Google Told You to Optimise for Google

In May 2026, Google published a guide to generative AI optimisation. The headline message: “Optimising for generative AI is still SEO.” Nick Fox, Google’s VP of Search, has said the same thing in public. A senior Google executive, using the authority of one of the most trusted technical brands on earth, telling marketers clearly: the skills you have are the skills you need.

The statement is correct. It is also written by Google, for Google. And Google is no longer the only AI your customers are using.

What the data actually shows

The arXiv study that tested this most rigorously (Chen et al., University of Toronto, January 2026) measured domain-level overlap between each major AI system and Google’s top-10 search results across 1,000 queries:

• GPT-4o: 4.0% overlap with Google
• Gemini: 11.1%
• Claude: 12.6%
• Perplexity: 15.2%

GPT-4o — the engine behind the most-used AI chatbot on the planet — shares four percent of its cited domains with Google’s top results. Not forty percent. Four.

BrandFeatured’s analysis adds a further dimension: less than 1% of citations overlap between ChatGPT and Perplexity. The two most-used AI answer engines are drawing from almost entirely different source pools.

Google’s own AI products sit at the high end of the overlap range (Gemini at 11%). This makes sense: Gemini is built on Google’s infrastructure and uses Google Search grounding. Of course it draws from Google’s index. That’s also why Google’s guide is correct — for Google’s AI.

The guide answers the right question about the wrong platform

Nick Fox’s statement is accurate. For Google AI Overviews, 76% of citations come from pages already in Google’s top-10 results (Ahrefs data). Rank in Google, appear in Google’s AI. The logic holds. The advice works.

The problem is the implicit premise: that optimising for Google is optimising for AI search. That premise was once defensible. A strong Google presence got you most of the way there across platforms.

By mid-2026, that premise has collapsed. A page that ranks well in Google has roughly a one-in-twenty chance of appearing in GPT-4o’s cited sources for the same query.

Google’s guide doesn’t mention this. It doesn’t need to. It is a guide to Google’s products.

The platforms diverge in ways that change the strategy, not just the tactics

The divergence is not just about which pages each AI retrieves. It runs through every lever brands can actually pull.

Earned media vs owned content. The arXiv study found GPT-4o cites 57% earned media; Gemini cites 46% earned and 46% brand-owned content. Google’s own AI gives substantially more weight to brand-owned pages than ChatGPT does. Optimising your website for Google’s AI is rational. That same owned content is a marginal signal for ChatGPT, which is dominated by independent editorial coverage.

Freshness windows. ChatGPT’s median citation age in consumer electronics is 80 days. Google’s is 130 days. A content programme calibrated to Google’s freshness cadence — one that refreshes every quarter — may be running too slow to reach ChatGPT’s retrieval window for the same queries.

Outlet profile. Claude cites Reuters roughly 50 times less than ChatGPT. Claude’s top-cited journalism outlets include Good Housekeeping, TechRadar, and Harvard Business Review. ChatGPT and Gemini share an almost identical outlet profile anchored by Reuters, the Financial Times, Time, Forbes, and Axios. A PR programme built for Google’s citation preferences will serve Gemini well and largely miss Claude.

These are not marginal differences. They are structural divergences across the platforms that now collectively account for the majority of AI-mediated information discovery.

The market share shift makes this urgent

ChatGPT’s share of AI chatbot traffic was 87% in early 2025. By May 2026 it had fallen to 64%. US mobile share had dropped below 40% for the first time. Gemini went from 5.7% to 21.5% in the same period.

Most AI visibility advice was written for an 87%-ChatGPT world. That world is gone. The distribution across AI platforms is fragmenting rapidly — which means the gap between “optimise for Google” and “optimise for AI” is widening every quarter.

Google’s guide was accurate when it was published. The market it describes is changing underneath it.

What Google’s guide gets right (and why that still matters)

The guide’s underlying recommendations are not wrong. Demonstrate expertise, publish original research, earn independent editorial coverage, maintain consistent entity signals. These are correct — and they are the inputs that help with ChatGPT, Perplexity, and Claude as well as Google’s AI.

The fundamentals converge. The execution diverges.

Which outlets you target for earned coverage. How fast you refresh content. Which platforms you prioritise when budgets are constrained. Those decisions are platform-specific. Google’s guide does not help you make them, because it was not written to.

The question the guide doesn’t ask

Google’s guide starts from the premise that you are optimising for Google. That is a reasonable premise for a guide published by Google.

It is not a sufficient premise for a brand strategy in 2026.

Your customers are not on one platform. A B2B buyer researching software options may be using Claude or Perplexity. A consumer comparing products may be in ChatGPT on iOS. The AI-mediated moments that shape purchase decisions are distributed across platforms with fundamentally different citation logic.

The question the guide doesn’t ask is the one that matters most: which AI is your customer using when they form the opinion that drives the decision?

Google’s guide is the right answer. It’s just the answer to a narrower question than the one you need to be asking.

Sources: Google AI Optimization Guide (https://developers.google.com/search/docs/fundamentals/ai-optimization-guide, 2026-05-15); Chen et al., arXiv 2601.16858 (University of Toronto, January 2026); BrandFeatured AI ranking factors analysis (2026); Ahrefs AI Overview citation study; Muck Rack Generative Pulse (via Nieman Lab, July 2025).