helps AI answers surface local and product results

The new guide barely touches this; the authoritative source is still AI features and your website.

- `nosnippet` / `data-nosnippet` / `max-snippet` — restrict what AI surfaces show - `noindex` — block indexing entirely - `Google-Extended` — restrict Gemini-app and similar non-Search AI uses

The trap: blocking Google-Extended does not exclude you from AI Overviews or AI Mode, because those features ride on Google's regular ranking systems. To opt out of those, you need the `nosnippet` family. Reflection requires URL Inspection and a recrawl, which can take days to months.

As of May 2026:

- AI Overviews and AI Mode clicks/impressions are consolidated into the Performance report (Web search type), alongside organic and Featured Snippets — no separate breakdown - If a URL appears in both AI Overviews and the Blue Links, that counts as one impression - Feb 18, 2026 — Search Console released AI-powered configuration (natural-language report filtering) - Apr 3, 2026 — Google's Changelog disclosed that impressions were over-counted from May 13, 2025 through April 2026 (~11 months). Clicks were not affected. CTR appears artificially low for that period

Google's line is "clicks are higher quality." Volume is a separate question. These are independent (non-Google) figures:

Methodologies differ, so don't simply compare 58% vs 34.5%. But both agree that the top spot loses a measurable share of clicks on AI-Overview queries.

The guide explicitly mentions AEO and GEO, then stakes out:

> "From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience — and that is still SEO."

So AEO and GEO are not treated as separate disciplines. LLMO doesn't appear in any official Google doc — it's industry slang only.

Do: - Double down on regular SEO (E-E-A-T, original primary research, technical quality, Core Web Vitals) - Keep structured data — for *normal* SEO - Maintain accurate Google Business Profile / Merchant Center data - Use Search Console's AI-powered configuration - Discount CTR data from May 2025 – April 2026 because of the impression-counting bug - Treat AI Overviews exposure as a downstream effect of ranking — invest in normal SERP position first

Don't (officially disavowed): - Publish llms.txt - Ship AI-only Markdown versions of pages - Chunk content "for AI" - Rewrite in an "AI-friendly" voice - Add AI-only structured data - Spin up a page per query variant - Manufacture mentions

Our SEO column index has consistently covered core updates, E-E-A-T, and structured data. With this new guide, we've consolidated our advisory stance into two points:

1. We don't sell "AI-specific optimization" packages. Vendors now do — built on tactics Google just officially disavowed 2. AI Overviews visibility is downstream of original primary content × disciplined SEO, which is exactly what we already deliver. We also help via AI consulting when AI search strategy intersects with broader AI adoption

Q1. Do I really not need llms.txt? A. Right. Google explicitly says no, and Mueller has stated no AI system uses it. It won't hurt, but it doesn't help either. Q2. Should I add schema specifically to appear in AI Overviews? A. No AI-specific schema exists. Keep your normal SEO structured data as-is. Q3. Will blocking Google-Extended remove me from AI Overviews? A. No — those features use the regular search ranking pipeline. Use the `nosnippet` family instead. Q4. Can I see AI Overviews clicks separately in Search Console? A. Not as of May 2026 — counts are consolidated with normal organic. Q5. Is "write short, bullet-heavy content for AI" good advice? A. Google explicitly says no special writing style is needed. Write for readers. Q6. Should I trust the 58% CTR drop number? A. It's a third-party Ahrefs figure, not Google's. Google says click quality has improved. Both dimensions matter. Q7. Should I hire a GEO/AEO specialist? A. Google has effectively said GEO/AEO is not a separate discipline. A strong, evidence-based SEO partner is enough. Be cautious of vendors selling AI-specific tactics Google just disavowed.

Google's AI Optimization Guide (May 15, 2026) is almost defiantly simple: don't add new things for AI. Keep making original, high-quality content under normal SEO best practices. A large fraction of the GEO/AEO advice the industry has been peddling — llms.txt, AI-only schema, content chunking, AI-voice rewrites — was just officially disavowed.

The hidden message is that sites already strong on E-E-A-T, original primary research, and technical quality keep winning into the AI-search era. But the top-of-funnel #1 CTR is measurably eroded on AI-Overview queries, so the strategic move in the back half of 2026 is to rebalance from "click capture" toward branded search, recall, and loyalty.

Primary (Google): - Google Search Central — Google's Guide to Optimizing for Generative AI Features on Google Search - Search Central Blog — A new resource for optimizing for generative AI in Google Search (May 15, 2026) - Search Central — AI features and your website - Search Central Blog — Mueller, May 2025 - Search Central Blog — AI-powered configuration in Search Console (Dec 2025) - blog.google — AI Mode in Search Third-party: - Search Engine Land — Google's guide on optimizing for generative AI features - Search Engine Journal — Google's official advice on optimizing for AI Overviews / AI Mode - Search Engine Land — Query fan-out explained - Ahrefs — AI Overviews reduce clicks (Feb 2026 update) - Semrush — AI Overviews study (10M keywords) - Search Engine Roundtable — Google does not endorse llms.txt - Stan Ventures — Google dismisses llms.txt - emarketed — GSC impression bug 2026 Note: the 58% (Ahrefs) and 34.5% (Semrush) CTR-drop figures are third-party measurements with different methodologies — not directly comparable, and not official Google numbers.