YouTube is the second most cited social platform in AI-generated answers. Not a blog. Not a press release. Not a brand website optimized for every keyword in the book. A video platform.
That finding, from a large-scale study published this week by AI search monitoring platform OtterlyAI, should stop every agency owner mid-sentence during their next strategy call. Because it confirms what’s been building quietly for the past 18 months: the rules for getting found have fundamentally changed, and a lot of agencies are still playing by the old ones.
This is not a story about YouTube SEO. This is a story about what AI models actually trust when they synthesize answers, and what it means for how you structure your clients’ content, authority, and brand signals going forward.
What the OtterlyAI Study Actually Found
OtterlyAI analyzed AI citations across major answer engines, tracking which sources, domains, and content formats were most frequently cited in AI-generated responses. Their data spans millions of queries across tools including ChatGPT, Perplexity, Gemini, and others.
The headline: among social platforms, YouTube ranked second only to Reddit for AI citation frequency. LinkedIn came in behind it. Twitter/X barely registered. Facebook and Instagram were largely invisible.
That ranking should reframe how you think about content investment. Reddit and YouTube are both platforms built around depth, specificity, and community authority. Reddit surfaces because it is full of direct, firsthand answers to specific questions. YouTube surfaces because video transcripts, structured descriptions, and contextual authority signals give AI models something to extract and summarize.
The brands and agencies that built video content with intent, clear spoken explanations, structured scripts, keyword-rich descriptions, and chapter markers, are now getting cited in AI answers. The brands that built 30-second brand awareness reels optimized for likes are not.
The same study noted that major brands including some of the most SEO-dominant names in their categories are losing AI citation share. Not because their websites are poorly optimized for Google. Because their content is not structured in a way that AI models can extract clean, authoritative answers from.
Why YouTube Outranks Most Brand Content for AI Citations
AI models are not ranking pages. They are evaluating whether a piece of content can reliably answer a question. That distinction matters more than most agencies realize.
When a user asks an AI tool “what is the best way to structure a local SEO campaign,” the model does not return the site with the highest domain authority or the most backlinks. It returns the source that most directly, clearly, and completely answers that question. YouTube videos, particularly those with spoken explanations and detailed descriptions, score well on all three counts.
Here is why video specifically clears the bar:
Spoken content indexes like dense text. A 10-minute explanatory video, properly transcribed, contains thousands of words of contextual, conversational content. AI models process that content the same way they process a long-form blog post, but the video often has a cleaner question-and-answer structure because it was designed to be explained verbally.
Chapter markers act like headers. When a creator uses YouTube chapters, those timestamps signal the structure of the content to both search engines and AI extraction systems. It is the video equivalent of H2 tags. A video titled “How to Run a Google Ads Campaign” with chapters for “Setting Up Campaigns,” “Keyword Match Types,” and “Optimizing Bids” gives AI models a table of contents for the answer.
Description fields carry authority signals. A well-written YouTube description with entity mentions, clear topic framing, and links to supporting content gives AI models additional context for what the video is about and who is publishing it.
The agencies getting cited are treating YouTube as a knowledge asset, not a social media channel. That is the reframe.
The Brands Losing Ground in AI Search
The OtterlyAI study pointed to a pattern that should be uncomfortable reading for big brand SEO teams. Several companies with strong traditional SEO authority, the kind of brands that dominate first-page rankings, are not maintaining that authority in AI-generated answers.
This is the core GEO gap. Generative Engine Optimization is not an extension of SEO. It is a different discipline. The signals that make you rank in a ten-blue-links result, domain authority, backlink profiles, keyword density, and click-through rate signals, are only partially relevant to AI citation. GEO requires different inputs.
What AI citation actually rewards:
Specificity over breadth. A focused article that directly answers one question in full detail outperforms a comprehensive guide that covers ten questions shallowly. AI models extract the clearest answer, and a piece that hedges, qualifies, and elaborates around a topic without ever landing cleanly tends to get skipped.
Named entities and credentialed sources. AI models weight content higher when it references identifiable experts, research studies, or institutions by name. A post that cites “a recent study from Harvard Medical School” is more likely to be pulled into an AI answer than one that says “experts say.”
Conversational structure. AEO, answer engine optimization, has always pushed toward FAQ-style content and direct question-answer formatting. That principle holds here. Content structured around the questions people actually ask, with direct answers in the first one or two sentences, is what AI extraction systems are built to find.
Schema markup. FAQPage, HowTo, Article, and Speakable schema are not legacy SEO tricks. They are signals that AI crawlers use to identify structured, answerable content. Brands and agencies that have not implemented schema across client content are handing citations to competitors who have.
What Agencies Need to Do Right Now
The OtterlyAI data does not just describe a trend. It gives agencies a very specific to-do list. Here is what translating that research into client strategy actually looks like.
