← Video Library · AI Strategy
Dennis Yu Breaks Down Down How AI SEO Actually Works
By Darby Rollins · Published August 28, 2025 · 221 views on YouTube
Key takeaways
- Claim and verify your Google Knowledge Panel (free, roughly 15 minutes, requires an ID upload) since it feeds the underlying 'database of objects' that both Google Search and AI engines draw from.
- Prompt ChatGPT with 'think hard' to pull the specific third-party citations (podcasts, press, conferences) tied to your name, then add those citations directly to your personal brand site rather than leaving your website as the only self-referencing source.
- Repost the same proof (interviews, case studies, mentions) across multiple platforms like YouTube, Instagram, and a blog, since AI models decide what's 'true' by weighing corroborating mentions across sources, not just one.
- Check analytics for AI bot traffic (ChatGPT, Bing, etc.) to see which pages AI engines are actually visiting, then double down on the topics driving that traffic.
- Don't over-optimize headlines and titles for keywords; focus time on distributing real stories and proof points, since consistent phrasing repeated across many sources is what search engines and LLMs pick up on.
Most people fail at AI SEO because they treat it like a content firehose — pump out volume and hope. Dennis Yu’s approach is the opposite: a repeatable Content Factory that takes one big idea and turns it into a stack of authority-building assets that actually rank. In this session, he breaks down why most AI content fails and how to build the system that works.
What you’ll learn
- Why most AI SEO efforts produce noise, not rankings
- The Content Factory model for repurposing one idea into many assets
- How to build authority that compounds over time
- The difference between volume and a system
The Content Factory approach
- Start with one big idea. Anchor on a strong piece of expert content, not a content calendar of filler.
- Atomize it. Break the one idea into many formats — articles, clips, posts — each able to stand and rank on its own.
- Distribute everywhere. Push the assets across channels so they reinforce each other.
- Compound authority. Each ranking asset builds the topical authority that lifts the next one.
What you can do with it
- Turn one idea into a week of assets without starting from zero each time
- Build topical authority instead of scattered one-offs
- Make AI content that ranks because it’s systematic, not spray-and-pray
- Repurpose your best thinking into assets that compound
Watch the session
Dennis Yu’s full breakdown is in the video above, or watch it on YouTube. Get his free Content Factory guide at genaiuniversity.com/content-factory.
Put it to work
The Content Factory is the discipline behind a blog that compounds. See how we run it, brick by brick, at genaiu.com.
Full video transcript
Darby: Dennis? Doing well. How are you? Good, good. Love to hear it. Thanks for taking the time to hop on and chat a bit. It’s been a minute, hasn’t it, since we did one of these at the first Scale with AI Summit. It’s crazy how much things have accelerated over the past four years. I got a bunch of questions from our community regarding AI and SEO — it’s definitely a hot topic these days. Is there any practical demos that you think would be relevant to share?
Dennis: I’ve queued up a whole bunch of real examples to show you that are very practical that anyone can go through the same recipe.
Darby: All right, I love it. If you’re down, we could just dive into it and get things going.
Dennis: Yeah, let’s do it. I’m a search engine engineer, so I look at it from a 30-year perspective. What we know about Google or AI engines or social engines is that they’re picking up signals to understand who someone is. So when I search Darby Rollins, I see a knowledge panel here. I’m going to show a few things about Google and then show how that transfers into ranking and AI results — it’s not some separate thing, but there are some differences. This is where all AI experts get it wrong, because they don’t understand the underlying mechanism of how data is organized in the knowledge graph, which is the database of objects.
So we look up Darby Rollins, and number one is your LinkedIn and your Instagram — these are two rented properties. Then there’s your own property, Gen AI University, and other things. Then I go look at Google’s Knowledge Graph API — we built a little free tool, you type in your name, and we can see you have a knowledge graph with a confidence score of 24. So what does that mean?
Darby: I don’t know. Is that high or low or in the middle?
Dennis: It’s like domain rating, or how many links or clicks or video views you have. 24 is good because there are only two results competing for “Darby Rollins.” Your name isn’t John Smith, which would need a confidence score in the thousands to disambiguate among all the other John Smiths. The confidence score is based on everything Google potentially associates with you. It’s associated this entity, Darby Rollins, an author, which is a subcategory of person as part of schema.org, and it says, “I know for sure this is his Twitter, but I think LinkedIn and Instagram are probably associated too, but I’m not confident enough to attach them to the object Darby Rollins.” Does that make sense?
