Google's AI optimization guide says there is no separate playbook for AI search. Showing up in Google's AI answers is still Search Engine Optimization (SEO). The guide went live on May 15, 2026. It tells site owners to write original content, keep pages open to Google's crawler, and skip five popular tricks. Two of those tricks are llms.txt files (a text file meant to brief AI tools about your site) and "chunking" (cutting pages into small pieces for AI).
We read the whole page so you don't have to. Below, each part of the guide gets a plain summary, a short quote, and what it means for a business website. Our angle is practical, not a rebuttal. The guide describes how Google's own systems work, and knowing that helps you decide what to skip.
Key facts
- The full title is "Optimizing your website for generative AI features on Google Search" (official page).
- It was announced on May 15, 2026 in a short blog post. The last update was on July 10, 2026.
- The guide covers the AI answers inside Google Search only. It says nothing about ChatGPT or other AI tools.
- Its core message is that AI search work is, in the guide's words, "still SEO".
Why it matters: at the May 2026 developer keynote, Sundar Pichai said AI Overviews "now has over 2.5 billion monthly active users" (keynote text).
What is Google's AI optimization guide?
The guide is the official advice for site owners who want their pages to show up in AI Overviews and AI Mode. It sits in the same documentation set as the SEO starter guide, so it carries the same weight as the rest of the official search docs.
John Mueller wrote the announcement. It lists what the guide covers: unique content, tips for local, shopping, image and video results, myths about AI search tactics, early notes on AI agents, and why SEO still applies. He ended with: "We hope this guide provides more clarity to help you continue to create great content that reaches your audience effectively".
The guide also explains two terms you will hear in every AI search pitch:
- Retrieval-augmented generation, also called grounding. The AI pulls current pages from the normal search index, writes its answer from them, and links to them.
- Query fan-out. The AI runs several related searches at once. The guide's example: a search for "how to fix a lawn that's full of weeds" may also run "best herbicides for lawns" and "remove weeds without chemicals".
Put simply, the AI picks its sources from the same index and the same ranking systems as the regular results. That one fact explains most of the guide.
What does it say to do?
The guide asks for three things: original content, a sound technical setup, and accurate local or product details. Content comes first. In the guide's words, unique, useful content "will likely influence your website's presence in generative AI search in the long run more than any of the other suggestions in this guide".
Content: write what only you can write
The guide splits content into commodity and non-commodity. Its commodity example is a post called "7 tips for first-time homebuyers", which anyone could write from common knowledge. Its non-commodity example is a first-hand story about waiving a home inspection and checking the sewer line instead. Only someone who did it could write that one.
The other content points:
- Organize pages with paragraphs, sections and clear headings for people.
- Add good images and video, since AI answers can show them too.
- Don't build a separate page for every fan-out query. Doing that "primarily to manipulate rankings or generative AI responses" breaks the scaled content abuse policy.
- If you use AI writing tools, the result still has to follow the spam policies.
Picture a hypothetical Toronto accounting firm. A post called "5 tax tips for small businesses" is commodity. A post on mistakes the firm itself caught in client sales-tax filings is not, because no other firm has those files.
Technical: be indexable first
To qualify, "a page must be indexed and eligible to be shown in Google Search with a snippet". After that come the usual technical checks: crawlable pages, JavaScript that renders, a good experience on every device, and less duplicate content.
Two details matter. First, the guide says tidy HTML helps but "don't worry about perfect code". Second, there is a new switch. A site "must be included in Search generative AI features in Search Console" to appear. The help page for that switch says it lives under Settings and includes your site by default. It reached every website on August 31, 2026.
Local and ecommerce: keep your listings current
AI answers can include business details and product listings. The guide points owners to their business profile and to product feeds as the way to supply them. For a local service firm, the profile's hours, services and photos are the first thing to check.
Which AI-search myths does it reject?
The guide names five tactics you can ignore. It says many of these "hacks" "aren't effective or supported by how Google Search actually works".
| Common AEO/GEO advice | What Google's guide says | What we'd do on a business site |
|---|---|---|
| Add an llms.txt file so AI can read your site | "Google Search itself doesn't use them". Creating one "will neither harm nor help" your Google visibility. | Optional. Only add one if another service you care about reads it. |
| Break content into small chunks for AI | "There's no requirement to break your content into tiny pieces". Google can find the relevant part of a long page. | Write for the reader. Clear headings help people, which also helps machines. |
| Rewrite pages in a special "AI style" | "You don't need to write in a specific way just for generative AI search". AI understands synonyms. | Stop padding pages with long-tail keyword variants. |
| Buy or seed mentions across the web | Seeking inauthentic mentions "isn't as helpful as it might seem". Spam systems filter it. | Earn real mentions: reviews, partner pages, press, client case studies. |
| Pile on structured data for AI | "Structured data isn't required for generative AI search, and there's no special schema.org markup you need to add". | Keep schema for rich results and accuracy. Don't expect it to unlock AI answers by itself. |
Two of these need context.
llms.txt. This is a proposed text file that summarizes a site for AI tools. The guide says search ignores it. It also says keeping one for "other services or systems that use these files" is "completely fine". So the guide tells you how one search engine behaves, and nothing about other tools.
