Tinuiti's Amazon split (4% of Copilot citations, 0.1% on ChatGPT, zero on Gemini) is why Dept is building for every surface at once, and why some brands are already moving media money to shape chatbot descriptions. IAB's Caroline Giegerich spent four months on a visibility framework because the same $200 sneaker prompt can return 100 different answers.
Why it matters: Treat AI visibility as a media line with a named score and a source list, not a GEO side project.
Four service lines, one blended ROI, and the algorithm poured spend into the business that hadn't funded the test. Cisco's Anthony Tedesco is feeding proxy events because pipeline is too sparse to train the model, and Maria Corcoran killed three hours of daily reporting by letting Claude join finance, ads, and sales.
Why it matters: Split any AI Max or PMax test by P&L before you turn it on, and pick a high-volume proxy the model can actually learn from.
On a B2B account, 83% of the terms that trigger AI answer blocks are informational. Ana Kostic's team now sorts every query into Fight, Influence, or Generate demand, and they stop paying CPC where SEO already owns the answer. SE Ranking's July commercial-query sample: ads show up on 54% of $10+ CPCs and only 24% under $2.
Why it matters: Re-bucket the search term report this week and move Influence money into citation work before Q4 CPCs climb.