
Only 6.9% of ChatGPT prompts involve researching something to buy. On Google, the figure is more than 15%. The first real panel data on AI assistant demand is smaller than the hype, and it landed in the same week AI got a lot more permission to act.
Google Search now builds comparison tables and custom tools for queries worldwide. Ask Maps phones local businesses to book appointments for the customer. Meanwhile, two AI agents switched off the systems built to monitor them, and a regulator filed the first lawsuit over rogue agent incidents.
The pattern repeats everywhere. Machines are moving from answering to acting, while the controls, metrics and org charts around them were built for tools that wait to be told. Anthropic's $900.9 billion compute bill shows what that capacity costs. A study of 1,819 scientists shows how easily we mistake feeling faster for being faster.
Seven pieces below on what to let the machines do, where to put the fence, and how to tell whether any of it is working.
Split your Search Console data into branded and non-branded
Google Search Console now classifies queries as branded or non-branded, in the Performance report's Query filter and, as of this week, in the Search Analytics API through a new brandedClassification dimension. This week, open Performance, filter to "Non-branded queries", and compare the last three months with the period before. Non-branded clicks show whether you are winning new demand; branded clicks show whether AI answers and other channels are sending people who already know your name. The data starts on Feb. 21, 2026, covers top-level properties only and is AI-classified, so treat it as directional and annotate anything that looks misfiled.
Happy Growth,
Dan
Subscribe to The Weekly Vibe
Every Tuesday. 5-7 original takes on what matters in AI, Marketing, and Business Growth. No spam, no fluff, unsubscribe anytime.