Structured data. It isn't required for AI answers, but the guide adds that "it's a good idea to continue using it as part of your overall SEO strategy". Our structured data guide covers what schema is actually used for. The organization schema and local business schema posts give copy-paste examples.
Does Google say AEO and GEO are different from SEO?
No. Answer Engine Optimization (AEO) means shaping pages so AI answer tools quote them. Generative Engine Optimization (GEO) is a near-synonym. The guide treats both as the same work as SEO: "optimizing for generative AI search is optimizing for the search experience, and thus still SEO".
The guide also asks anyone weighing AEO or GEO services to read the advice on judging outside SEO help. Later it adds a warning: "No third-party tool has access to our internal ranking or AI systems".
We agree with the substance. What we add is execution. A page can rank and still bury its answer too deep to be quoted.
Our post on how to get cited in an AI Overview covers the habits that make an answer easy to lift. Put the answer first, keep one question per section, and make the schema match the visible text. None of that contradicts the guide. Call it SEO done carefully.
The extra terms still help outside this one engine. ChatGPT and Perplexity run their own systems, and the guide makes no claims about them. Our SEO vs AEO vs GEO post explains where the three overlap and where they split.
What should a business website change?
Most sites need fewer changes than the sales pitches suggest. The guide's own recap says "plenty of content thrives in Google Search (including generative AI experiences) without any overt SEO at all". Here is the short list we would work through this quarter, in order.
| Priority | Task | Where the guide supports it |
|---|---|---|
| 1 | Confirm key pages are indexed and show a snippet. Check that no nosnippet tag or robots rule blocks them. | Technical requirements section |
| 2 | Open Search Console, go to Settings > Search generative AI, and confirm the site is included. | Technical section, Search Console help page |
| 3 | Open the Generative AI performance report and note which pages earn impressions. | Measurement section |
| 4 | Rewrite your two or three most important commodity pages with first-hand detail: real numbers, real projects, real process. | Non-commodity content section |
| 5 | Update your Google Business Profile hours, services and photos. Ecommerce sites: check Merchant Center feeds. | Local and ecommerce section |
| 6 | Merge thin near-duplicate pages written for keyword variants. | Fan-out and duplicate content guidance |
| 7 | Keep structured data accurate. Don't add more for AI's sake. | Myths section |
On step 3, the report's help page says it counts how often your links appeared in AI answers. You can split it by page, country, date and device. Impressions are its only number, and small sites may not see the report at all. It reached every site on August 31, 2026.
On step 1, remember that robots.txt controls crawling, not indexing. If you've added rules for AI crawlers, read our robots.txt in 2026 post before you change one.
What you can safely stop doing: paying for llms.txt setups aimed at this engine, splitting long pages into fragments, and publishing dozens of near-identical "best X in Y" pages.
Agents: low priority for most sites
The guide's last section covers AI agents, meaning software that acts for a person, such as booking a table or comparing products. It frames this as optional: "If this is something that's relevant to your business and you have extra time". It links to a web.dev page on agent-friendly sites and a new commerce protocol for shopping agents. If your site takes bookings or sells online, plain forms and clear button labels are a sensible start. For a brochure site, this one can sit at the bottom of the list.
What does the guide not cover?
The guide covers one search engine and nothing else. Several questions owners ask us fall outside it:
- Other AI engines. The myth list is framed as things "you can ignore for Google Search". It says nothing about how ChatGPT or Perplexity choose sources.
- AI training. The opt-out switch only controls whether content appears in AI answers. Its help page says: "This control doesn't affect AI training". Training opt-outs go through a separate robots.txt token. Two more gaps matter for planning:
- How sources are picked. The guide says AI answers rely on the core ranking systems. It does not explain why one ranking page gets quoted and another doesn't.
- Clicks and leads. The new report counts impressions only. Clicks and leads still have to come from your normal reports and analytics.
- Formulas. There are no word counts or schema checklists. On length, it says plainly: "There's no ideal page length".
The missing formula is useful to buyers. When a proposal promises AI search results through llms.txt, chunking or "AI-optimized" rewrites, you can point to the official page.
If you want to know where your site stands on indexing, the opt-in switch and content, our free site check reviews them.