Welcome to episode 39 of The GEO Show, the briefing on Generative Engine Optimization and AI visibility. Brought to you by GEOforge.
In this episode, Paris Childress, founder of Hop AI and co-founder of GEOforge, covers ten stories: Profound's Dynamic Bot Rendering, Google Lighthouse Agent Resource Discovery, citation fidelity near 50%, crawler spoofing, YouTube's 25.66% share of GEO citations, weak country-code domains, llms.txt placeholders, the AI Recommendation Gap, Swap Discovery inside Shopify, and Thailand's Recharge Land tracker.
In this episode:
- 🔦 Google Lighthouse makes agent discovery a web audit
- Lighthouse 13.5.0 (September 18) adds Agent Resource Discovery (ARD) and groups ARD with llms.txt under a new agent discovery category. Google expected Chrome 156 DevTools and PageSpeed Insights within roughly two weeks. Paris tested getgeoforge.com live: this is Google treating agent crawl readiness like a first-class web quality signal.
- 🤖 Profound moves from observing agents to serving them
- Profound's September 18 Dynamic Bot Rendering detects answer-engine agents and can serve a fully rendered page without CMS changes.
- Policies vary by bot, page, pattern, or domain, with CloudFront, Cloudflare, and Vercel support. Delivery is becoming part of GEO, not just dashboards.
- 📊 Only about half of resolvable citations support the claim
- Across 15,525 citation events, 54.8% of resolvable ChatGPT citations and 50.1% of resolvable Perplexity citations actually supported the anchored claim. Citation presence is not evidence quality.
- 🕵️ 73% verified, 27% fake: crawler dashboards without IP checks
- Machine Relations audited spoof-prone AI crawler traffic. On a corrected one-day run with IPs retained, 73% matched vendor ranges and about 27% did not. User-agent matching alone contaminates AI bot analytics.
- ▶️ YouTube is 25.66% of GEO category citations
- In the AI visibility / GEO buying category, editorial publications were only 5.9% of citation weight versus 11.3% across 16 categories. YouTube led the source set at 25.66%. For this category, YouTube is the highest-leverage citation tactic Paris sees.
- 🌐 Country-code domains barely register
- Across 15,883 answer runs, national ccTLDs were 6.58% of cited domains but only 2.73% of citation volume. .ai and .io together captured 8.6%, about 3× all national ccTLD volume. A local suffix is not proof of local retrieval.
- 📄 More than half of llms.txt adopters ship placeholders
- GeoReady audited 282 domains: 62.8% had llms.txt, but 52% of adopters were placeholders. There is still no confirmed causal link to citation wins. Existence is not quality.
- 🎯 The AI Recommendation Gap
- Polish agency Funky Media frames Known → Mentioned → Cited → Considered → Recommended. A brand can be cited and still disappear when the user asks which provider to choose. Visibility is an intermediate metric.
- 🛒 Swap Discovery writes GEO fixes into Shopify
- Swap measures e-commerce visibility by assistant and country, then writes product, collection, FAQ, and structured-data improvements into Shopify. Vertical CMSs are absorbing measurement plus execution.
- 🇹🇭 Thailand gets a multi-run GEO tracker
- Recharge Land launched a locally built tracker covering Google AI Overviews, ChatGPT, Gemini, and Copilot with multi-run and competitor analysis. GEO is going regional.
Subscribe to The GEO Show wherever you get your podcasts, and watch the full episode on YouTube: https://youtu.be/nKsrvaP4oJk
💬 Question: If Google is auditing agent discovery in Lighthouse, did you run PageSpeed Insights on your site this week, or are you still guessing whether agents can read you?