Session library

AI GEO and the Future of Brand Discovery Insights from Ongoing Research

Sunny Youn · AI Marketers Guild · November 5, 2026

Sunny Youn presents data-driven research exploring how generative engines cite brands, evaluate user queries, and reshape digital search discovery.

geoai search

Sunny Youn presents data-driven research exploring how generative engines cite brands, evaluate user queries, and reshape digital search discovery. Hosted by David Berkowitz on the AI Insiders series from AI Marketers Guild, the session examines live API testing across five commercial categories—consumer electronics, sports merchandise, skincare, toys and games, and nonprofits—across ChatGPT, Google Gemini, Perplexity, and Anthropic Claude.

Youn breaks down critical Generative Engine Optimization (GEO) concepts, including background search fan out, the divide between keyword SEO and buyer story Answer Engine Optimization (AEO), and conversational dropout at price pivots. The discussion provides agencies, brands, and marketing practitioners with actionable methods to diagnose where brands vanish during multi-turn AI interactions and how to produce structured, scenario-specific content that wins AI engine citations.

Key moments

[06:09] What Is Search Fan Out in Generative Engine Optimization?

Sunny Youn defines fan out in Generative Engine Optimization (GEO) as AI query expansion, where a generative engine automatically runs multiple background web searches to answer a single user prompt. Rather than executing a direct, literal keyword lookup, the AI broad-matches the underlying intent of the question by generating several exploratory search queries before synthesizing a response.

Youn notes that conversational AI transforms the path to purchase into a multi-turn journey. Users begin by exploring broad topics, move into comparative evaluation, and finally reach a purchase decision. Because each follow-up turn in an AI dialogue triggers an entirely new set of background searches, brands risk experiencing mid-conversation "dropout," where their share of voice disappears at critical decision pivots.

Notable Quotes:

"fan out is query expansion. You ask one question, and then the engine runs several searches for you. So, it's really like a broad match." — Sunny Youn [06:09]

"conversation is the path to purchase because people discover, then compare, and then decide, and each new question is a new search." — Sunny Youn [06:19]

"And then the dropout is the share of voice lost at the pivot. So, it's the moment your brand stops being named as your conversation deepens." — Sunny Youn [06:43]

[07:35] What Core Metrics Measure How AI Engines Retrieve Information and Cite Brands?

Sunny Youn identifies three foundational metrics for benchmarking GEO performance: searches per answer (fan out depth), keyword versus story overlap, and run-to-run overlap across non-deterministic engine outputs. Searches per answer counts how many distinct background queries an engine triggers for a single prompt, which can range from zero to multiple queries depending on the engine and industry.

Keyword versus story overlap assesses whether an engine runs the same background searches when presented with a terse keyword phrase versus a detailed, scenario-based user question. Youn observed that this overlap ranged between 21% and 42% in her tests. A high overlap indicates that conventional SEO keyword rankings may carry over into AI citations, whereas a low overlap signals that brands must deploy targeted Answer Engine Optimization (AEO) that speaks directly to real-world buyer situations.

Notable Quotes:

"There's three numbers to watch. One is searches per answer. How many searches do the engines run behind one answer? That's the fan out." — Sunny Youn [07:36]

"The key point is that keyword versus story overlap, at least in this study, ran 21% to 42%." — Sunny Youn [08:03]

"The higher the overlap, the better you can just stand on SEO alone. The lower the overlap, the more there's a need to look at the detailed story from an AEO perspective." — Sunny Youn [08:19]

[10:48] How Do Perplexity, Gemini, and ChatGPT Differ in Live Search Depth?

Sunny Youn explains that AI engines display vastly different search behaviors: Perplexity conducted live searches on 99% of queries, Gemini on 64%, and ChatGPT on only 24%. Perplexity executed more than 10 times as many background searches per answer as ChatGPT. This disparity means ChatGPT relies primarily on pre-trained internal knowledge, making foundational earned media sources like Wikipedia, press coverage, and established product reviews critical for ChatGPT visibility.

However, Youn emphasizes that a higher volume of live searches does not automatically result in more brand citations. In testing sports merchandise queries, Gemini named a dedicated retailer 73% of the time—more than twice as often as Perplexity—despite Perplexity conducting significantly more searches. Youn cautions that practitioners must track and optimize performance engine by engine rather than treating generative AI as a single search surface.

