Google Just Told You What Drives AI Visibility. Most of Your AEO Budget Is Waste.

Google published its first official AI optimization guide in May 2026. The core message: AI Overviews and AI Mode are built on the same ranking and quality systems as traditional search. If you have been paying for AEO-specific tactics — llms.txt files, content chunking, AI-targeted rewrites — Google just told you those do not work. The three inputs that actually drive AI citation are original content, clean technical structure, and current business data. Everything else is noise.

Google Debunked Five "AI Optimization" Tactics by Name

The guide is unusually direct. Google explicitly listed five common tactics it says to abandon:

1. llms.txt files — treated like any other text file; no special indexing pathway.
2. Content chunking — AI systems understand multi-topic pages without pre-fragmentation.
3. AI-specific rewrites — the models handle synonyms and intent variations already.
4. Manufactured mentions — inauthentic brand references across the web do not improve AI citations.
5. Over-reliance on structured data — schema helps rich results, but it is not an AI citation hack.

If your agency sold you a "GEO audit" that recommended any of these, the ROI case just collapsed. Google's own documentation routes the work back to core Search: the AI features guidance tells site owners to apply the same foundational SEO best practices they already apply to Search overall.

What Actually Drives AI Visibility: Three Inputs

As Semrush's analysis noted, the guide's core statement is clear: "The best practices for SEO continue to be relevant because our generative AI features on Google Search are rooted in our core Search ranking and quality systems." Google's guide identifies three factors that determine whether your content appears in AI Overviews and AI Mode:

Non-commodity content. First-hand expertise, proprietary data, original analysis. Google's phrasing: content that offers "unique, experience-based insights rather than generic summaries." This is the highest-leverage input. If your content can be accurately reproduced in a two-sentence AI summary, it will be summarized away — not cited.

Clean technical structure. Indexed, crawlable pages with solid Core Web Vitals and semantic HTML. Not AI-specific markup. Standard SEO hygiene that most enterprise sites already have or should have.

Current business data. Google Merchant Center feeds and Google Business Profile for product and local information. If you are a B2B company, this means keeping your structured business presence current — not just your content calendar.

The CTR Data Makes This Urgent

The stakes are measurable. In Seer Interactive's first AI Overview CTR study (about 10,000 informational keywords where the clients ranked in Google's top 20, January 2024 to January 2025), organic CTR on queries where an AI Overview appeared fell from 1.41% to 0.64% year-over-year, a 55% decline. Seer notes that those queries already had lower CTR before AI Overviews arrived, so the drop is not all attributable to the AI Overview, and its April 2026 update reports that the 2025 decline did not continue into early 2026.

But here is the operational signal: Seer Interactive's April 2026 update puts paid CTR at 15.74% for brands cited inside the AI Overview against 11.19% for brands that are not (informational queries, 2025 full-year average). Being cited is not just a brand play: the cited side of the same results page carries the higher paid click-through, though Seer reports the gap as correlation, not proof the citation causes it.

The operators who built real content authority — original research, expert analysis, proprietary data — hold the inputs Google's guide says it relies on. AI-specific markup is not among them.

The Google Guide Only Covers Google

One critical caveat flagged by Frase's analysis: this guide governs Google's AI surfaces only. ChatGPT, Perplexity, and Claude operate independently with different retrieval systems.

Google confirmed that its AI features use retrieval-augmented generation (RAG) — pulling indexed pages, extracting information, and providing clickable citations. Other engines use different retrieval architectures. What earns a citation in Google AI Mode may not appear in Perplexity's answer, and vice versa.

This is why Machine Relations is the operational frame here: AI visibility is not a single-engine optimization problem. It requires source architecture that works across retrieval systems — original claims, structured evidence, entity authority. That is what compounds across engines, not engine-specific tricks.

Your Monday Audit

Three things to check this week:

1. Kill the AI shortcuts. If anyone on your team is maintaining llms.txt files, chunking content for AI consumption, or rewriting pages with AI-specific phrasing — stop. Google confirmed these are wasted effort. Redirect those hours to original content production.

2. Measure your citation rate, not just your rankings. Track whether your content appears inside AI Overviews for your target queries. In Seer's data, cited brands had higher CTR than uncited ones on the same results pages. AuthorityTech has covered this measurement framework — the click-tracking model is already broken.

3. Audit content for summarizability. Open your top 20 pages. For each one, ask: can an AI engine accurately answer the user's query by summarizing this page into two sentences? If yes, that page will be summarized away — not cited. The fix is adding proprietary data, original analysis, or expert perspective that cannot be compressed.

FAQ

Does Google's AI optimization guide mean AEO is dead?

Not exactly. AI engine optimization as a concept — making your content visible in AI-powered search — is real and growing in importance. What Google killed is the tactical layer that emerged around it: llms.txt files, content chunking, AI-specific rewrites, and manufactured mentions. The guide confirms that the inputs to AI visibility are the same inputs that drive traditional search quality: original expertise, technical crawlability, and current structured data. The discipline survives; the shortcuts do not.

Should I optimize differently for ChatGPT and Perplexity than for Google AI Mode?

Yes, but not with different tricks — with deeper source architecture. Google's guide only covers Google's retrieval system. ChatGPT, Perplexity, and Claude each use different retrieval approaches. What works across all of them is content that carries original claims, structured evidence, and clear entity authority. Engine-specific tactics are a treadmill. Source architecture that earns citation across multiple retrieval systems is what compounds.