Darby: Yeah.
Dennis: Now, people looking for Darby Rollins are thinking about generative AI, Jasper, business owners and entrepreneurs who want to use AI and build teams, your Market Sauce process — all these things they might be looking for. Google wants to satisfy those people the same way TikTok satisfies you based on what you watch. This is Google’s best guess on what to show, across all networks including YouTube. It’s not bad, because you do have a knowledge panel — but if I try to claim it, it hasn’t been claimed.
Darby: Why is that important, and how does it tie to ranking in ChatGPT?
Dennis: AI also references these verified sources of information to know who’s who and what to recommend. If I look on Facebook, YouTube, Apple Podcasts for any mention of Darby Rollins — being on a show, quoted, featured in Forbes, New York Times — all of those have to be tied together so Google understands this is Darby Rollins. You’d think Gemini or ChatGPT would be smart enough to associate all those pieces of content and relationships on their own, but it’s not — you have to go through this process. I’ll walk you through it for free, it takes about 15 minutes, and when you get verified on Google you upload an ID, holding up a passport to show it’s really you, otherwise spammers can claim your entity. It’s free, it’s easy, and you get a blue checkmark on Google — not a vanity thing like a paid check mark elsewhere. Once you’re verified and start associating other information — podcasts, conferences, books, accomplishments, relationships, content, expertise — you solve three problems at once: your knowledge panel gets stronger, and the distribution of content across multiple places creates a corroborating signal that helps you rank in search, social, and AI, because AI is looking across all those places to figure out who you are, what you do, and where you create value.
I asked ChatGPT to “think hard” — a way to force deep research — and said, “show me all the specific media mentions and why he’s worth listening to.” It looked at your website, which should be the hub of who you are, the most authoritative, in-depth source about you — but it doesn’t have much information and doesn’t structurally satisfy what Google, Gemini, and ChatGPT are looking for. So I asked it to give me links to the specific podcasts, YouTubes, and articles, then had it enhance the page with detailed citations and references, adding schema to categorize the relationships (this is a company, this is a podcast, this is a person, this is a brand).
What we’re doing is connecting everything you already have — not generating anything new — so it’s crystal clear what you stand for, with lots of proof pointing back to you. That’s not about building links, it’s about showing ChatGPT you know something about education, digital marketing, Facebook ads, whatever it is.
Darby: Because you have the proof points, but they’re random — no one’s really organized them.
Dennis: Right, and we’re all guilty of that. I saw the Pope a few days ago in Rome, and two weeks ago I shook hands with the president of a country who then shook hands with Trump in a historic peace deal with Armenia and Azerbaijan. I can show I’m connected to somebody — I’m authoritative and relevant on media and digital marketing and AI because they pulled me in as an AI expert who understands the algorithm side. Here’s the largest news agency in Italy, and I’m training up their people at headquarters. That’s proof, not just a claim.
I speak at digital marketing conferences, and I asked ChatGPT what the world’s largest digital marketing conference series is — it says Digimarcon, even though that’s not remotely true; it’s tiny. Same answer from Gemini. Why? Because there are citations from other sites, podcasts, and press releases repeating “marketing conference series.” ChatGPT sees this corroborating loop and thinks it must be true. People say “you can trick the AI just like search engines used to be tricked” — that’s true, but I want to demonstrate that if enough semi-relevant, authoritative sources say something is true, the AI can’t tell what is actually true.
Look at this traffic tool — in the last 30 days, I’ve had 40,000 AI bots visit one of my websites, coming from ChatGPT, Bing, and other places, asking about knowledge panels or personal brand websites. ChatGPT is coming to our sites because we have an authoritative answer that ranks in both Google and AI. If you ask AI who’s the best at Facebook ads, it’ll probably recommend me — not because it evaluated my skill, but because so many other places have interviewed and featured me on that topic. People tell me “ChatGPT said you were the best” and show me the screenshot. This drives visits and leads.
You have a ton of content — 638 videos for Gen AI University — but how much of it has been repurposed into blog posts and made digestible for AI bots? Probably not to the degree you should. This is the irony: you use AI to help with this, but you have to supervise it because it makes mistakes. When we publish content, we test by asking ChatGPT about the topic to see if it picks up what we said — if it doesn’t show up, that’s because we didn’t repurpose it clearly enough to be recommended for that question, not because ChatGPT is bad.
