Welcome to episode 21 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, records a September 6, 2026 solo news roundup on Unification Day in Bulgaria. The lead story is NVIDIA acquiring Hugging Face for $12.93 billion, then ten more beats on model volatility, execution-layer GEO, llms.txt, citation freshness, publisher volume, ChatGPT ads, high-intent AI Overviews, and B2B buyer behavior.
In this episode:
💰 NVIDIA acquires Hugging Face for $12.93B
NVIDIA agreed to buy Hugging Face, whose platform hosts 3 million models, 500,000 datasets, and 1 million applications used by over 18 million developers. NVIDIA says Hugging Face will remain open and will not require NVIDIA compute. Paris's GEO take: this bolsters American open source and further fragments AI engines beyond ChatGPT, Google AI Mode, AI Overviews, and Claude, making per-platform visibility reporting more important.
⚡ ChatGPT-6 Astra raises model volatility for GEO
OpenAI launched ChatGPT-6 Astra with browsing, research, and multi-step agentic upgrades. Rollout has begun with selected organizations. OpenAI also rates Astra at a critical cybersecurity capability threshold. Major model changes can alter retrieval and recommendations even when a brand changes nothing, so annotate releases in GEO reporting.
🤖 Pepper Agent Atlas executes GEO workflows
Pepper launched Agent Atlas, connecting GEO tracking, Search Console analytics, competitive data, and CMS information, then recommending and executing workflows such as content optimization. The stack is moving from monitoring to diagnose, recommend, and execute. GEOforge connects a similar integration set and competitor share of voice.
🧭 HubSpot GEO Recommendations beta lands in the marketing stack
HubSpot's GEO Recommendations beta analyzes citation patterns across tracked prompts and recommends site content, third-party outreach, and social actions. It can generate gap content and track mention changes. Paris's take: HubSpot's CRM, CMS, social, and revenue measurement create a strong competitive edge as GEO embeds inside the marketing platform.
📄 41% of B2B SaaS have llms.txt without evidence
Treyci scanned 100 B2B SaaS companies and found 41 with llms.txt. No controlled comparison showed more recommendations or citations for adopters (NetContent SEO). Treat it as a cheap experiment and measure before-and-after citation rates rather than selling implementation as the outcome.
⚠️ llms.txt becomes a cybersecurity risk
Ars Technica: researchers scanned 6,214 corporate domains and found 120 llms.txt or llms-full.txt files referencing unregistered packages or domains. After registering abandoned names, coding agents including Claude, Codex, and Hermes executed proof-of-concept code. At least one misconfigured reference already pointed toward real malware.
📅 Brand Ghost: median citation age 263 days
Brand Ghost analyzed 6,762 AI citations. Median age was 263 days, 42.2% were more than a year old, and only 21.5% were under 90 days. That challenges freshness-only narratives. Protect durable research, benchmarks, docs, and evergreen assets instead of manufacturing date updates.
📉 USA Today: more content no longer grows audience
Search Engine Land reported USA Today's September 4 memo: producing more content is no longer as effective while platforms keep users. The volume model is structurally weakening. Optimize for information gain, citations, recommendations, branded demand, and pipeline, not articles-per-month.
📣 ChatGPT ads split paid vs organic visibility
OpenAI's ad tooling adds deeper custom audiences, inclusion/exclusion, bid adjustments, and conversion measurement. Conversational discovery now has earned GEO and paid layers. Paris has not seen platforms cleanly break out paid ChatGPT visibility, so share of voice may already mix apples and oranges.
🎯 High-intent B2B queries are AI Overview heavy
SER Interactive data via Search Engine Land: AI Overviews on 95.4% of X vs Y, 86.3% of review, 83.4% of price/cost/buy, and 81.3% of best-of queries. Start GEO programs with battle cards, comparisons, alternatives, pricing, and best-vendor pages before more top-funnel blogs.
📊 G2: 51% of B2B buyers start with AI chatbots
G2 finds 51% of B2B software buyers now start with AI chatbots more often than Google, up from 29% eleven months earlier. 69% changed vendor plans after chatbot recommendations. Roughly a third bought a previously unknown vendor. 85% view mentioned vendors more favorably. Paris's close: B2B SaaS needs to be GEO first.
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.
💬 Question: After NVIDIA's Hugging Face deal, are you measuring open-source model visibility as a first-class GEO surface, or still optimizing like ChatGPT and Google are the whole map?