Welcome to episode 18 of The GEO Show, the briefing on generative engine optimization, LLM citations, and AI search visibility. Brought to you by GEOforge.
In this episode, Paris Childress, founder of Hop AI and co-founder of GEOforge, walks through a Friday, September 4, 2026 docket that starts with Gemini citation bugs and engine disagreement, then hits the biggest unlock of the day: turning video UGC into machine-readable GEO content, before closing on Cloudflare's September 15 default AI crawler block.
In this episode:
๐ Google says Gemini 3.8 citation gaps are a bug
Early Gemini 3.8 Flash testing in Google AI Mode showed many top-of-funnel answers with no links or citations. Google Search VP Robbie Stein said that is not working as intended and a fix is coming. Annotate share of voice swings that may have nothing to do with your GEO work.
๐ AI engines can disagree nearly 2x on whether a B2B brand exists
Treyci analyzed more than 1,200 scored AI answers. In one B2B software category, tracked brands appeared in 81% of answers from one engine versus 43% from another. A single ChatGPT visibility score is almost meaningless. Report engine-level share of voice and weight engines by where buyers actually research.
๐ A 12-month GEO field study finds no special AI hack
The HOTH says its strongest citation gains across dozens of campaigns came from technical accessibility, deeper commercial content, earned media, and authority. On its own data, AI-referred visitors converted at 22.79% versus 2.45% for standard organic. Roughly a 10x conversion gap. GEO looks less like a replacement for SEO and more like a different weighting of authority signals.
๐งญ Botify adds competitor citation intelligence
Botify added competitor mention and citation share of voice, including which domains generate competitor citations, plus semantic grouping of related prompts. Measurement alone is not a moat. AI visibility monitoring is becoming a commodity inside SEO suites. Value is moving to the execution layer.
Video UGC is becoming machine-readable GEO
Reported via the National Law Review, VideoWise launched tools that convert TikTok, Instagram, and user-generated video into structured content AI systems can retrieve and cite, including YouTube GEO. Paris called this the biggest story of the day: unlock proprietary knowledge stuck in webinars, customer interviews, demos, and SME videos by transcribing, structuring, and making it crawlable.
โ ๏ธ ChatGPT, Claude, and Grok went down almost together
On September 3, ChatGPT, Claude, and Grok experienced outages in roughly the same morning window. No shared cause confirmed. OpenAI pointed to a routine error. Treat AI engines as infrastructure, and annotate odd measurement windows the same way you would a Gemini bug.
๐งช New evidence weakens schema as a GEO citation layer
Search Engine Land reviewed evidence including an Ahrefs test of 1,885 pages adding JSON-LD. No meaningful citation uplift across ChatGPT, AI Mode, or AI Overviews. AI Overview citations declined 4.6% in the tested cohort. Cited pages on AI Overview SERPs generated roughly 120% more clicks per impression than uncited pages. Schema still helps machine clarity. The evidence does not show it drives citations.
๐ญ๐ฐ Moonshot AI's Kimi ecosystem is preparing for a possible $3B IPO
Reuters: Moonshot AI confidentially filed for a Hong Kong IPO that could raise about $3 billion, with valuation around $50 billion in an ongoing funding round. Do not assume AI search stays a four-engine Western market. Especially for APAC, plan market-specific engine coverage.
๐ง Profound moves from measurement toward a business knowledge brain
Profound published Context Manager Structure on September 1 (corroborated by Search Engine Watch on September 3): brand context, personal context, knowledge base, and synced data across Gong, Slack, and email, auto-distilled into six brand records. Enterprise only at launch. Architectural cousin to GEOforge BaseForge knowledge ingestion.
๐ ๏ธ AirOps is repositioning as platform plus services
CEO Alex Halladay's September 2 post says AirOps is combining platform with channel expert services across owned, earned, AI ads, community, and social/influencer, tied to pipeline. Proprietary figures cited: 6 billion citations, 47 million actions, 1.5 billion prompts tracked; 86.6% of marketing leaders prioritize AI search, only 23.2% feel confident measuring it. A hedge against tool commoditization.
๐ก๏ธ Cloudflare will default-block certain AI crawlers on September 15
Effective September 15 (announced July 1), Cloudflare will by default block training and agent category AI bots, but not search category bots, on ad-monetized pages for new domains. Multipurpose crawlers such as Googlebot, Applebot, and Bingbot can get blocked for customers who choose to block training. Paris's policy: allow legitimate bots, watch for spoofing, manage the new default consciously.
If this briefing is useful, subscribe to The GEO Show so we can keep walking through the studies, platform changes, and citation data that actually move AI visibility.
๐ฌ Question: If your best answers still live only as MP4s, what is stopping you from transcribing and structuring them this week?