Google's John Mueller said yesterday that there is nothing special to do for AI answers, which is the single objection our sales page has to beat. It is also a claim about one engine out of five.

A Bluesky user asked John Mueller whether there are industries where GEO simply does not matter yet, naming sectors where Google still drives most discovery. Mueller replied: "I'm not quite sure what you're asking; from our POV there's nothing really special you need to do for generative AI responses in search." SEJ's framing is that Google's AI results draw from the same crawled index and tend to pull from pages already ranking, so no separate optimisation is required.

Worth putting one older number beside it. Ahrefs analysed 540,000 query pairs of September 2025 US data and found that only 13.7% of cited URLs appeared in both AI Overviews and AI Mode for the same query, with 16.3% overlap in the top three citations (Ahrefs, 15 December 2025, https://ahrefs.com/blog/ai-overviews-vs-ai-mode). Two Google surfaces, the same question, and they mostly do not cite the same pages.

Older material, never reported here, and the closest thing yet to a measurement of our actual buyer. Searchable, an AI visibility vendor, ran two tests. In the first, 165 London businesses were put to ChatGPT, Gemini and Perplexity across 13,365 questions covering services, contact details, size and founding dates, and the answers were graded against Companies House records and official profiles: 93% of those businesses had at least one basic fact wrong or missing, and half of the smaller businesses got at least one outright false fact against 32% of the larger ones.

In the second, the same three engines were asked more than 72,000 questions about UK high street retailers, graded against each retailer's verified information. The engines answered 98% of the time. One in 16 answers was false. The most common error was the wrong postcode, at one in ten, even though the prompts named the town or city. Around one in 15 website answers pointed to a dead link, a lookalike, or a different business. Wrong brand attributed to a shop name ran at one in 40. Overall, 64% of the retailers had at least one false fact returned about them.

Read it with the caveats attached. Searchable sells the product this research recommends. It is UK only. The retailer count is not disclosed in any coverage found, no fielding window is stated, and no primary report page surfaced in this sweep, so every figure here is a press-release figure carried by trade press. The direction is still the useful part, and it is corroborated in shape by the size split: the smaller the business, the thinner the corroborating web evidence, the worse the answer.

Older, and it closes an item this radar logged as an anecdote on 13 August. Lily Ray ran 100 B2B "best [category] software" queries through Ahrefs Brand Radar at three checkpoints, 15 April, 15 May and 8 June. AI Overviews triggered on 80 of the 100. Across 323 instances where Google cited a brand's own self-promotional listicle, there were 224 cases, roughly 69%, where Google cited that brand's page and then recommended somebody else from it.