Notable Quotes:

"Perplexity searched live on 99% of the answers and Gemini on 64%. ChatGPT recorded a search on only 24% and Perplexity ran more than 10 times as many searches per answer." — Sunny Youn [11:09]

"on about three in four ChatGPT answers, it's working from what it already knows. And that makes the sources it learns from, like press coverage, reviews, and Wikipedia matter more for ChatGPT." — Sunny Youn [11:25]

"Gemini named a sports merchandise retailer 73% of the time. And you know, [Claude], ChatGPT, and Perplexity less so. But actually, Gemini named the retailer twice as often as Perplexity, even though Perplexity had more searches per answer." — Sunny Youn [12:46]

[14:00] Why Does Keyword Versus Story Phrasing Impact Brand Citations in AI Engines?

Sunny Youn explains that traditional keyword optimization fails in conversational AI because contextual buyer prompts alter the actual search queries engines run behind the scenes. In consumer electronics testing, a basic keyword prompt like "headphones for work calls home" produced completely different AI search queries than a situational prompt describing six hours of daily video calls with poor microphone quality.

When user prompts contain specific use cases and pain points, the search terms generated by the AI deviate significantly from standard commercial keywords. Youn notes that content strategy must move beyond generic keyword stuffing to target Ideal Customer Profile (ICP) scenarios and detailed buyer constraints, ensuring an engine finds pages that explicitly solve the user's practical predicament.

Notable Quotes:

"The higher the keyword to story overlap, the more likely it is that SEO will also earn you the AI citation cuz the engine is searching for roughly the same words you already rank for." — Sunny Youn [14:33]

"The lower the overlap, the less your SEO work will move AI citation because the AI engine is searching for things your keyword pages don't cover." — Sunny Youn [14:44]

"it's the extra detail the buyer adds that changes the search, not the writing style." — Sunny Youn [16:36]

[16:53] How Does Brand Citation Concentration Vary Across Different Product Categories?

Sunny Youn reports that brand citation behavior in AI engines is heavily dictated by category dynamics, separating into concentrated, keyword-led markets versus fragmented, situation-led markets. Highly established sectors such as consumer electronics and sports merchandise averaged 1.5 searches per answer, and AI engines consistently concentrated citations among a few dominant market leaders who won on comparison pages and reviews.

Conversely, categories such as toys and games, skincare, and nonprofits showed lower search fan out and as little as 21% keyword-to-story overlap. In these fragmented sectors, consumer choices depend on narrow situations (such as an eight-year-old interested in dinosaurs, or sensitive skin tones seeking sunscreen without white cast). For these long-tail categories, Youn advises brands to write directly for specific buyer situations because situational details dictate which brands the AI engine retrieves.

Notable Quotes:

"Sports and electronics has the most searching one and a half times per answer. With toys and games, it's less than one. And there's only a 21% match." — Sunny Youn [17:00]

"Sports and electronics are keyword-led where reputation and comparison of content helps them stand out. Toys and skin care is the opposite." — Sunny Youn [17:46]

"The less the engine searches, the more the buyer's situation steers the search. Write for the situation and track both." — Sunny Youn [19:15]

[22:36] What Is the Price Pivot and Why Does It Cause Brands to Drop Out of AI Conversations?

Sunny Youn explains that the "price pivot" occurs when a consumer introduces cost sensitivity into a multi-turn conversation, causing premium market leaders to abruptly disappear from AI recommendations. In scripted tests asking for custom sports jerseys, the leading official retailer dominated initial responses, but dropped to zero mentions when the user asked for high-quality alternatives that were less expensive, with Etsy capturing 50% of citations instead.

A parallel shift occurred in consumer electronics: when users queried top-tier wireless earbuds, engines initially cited Apple, Sony, Bose, and Samsung. Once the user stated that those options were too expensive, Anker Soundcore completely displaced the premium incumbents on subsequent turns. Youn warns that brands fail to retain visibility if they do not publish content that addresses pricing objections and positions them favorably before budget-conscious pivots occur.

Notable Quotes:

"when price becomes part of the AI search, it drops to zero and doesn't even make the short list. Etsy appears in the price pivot 50% of the time." — Sunny Youn [23:39]

"Top sellers hold at the price pivot. If your Eagles or 49ers, you stay on high on the short list, but the smaller fan bases slip." — Sunny Youn [24:49]

"what's the best earbud? So, Apple, Sony, and Bose, and Samsung show up. But, then when you say it's too expensive, Soundcore Anker actually replaces the brand citation in the second and the third hop." — Sunny Youn [25:59]

[26:44] How Can Marketers Defend Brand Visibility Against Mid-Conversation AI Dropout?

Sunny Youn presents a four-step framework to prevent brand leakage in generative conversations: map the buyer journey, identify the specific dropout hop, build objection-handling content, and seed external citation sources. Marketers should document the sequential questions buyers ask—from initial discovery through budget objections and competitor comparisons—using real feedback from customer sales conversations.

