What changed in AI search, 25 to 30 October 2026

Which pages AI answers cite

TJ Robertson, founder of the GEO agency TJ Digital, published a study of 193,507 AI answers on 25 October 2026. The answers covered 4,167 buyer questions in ChatGPT, Google AI Mode and Google AI Overviews from 27 August to 21 October 2026. In 195 blind-rated pairs, the cited page opened with a direct answer 20% of the time, against 5% for a Google top-10 page that never came up in the AI’s searches. Robertson reports these as associations without proof of cause. Brands should open every page built for a buyer’s question with the answer, because AI engines pick pages whose opening carries what the question asks, which is part of getting pages cited in Google AI Overviews.

Robertson’s two rounds of blind ratings compared cited and passed-over pages:

In an earlier round, among pages the same answer pulled up, cited pages opened with a direct answer 23% of the time and uncited pages 14%.

How people prompt AI versus Google

Cate Dombrowski, a researcher at iPullRank, published on 29 October 2026 a comparison of the same 16,310 users on Google and AI search, matched across two Datos panels. The panels cover US desktop users in November 2024, and the AI panel recorded prompts typed into Perplexity, Copilot and you.com. The same users averaged 8.4 words in AI search and 4.2 on Google, and Alphabet has published its own figure on longer AI Mode queries. For 96% of AI prompts, the same user didn’t search that topic on Google within 30 minutes. Buyers bring an AI assistant their whole situation and get one recommended solution back, so a brand needs pages that answer those situations in the buyer’s words.

The same users asked different kinds of queries on each surface:

Informational questions made up 26.8% of AI prompts and 8.1% of Google queries, and 8.0% of AI prompts revealed the asker’s role, situation or purpose, against 1.2% on Google.

Knowledge Graph facts in AI Mode

Dan Petrovic of DEJAN published on 28 October 2026 that Google AI Mode receives Knowledge Graph biographies of people and companies as search results with no URL. Asked for Petrovic’s age and the raw source data, AI Mode returned a “Person Biography” record with no link in two incognito sessions, and cited only a Wikipedia page about a footballer. A test on owayo returned an “Organization Biography” with the company’s founding year, founders and headquarters. DEJAN notes that AI Mode itself wrote these records when asked. What AI says about a brand outweighs how often it says it, so a company should check what Google’s Knowledge Graph holds about it.

ChatGPT search volume against Google

Mitul Gandhi, co-founder of seoClarity, published on 28 October 2026 an estimate that ChatGPT’s daily search-equivalent activity is approximately 7.5% of Google’s search-query volume. The estimate puts ChatGPT at about 1.24 billion search-like messages a day against 16.44 billion modeled Google queries. seoClarity built the estimate from OpenAI’s Signals v3.0 usage data plus its own assumptions, such as 20% growth on Google’s five trillion annual searches, and calls it a modeled comparison. ChatGPT at that volume needs its own count of how often AI answers mention a brand, because buyers who met a brand in an answer return later through Google or direct, and analytics undercount them.