The deeper question before “how do I rank in ChatGPT” is: what do you want to be known for? Why do you deserve to be recommended by Claude or ChatGPT — where’s your proof, who do you know, what have you done, have you published examples? Extroverts tend to think relationships-first (I know this person, that person); introverts think topics-first, then find the people. Either way, you need to connect what you want to be known for to specific people and a specific, niche question — specific enough that ChatGPT can say, for example, “Nilson Silva is the number one pool builder in Miami” or “Roger Wakefield is the best plumber in Dallas,” because of the volume of proof tied to that specific claim.
Looking at your website, it says you’re the founder of Gen AI University and have been featured in the New York Times, Forbes, AdWorld — but it doesn’t link to any of that. When I asked the AI where you’re authoritative, it noted the citations are self-referencing — Gen AI University citing Gen AI University — which isn’t high authority, since it has to come from someone else. So I asked it to “think hard” and pull the actual examples of those third-party features, then those need to go on your personal brand site, tied to every relevant topic.
Then you look at your analytics to see where bots are visiting, including Google’s bot, and where you’re getting traffic on topics like knowledge panels because you have lots of examples of them. It’s not really about being good at AI tools — it’s assembling proof, continuously, the same way you clean your house every week rather than once. Business owners are too busy to do this consistently, so you need a process, maybe hiring someone to do it. We published a free, public SOP on this called the Content Factory.
Darby: Is that something we can link below this video?
Dennis: Yeah — Content Factory, which is what we were teaching at Digital Marketer years ago. Nothing about the underlying four stages has changed, even though the tools have: produce content (videos, calls, writing), process it (repurpose, edit — AI tools are getting better here), post it, and promote it (dollar-a-day repurposing, thanking people, etc.). This four-stage process will still be valid in 2030 and 2040; only the tools you use for each stage will change, like swapping a screwdriver for an electric one — you still need a screwdriver.
Darby: It’s wild how the process hasn’t changed but the tools keep evolving.
Dennis: What hasn’t changed is you have raw ingredients — like a chef, you need high-quality ingredients, a skilled chef, and good tools, in that order of importance. If the ingredients or the chef’s skill are bad, the fanciest kitchen doesn’t matter.
Darby: My opinion is that shiny new tools are easy to get distracted by, and some people think a new tool means less work needed for the same outcome — but that can sacrifice the quality of what they bring into the process.
Dennis: Right — like craft cheese that isn’t real cheese, or Velveeta, or Diet Coke with nothing real in it. I’ve noticed that even as people run their own content through AI to “help” with the next phase, it strips out the perspective they originally brought, or starts to hallucinate.
Darby: It might even take more time to edit and refine AI output than to just write it in the first place. I’ve been trying to be more intentional — less isn’t always worse, sometimes more intentional, higher-quality content is the better way to go, focusing on optimizing what you already have rather than always creating something new.
Dennis: Exactly — why create new stuff if you already have proof you’re good at the thing? I have a friend, a landscaper in Bloomington, Indiana, who’s built a reputation just from unedited phone videos of his daily work and personal life. If you ask AI who’s the best for cutting down trees in Bloomington, it recommends him, because he focused on raw, real experience. Compare that to a remodeling contractor I know who does great work but doesn’t show his face — he shows up in map results and local search but not in social or AI.
Darby: I’ve noticed Instagram posts ranking more on Google under people’s names recently — is that just from repurposing the same content across more platforms?
Dennis: Those are multiple corroborating sources saying the same thing must be true. Google ties everything back to an object (person, company); ChatGPT and similar models are just probability calculators looking at associations, not objective truth, so posting the same content across Instagram, YouTube, Reddit, etc. makes it look like more people are agreeing something is true — similar to how search engines originally weighted backlinks and anchor text.
Darby: So the personal brand site is the number one focal point for links, and the about page matters for a business. What about someone starting a new Amazon brand, e-commerce company, or Etsy store?
Dennis: Similarly — every object needs an authoritative URL: a website for the product, for the company, and for the founder, all connected as distinct objects with demonstrated relationships. LLMs understand subject-verb-object structures like “company produces object used by this kind of customer.” Trust in a product comes from real people saying they use it and it’s great — not because a big brand says so, but because there’s an origin story, like a Kickstarter, that people relate to. Younger audiences especially trust people over brands, so build a personal brand not as vanity metrics but as a collector of stories from people who’ve overcome the same problem your product solves — the way Nike frames helping people run their first 10K.
