Key Takeaways
- Google’s May 15 guide confirms AI search optimization is still SEO, with GEO and AEO no longer separate disciplines.
- Five tactics no longer help, including llms.txt for ranking, content chunking, machine-friendly rewrites, inauthentic mentions, and schema as an AI hack.
- The three priorities to maintain are people-first content, a clean technical structure, and accurate local or product signals.
- An llms.txt build helps only when agentic browsing matters for your customers, since Google Search does not need one.
On May 15, 2026, Google released its first official guide for optimizing websites in AI search. The document closes two years of speculation about how AI Overviews, AI Mode, and other generative results work. The headline finding lands clearly. Most of what the industry called GEO and AEO is still SEO. A few tactics shifted. Most stayed the same. This post breaks down what changed, what held, and what your team should plan for next.
What Google Released
Google’s AI Optimization Guide explains the two pieces behind AI search. Retrieval-Augmented Generation (RAG) pulls ranked pages from the Search index, then grounds the AI response in those pages. Query fan-out runs several related searches at once to fill out an answer. Both pieces run on the same ranking systems behind the standard search results.
The argument lands in one sentence. A page blocked from ranking for a query will not appear as a cited source in the AI response for the query. SEO drives the path. AI citations follow as a result.
Five Tactics Google Tells You to Drop
The mythbusting section of the guide names five practices the industry sold as GEO or AEO services over the past two years. Google says none of them help with Search visibility.
1. llms.txt files for SEO
The llms.txt proposal suggests websites publish a markdown summary at the root to help AI tools understand the site. Vendors pitched the file as a path into AI Overviews. Google’s response: you do not need new files, AI text files, or markdown to appear in AI search. The file does have a real use outside of Search. More on llms.txt below.
2. Chunking content into AI-friendly fragments
A common pitch over the past 18 months: rewrite articles into 40 to 60 word answer blocks under question H2s. Google’s stance is direct. Content does not need to be broken into tiny pieces for AI to understand. Question-and-answer structure still works for featured snippets and human readers. Mechanical chunking earns nothing extra from AI.
3. Machine-friendly content rewrites
Some vendors recommended stripping voice, swapping pronouns for proper nouns, and flattening prose into bullets to help AI parse the page. Google says modern AI handles synonyms and meaning well. Brand voice and expert perspective work in your favor, not against you.
4. Inauthentic mentions
Paid placements in forums, low-trust syndication, and comment-spam networks have been sold as AI citation building by some vendors. These overlap with Google’s existing spam rules. Earned mentions on trusted publications still help, but manufactured ones add risk.
5. Schema as an AI hack
Schema.org markup helps you qualify for rich results. Google says schema is not required for AI search, and no special schema type drives AI citations. Keep using schema for what schema does.
Three Priorities Google Wants You to Keep
The guide reduces to three priorities:
1. Publish non-commodity, people-first content
Google draws a clear line between commodity content (general advice anyone produces) and non-commodity content (firsthand experience, original data, expert insight). Primary sources earn citations. Restated summaries do not. Search Engine Journal covered the same point this month, calling primary-source content the new minimum bar for AI visibility. For background on how this fits the broader picture, see our piece on AI’s impact on search.
2. Build clean technical structure
Indexability is the price of entry. A page blocked from Search is blocked from AI search. The technical checklist holds:
- Pass Google’s technical requirements and snippet eligibility
- Protect crawl budget on larger sites
- Hit Core Web Vitals targets
- Render important content server-side, not only through JavaScript
- Resolve duplicate content and canonical issues
For an entry-level walkthrough of how search engines crawl and rank pages, see our guide on SEO basics.
3. Tune local and product signals where you sell
Brands with physical locations or products in the catalog should keep Google Business Profile and Merchant Center feeds accurate. AI Overviews pull from both. The new Business Agent product, a conversational layer on the SERP, draws from the same sources.
The llms.txt Question
Two Google teams shipped different guidance in the same week. Search said you do not need llms.txt. Chrome’s Lighthouse audit, published 10 days earlier, recommends llms.txt for agentic browsing. Not a contradiction, two different audiences.
Google Search indexes the entire public web at scale. The system does not need a curated index file because the system already holds the data. An AI agent acting on your site at inference time has no index, no time, and no patience. A llms.txt file lets the agent find what the agent needs in seconds.
Build llms.txt when:
- Customers will use agents to shop, book, schedule, or self-serve on your site
- Your documentation or product catalog is dense and benefits from a curated entry point
- Your dev team or procurement runs Lighthouse audits and flags the missing file
Skip llms.txt when:
- Your site is a brand or service overview with no agent-driven flows
- Goals tie to traditional Search visibility alone
What Agencies and In-House Teams Should Do This Quarter
Three actions handle the bulk of the response:
- Fold any GEO or AEO line item back into SEO. One discipline, one budget, one set of priorities.
- Run a robots.txt and crawlability audit for AI bots. Many sites block them by accident.
- Decide whether agentic browsing is real for your audience. If yes, scope an llms.txt build. If no, skip the file.
The Bottom Line
Google’s guide rewards work most strong SEO teams have done all along: clean indexing, primary-source content, real expertise, and earned authority. The GEO/AEO category did not vanish. The work folded back into SEO where the work has always lived.
If your program produces unique research, expert content, and a fast clean site, the AI search shift works in your favor. If your program leans on hacks, the next few quarters get harder.
Need a second opinion on your SEO and AI search readiness against the May 2026 guide? Reach out to our team, or browse our services.