Brands must determine which alternative brands replace them during conversational pivots and author dedicated on-site pages answering those exact queries, using the buyer's phrasing as page headings. Furthermore, because AI engines rely on external validation, brands must identify the third-party platforms engines cite during objections—including Reddit threads, buyer guides, and review portals—and ensure accurate, positive inclusion in those citation hubs.

Notable Quotes:

"Map the buyer's questions from discovery to final objection, including what your sales team hears. Find the step where you drop out and sort who replaces you" — Sunny Youn [27:36]

"write one page per objection, like too expensive or alternatives, and use the buyer's exact question as the heading." — Sunny Youn [27:54]

"look at the sources the engines cite at that step, like comparison lists, Reddit, review sites, and get listed there accurately." — Sunny Youn [28:24]

"continue to test and keep a holdout set of questions that you never write content for. And if only the questions you wrote for go up, then that's overfitting, and not a real win." — Sunny Youn [29:03]

[33:26] How Should Content Strategy and Website Ontology Adapt for Answer Engine Optimization?

Sunny Youn outlines that enterprise AEO requires restructuring a website's informational ontology around core concepts and the high-demand questions the brand wants to answer. While generative tools can assist in rapid content drafting, human strategists are essential for establishing editorial taste, institutional positioning, and topical architecture.

Youn stresses that content recency strongly influences generative AI engines. In live client experiments, newly published, highly specific answers achieved number-one placement in ChatGPT within one week, but citations decayed within two weeks when the content was left unrefreshed. Maintaining AI visibility requires ongoing updates and continuous iteration against real search demand signals.

Notable Quotes:

"start with prompt engineering, which is to center on which questions you want to be the answer for." — Sunny Youn [33:39]

"generate content that's refreshed on a regular basis because the LLMs do penalize recency. If you have stale content, then I'm working with a small business in the Midwest and she was number one response on chat GPT within a week for her category. But it did the numbers started to soften after 2 weeks." — Sunny Youn [34:28]

"that strategy of the ontology is human because it's strategic. The creation of it you know, is where AI can be efficient once you give it the right prompt." — Sunny Youn [37:52]

[39:41] Why Are Off-Site Sources and Video Transcripts Essential for Generative Engine Visibility?

Sunny Youn explains that GEO extends far beyond on-site optimization to include earned digital footprints across external platforms, describing the workflow as "PR in a box." Because large language models synthesize answers across the wider web, brand visibility depends heavily on presence across Reddit discussions, LinkedIn, YouTube transcripts, and Wikipedia entries.

Engines such as Google Gemini and Google AI Overviews actively process YouTube video transcripts to ground answers, making synchronized text captions essential for video discoverability. Additionally, maintaining entity consistency across digital assets—such as aligning website naming conventions with Google Business Profiles and using structured FAQ formatting—directly improves the ease with which AI models parse, comprehend, and cite brand data.

Notable Quotes:

"The tool that I built doesn't only do on-site optimization of your own site. We also have a training for AI sources. So, what should you put on YouTube, LinkedIn, TikTok? How should you comment on Reddit or which threads on Reddit are the ones that you need to do what I call syndicated content." — Sunny Youn [39:41]

"YouTube video transcripts are being captured by Gemini and AI overview. So, I add the transcripts now" — Sunny Youn [40:25]

"Everyone knows FAQs are the way that LLMs understand better, and there has to be consistency." — Sunny Youn [40:52]

[44:26] Why Does High Brand Awareness Fail to Protect Legacy Brands in AI Search Engines?

Sunny Youn demonstrates that widespread consumer brand awareness in traditional channels does not guarantee representation in AI answers. In comparative testing, Intuit's TurboTax secured 69% of AI citations for tax preparation software, but legacy antivirus and security giant Norton received 0% of citations in AI queries for password managers.

Instead of citing Norton, AI engines recommended alternatives such as Bitwarden, 1Password, and LastPass. Youn and David Berkowitz emphasize that AI-driven discovery forms new consumer habits, particularly among younger demographics who rely on conversational search. Brands that assume past market awareness guarantees AI discovery risk being completely excluded from answer engine recommendations unless they actively cultivate structured citations and differentiated authority.

Notable Quotes:

"for password managers, everyone knows about Norton. But, it was cited in AI search engine 0% of the time." — Sunny Youn [44:48]

"If a brand says, 'We can rest on our brand value recognition,' they're aging themselves out of the next generation of kids who are turning to AI search for their answers." — Sunny Youn [44:58]

"brand awareness does not carry into AI answers and any brand that is ignoring it is doing themselves a disservice." — Sunny Youn [46:03]