Welcome to episode 17 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 Thursday, September 3, 2026 docket that starts with Perplexity citing mass-generated best-software pages, then moves through model-version visibility, recommendation versus citation, vendor deals, and bot management as part of the GEO stack.
In this episode:
📉 Perplexity is citing pages Google would ignore
Trelner tested 380 software categories across Perplexity, Sonar, and Sonar Pro and collected 7,534 citations. 59.8% came from domains outside the top 100,000 on Tranco Traffic. GuideFlow, a product demo vendor, was the third most cited domain, ahead of Gartner. Three related sites held 215,128 generated best-category software pages, some titled "Facts and Grounding Page." Retrieval can reward sources with little conventional authority if they align with machine retrieval. SEO authority and GEO citation authority are diverging. Audit which URLs ground competitor recommendations, not just backlinks.
⚡ Google AI Mode is rolling Gemini 3.8 Flash
Search Engine Land: Google began rolling Gemini 3.8 Flash into AI Mode for Google AI Pro and Ultra users globally. The model is described as materially improved in multi-step reasoning and agentic tasks. AI search visibility is partly model-version dependent. A prompt benchmark can shift without the brand changing anything. Record model and date in GEO measurement and rerun controlled baselines after major model changes.
🎯 Citations are intermediate. Recommendation is the goal
At Search Seoul, Yoonhee Choi, RTN's SEO director, argued GEO should move from get-cited toward make-AI-choose-us: structure the relationships among brand, product, capabilities, user needs, and use cases so AI knows when to recommend a brand. Knowledge structuring and ontology beat publishing more AI-friendly text. Inside GEOforge, SignalForge weights a mention 3x a citation in Share of Voice.
🏢 Adobe acquired Rilo
Adobe acquired India-based Rilo in a licensing/team transaction. Terms undisclosed. Rilo built agentic workflows for competitor intelligence, content repurposing and distribution, and sales call analysis. More than 10,000 people had tried the product. The six-person team plus parts of its IP move into Adobe. Rilo itself shuts down. Follows Adobe's Semrush deal. Big platforms are assembling an AI marketing operating layer.
📊 Brainlabs: less volume, stronger intent
Brainlabs analyzed Google Analytics from 54 clients across 19 sectors. Organic sessions fell 10.5%. AI platform referrals rose 163%. AI-driven key events rose 335%, with AI referrals generating key events at 1.5x the organic search rate. Brainlabs associates the inflection with wider AI Overview adoption. The dataset does not establish causation. Sessions are a poorer proxy. Track visibility, recommendation, qualified visits, and pipeline.
🔀 SEO Pulse: engines do not pick the same vendor
SEO Pulse tested 1,500 commercial prompts. Different AI engines selected the exact same top vendor in under 1.5% of queries. More than 50% of answers mentioned brands. About 10% both mentioned and cited one. Consumer-style sessions, not API-only, covering Western and Asian engines. Do not blend share of voice across engines. SignalForge runs 40 times per prompt for under 2% margin of error.
🛠️ Yext Scout moved from tracking into actions
Yext expanded Scout from visibility tracking into brand-level GEO / AI search visibility and actions across listings, reviews, social, and its knowledge graph, plus Corvo AI for SMBs. Internal 147% visibility increase. One customer doubled inbound leads. Measurement and diagnosis are table stakes. Differentiation is execution. A dashboard-only GEO service has a short shelf life.
🗄️ Dun & Bradstreet is feeding Perplexity via MCP
D&B integrated its commercial graph and risk analytics with Perplexity Computer using Model Context Protocol, making verified corporate identity, relationships, and risk information directly accessible to AI workflows. GEO now includes licensed databases, APIs, and MCP connectors, not only websites. Feed brand knowledge directly rather than hoping a crawler formats it.
🛡️ Huskies raised $27M. A bad WAF can make you invisible
Blackstone Innovation Investments led a $27 million Series A in Huskies, valuing it above $100 million (Wall Street Journal). Bot management is intertwined with AI discoverability. Brands want legitimate AI retrieval while blocking malicious automation. A badly configured web application firewall can make a brand invisible. Cloudflare is part of the GEO stack. AI crawler accessibility plus bot security belongs in technical GEO audits.
📋 Kevin Indig: earn the recommendation from somebody else
Kevin Indig analyzed 3,000 ranking listicles in a 60,000-query set. Publishers accounted for 46.8% of highest-ranking listicles versus 19% for brands. AI Overviews appeared on 93.4% of B2B queries. Freshness showed the strongest stable association with high rankings. Putting your own product first did not independently explain performance. 50-best pages can still work in SEO. Self-serving comparison content is not a defensible GEO advantage. Combine freshness with third-party proof. Earn the recommendation from somebody else.
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: If 59.8% of the citations grounding a buying answer sit outside the top 100,000 domains, would you still call domain authority your GEO KPI?