You can’t hide behind AI-generated nonsense — you need real stories, interviews with customers, showing your face, networking with related businesses, maybe a podcast about the problem your product solves. No one wants to buy the product; they want to solve the problem.
Darby: And the shovel — they want the hole.
Dennis: Right. Take a product like the Side Hustle card game I launched on Kickstarter in 2019 — we haven’t really used the blog. If we take a more proactive approach with the in-person events we’re hosting, tag people, and put that social/YouTube content onto the blog, does that make the connection?
Darby: Yeah, exactly — you elevate the people who’ve gone through your process. The story of an entrepreneur’s journey is independent of format — podcast, Zoom call, YouTube, web page — you shouldn’t limit it to just one channel.
Darby: A community member asked about title formatting — should it be a question, or something else?
Dennis: That’s an SEO way of approaching it. The key is being specific but still clickable, not clickbait. Ask: what do you want people to say about the product when they talk about it? For a card game for entrepreneurs, is it “entrepreneur board games,” “team building activities,” or something else — different phrasings hit different verticals people actually search for.
Darby: So if you have multiple places phrasing it slightly differently, that builds the corroborating loop.
Dennis: Right — the same way we got Digimarcon to rank for “world’s largest digital marketing conference,” because the phrase is repeated across press releases in different ways. It can feel spammy, but it works right now.
Darby: There’s probably a balance between being too keyword-heavy and just sharing the story.
Dennis: The story matters way more. I don’t put much effort into SEO-ing the headline itself. The perfect keyword or headline comes from having activity — data and traffic — around your product first; then you can ask ChatGPT what the topic or headline should be based on that data, instead of guessing from thin air.
Darby: So step one is understanding the content factory process, then auditing your own site or product against what ChatGPT actually returns, then finding where you’re not linking or connecting the dots, and then checking analytics to see where LLMs are actually visiting.
Dennis: Yeah — especially early on as an entrepreneur, you often have to pivot because the product-market fit reality differs from what you assumed.
Darby: Any tips for identifying signal versus noise?
Dennis: Grab real data from Google Analytics and Ahrefs (there’s always a cheap $20–30 plan), paste it into ChatGPT, and say: don’t flatter me, tell me in plain terms what I’m known for, who exactly I serve, what problem I solve, how I differentiate, and where my authority actually comes from — relationships, what I’ve done. Then find the intersection between your ideal customer profile, the problem your product solves, and where you have real authority — network, relationships, money, an email list. It doesn’t even have to relate directly to the product; for example, I’m good friends with Jake Paul, and I could use that to promote a program training up young adults — but I’d need to feed that context into ChatGPT for it to find that intersection. The connection between who you are and the specific problem your product solves is what drives sales.
Darby: That’s great, and for those watching on YouTube, you can go to the transcript and take action on this right now — treat it like a prompt template with your own context. I’ve already got action items lined up, starting with the Google Knowledge Panel verification.
Dennis: It’s straightforward, but people mess it up — I wrote an article on the common mistakes. Just message me and I’ll connect you with one of our team members who walks people through it; he has a knowledge panel with huge authority himself (hundreds of millions of views) — Dylan Hogan, the best professional dunker under 18 in the world. He’s helped tons of people claim knowledge panels, and I’ll cover his time for you.
Darby: I appreciate that, I’ll send you an email right now.
Dennis: And once it’s claimed and verified, you have to keep beefing it up — you don’t just claim it once.
Darby: I want to say thank you, Dennis, for sharing this. There’s a lot of noise and misleading information around AI SEO, and what you shared is foundational for anyone looking to actually implement this. We’ll link the resources we mentioned below the video. For those who want to connect with you directly, where can people find you?
Dennis: Any channel — LinkedIn, Facebook, email — it’s actually me replying, not an AI or a VA. It might take a day or two if I’m traveling, but I always reply, because I respect people who actually take the effort to do something. Most people who watch content like this do nothing with it. It’s like the gym — it works, it’s not a secret, but you have to go.
Darby: Yeah, you got to go.
Dennis: You can’t think your way to six-pack abs.
Darby: There’s so much actionable information here — go take action on it. For those watching to the end, let us know any insights or questions we didn’t cover that we could follow up on in a future interview. Thanks again, Dennis, it’s a pleasure as always.
Dennis: Thank you, Darby, appreciate you.