Audit Your Clients’ Citation Footprint
Before you can improve citation share, you need to know where you stand. Run your clients’ brand names and core service keywords through ChatGPT, Perplexity, and Gemini. Note which questions surface AI answers at all, which competitors are cited when they are not, and which content formats are being pulled.
This is the new competitive analysis. You are not just benchmarking rankings anymore. You are benchmarking whose content AI models trust.
The AI Search Optimizer at Emarketed can surface how your current content is positioned for AI extraction, which is a useful starting point for this audit.
Build Video Into the Content Strategy
If your clients do not have a YouTube presence, the OtterlyAI data is your pitch deck. Not every client needs a full video production operation. What they need is consistent, substantive video content where someone explains the core questions their customers are asking.
For a healthcare facility: explainer videos on treatment approaches, what patients can expect from intake, and FAQ responses from clinicians. These are exactly the kind of credentialed, specific answers that AI models cite.
For a B2B agency client: thought leadership videos on service categories, answered clearly, with proper descriptions and chapters. Ten minutes of clean spoken content on “how to choose a marketing agency” does more for AI citation than a blog post covering the same topic in generic terms.
Restructure Existing Content for Extraction
Not everything needs to be rebuilt from scratch. Most agencies have clients sitting on substantial content libraries that are not optimized for AI extraction. The fix is often structural rather than substantive.
Go through the highest-value pages and do three things: add FAQ sections that directly answer the questions people ask about that topic, implement FAQPage schema, and rewrite introductions so they lead with the direct answer rather than burying it after three paragraphs of context.
That pattern maps directly to the GEO and AEO principles covered here on the blog, and it pays dividends across Google’s AI Overviews as well as third-party AI tools.
Make YouTube a Citation Channel, Not a Branding Channel
The mental model shift: your YouTube channel is not for brand awareness. It is for being the cited source when someone asks an AI tool a question in your category.
That means optimizing every video for extraction, not engagement. Lead with the direct answer. Structure scripts around specific questions. Write descriptions that explain exactly what the video covers, in full sentences, not hashtag chains. Use chapters for any video over five minutes.
The brands that get this right in the next 12 months will build AI citation authority that compounds the same way backlink authority used to.
FAQ
What is GEO and how is it different from SEO? GEO, or Generative Engine Optimization, is the practice of structuring content to be cited in AI-generated answers. Unlike traditional SEO, which focuses on ranking in keyword-based search results, GEO targets the extraction and synthesis layer of AI tools like ChatGPT, Perplexity, and Gemini. The signals that drive GEO success include content specificity, named entity authority, conversational structure, and schema markup.
Why is YouTube ranked so high for AI citations? Video content, when properly structured with transcripts, clear descriptions, and chapter markers, gives AI models high-quality, extractable content that reads like dense, well-organized text. YouTube also benefits from its position as a broadly trusted, high-authority domain. The combination of transcribed spoken content and structured metadata makes it one of the most citation-friendly formats in the current AI search landscape.
Are traditional SEO rankings still relevant in 2026? Yes, but their relationship to traffic and visibility is changing. AI Overviews and third-party AI tools now intercept a significant percentage of queries before the user ever reaches traditional organic results. Brands that rank well but are not cited in AI answers are seeing meaningful traffic erosion. The most effective approach combines both: optimizing for rankings in traditional search while building separately for AI citation.
What types of content perform best for AI citations? Direct-answer content structured around specific questions, FAQPage and HowTo schema, long-form video content with transcripts and chapter markers, content authored by or citing named credentialed experts, and regularly updated content that reflects current information. Thin, broadly written content that covers a topic without answering any single question clearly tends to perform poorly.
How can my agency measure AI citation share? Tools like OtterlyAI, Semrush’s AI tracking features, and manual sampling through major AI tools are the current options. The practice of “AI SERP monitoring” is still maturing, but regular manual auditing of how your brand and clients appear in AI answers is a meaningful starting point. Tracking citation share alongside traditional rankings gives a more complete picture of search visibility in 2026.
Does this strategy apply to healthcare marketing specifically? Yes, and it is arguably more important in healthcare than in most industries. Patients increasingly use AI tools to research symptoms, treatment options, and care providers. Healthcare facilities that have structured, credentialed, specific content answering the questions patients are asking are positioned to show up in those answers. Given the trust and sensitivity around healthcare decisions, being the cited source carries significant weight.
The Window Is Open, But Not Indefinitely
The brands and agencies that built strong SEO authority in 2012 to 2015 compounded that advantage for years. The same dynamic is playing out now with AI citation authority. The citations are being distributed, the precedents are being set, and the AI models are learning which sources to trust.
The OtterlyAI study is not just a data point about YouTube. It is a signal that the content formats, structures, and platforms that AI models trust are not the same as the ones traditional SEO rewarded. The agencies that internalize that difference and build GEO strategy into their service offering now are the ones that will be explaining it to prospects as a case study two years from now.
If you want to start measuring where your clients stand in AI search today, the AI Search Optimizer is a practical first step.
