Always-On AI Agents Will Outgrow Prompt Tracking

Welcome to The GEO Show, the briefing on Generative Engine Optimization and AI visibility. Brought to you by GEOforge. In this solo roundup, Paris Childress, founder of Hop AI and co-founder of GEOforge, covers nine stories from the week in AI search. The lead story: OpenAI's always-on Dots agents, and why prompt tracking will capture less of a brand's AI visibility as agents do more of the prompting. In this episode: - **🤖 OpenAI Dots and the future of prompt tracking** OpenAI launched Dots on September 29: always-on agents powered by GPT-6 Astra, each with its own cloud computer and access to more than 4,000 apps. As agents start conversations on their own, stateless prompt tracking becomes a less complete picture of a brand's AI visibility. - **🔎 Perplexity trains retrieval to return evidence** Perplexity released pplx-embed-v2-context-9b-preview on September 30, a contextual embedding model trained to retrieve both the answer chunk and the supporting context. The unit being optimized for retrieval is becoming an evidence bundle, not an isolated quotable paragraph. - **⏳ The median AI citation half-life is 11 days** Profound tracked 883,000 pages across seven AI engines. The median page loses half its citation share within 11 days of peaking, and 78% fall to half their peak within two weeks. Engines decide independently, so watch citation persistence and refresh or replace pages as they decline. - **🏘️ Third parties win commercial citations** In Writesonic's analysis of 9 million AI answers for 400+ enterprise brands, published on Search Engine Land, third-party sources got 82% of citations on bottom-of-funnel commercial prompts and owned pages just 3%. Owned citations lasted 3 to 9 times longer. The data comes from Writesonic customers, not the whole market. - **🔌 Buyer intent data inside AI tools** Informa TechTarget launched an MCP server for its Buyer Intelligence, bringing person-level intent data from its audience of about 58 million permissioned members into Claude, ChatGPT, Copilot, Gemini, and custom AI systems. - **📊 AI visibility moves into communications reporting** 10Fold launched MetricsMatter 5.0. In its survey of 400 US and European B2B technology marketing decision-makers, 58% include AI search or LLM visibility in integrated reporting, but only 38% connect communications and visibility metrics to pipeline or revenue. Paris's fix: add an AI chatbot option to "How did you hear about us?" on your lead form. - **📰 PR distribution meets GEO measurement** Brandpoint launched Optimize360, which tracks AI visibility with more than 30,000 prompts a month per brand. Brandpoint advertises a network of 3,000+ publications and 800+ guaranteed placements, the hardest part of off-site citation building. - **🏭 Vertical GEO needs vertical logic** Zen Media ran 1,000 industrial automation buyer prompts across ChatGPT, Gemini, Perplexity, and Grok. The top 100 vendors averaged 4% visibility, versus 21% for Siemens, and distributors and authorized channels showed up prominently. - **🇦🇺 Source authority is local** Findrix's Australian AI Source Index covers 23,396 buying prompts in nine markets. More than 85% of retrieval in clinics, home services, and finance and accounting came from .au domains, versus 32.9% in B2B software. There is no universal best-sources list. Subscribe to The GEO Show wherever you get your podcasts. Full episode on YouTube: https://youtu.be/t6YnwMMyMe4 💬 Question: How will you measure AI visibility when agents, not people, start the conversation?