Dennis Yu on Real Signals, Reputation, and Why AEO and GEO Are Just SEO
Session identifier
- Session: Real Signals, Reputation, and Why AEO and GEO Are Just SEO
- Day: 2 (June 25, 2026)
- Speaker: Dennis Yu, CEO, Local Service Spotlight and the AI Builder Program
- Framework concept: Operator Convergence — Paid and Organic Collapse (operator side), coined by Jason Barnard in 2025
- SEL article: How AI Is Merging Paid and Organic Visibility — https://searchengineland.com/ai-paid-organic-visibility-480229
Speaker bio
Dennis Yu has spent 25 years building search and social at scale, from running engineering at Yahoo to scaling agencies across home services and other local verticals where the ad spend tells you what is actually working before the strategy decks catch up. He now partners with universities like DSDT and Johns Hopkins to train students and military veterans on the AI and digital marketing micro skills that turn into paying clients in weeks, not years. On the AEO and GEO question, his take is simple. The labels keep changing but the job is the same, which is generating real signals and real reputation that Google and the LLMs cannot ignore
Session’s place in the Kalicube Framework
Dennis’s session covers the operator side of the Paid and Organic Collapse — the convergence of paid and organic visibility across AI surfaces from the ground-level vantage of an operator who has been reading signal data since Yahoo’s early engineering days. Dennis covers it from the operator’s side: what it actually looks like when real signals, real reputation, and real user behavior determine whether AI engines recommend a brand. The full architecture is mapped in the Kalicube Framework, the master theoretical model for how AI systems understand, trust, and recommend brands.
Episode summary
Dennis Yu spent twenty-five years at the signal layer — running engineering at Yahoo when the same search capabilities were being built at Google and released within weeks of each other, then scaling agencies across home services and local verticals where ad spend tells the truth before any strategy deck does. His session at Kalicube Summit 2026 opened with a direct dismantling of the label debate: Demis Hassabis and Dario Amodei have publicly agreed that the LLMs all do roughly the same thing and release capabilities at roughly the same time. Dennis confirmed it from the inside: the engineering culture across Google, Bing, OpenAI, and Anthropic is functionally one ecosystem releasing the same capability set in different wrappers. AEO, GEO, and SEO are the same job. The mechanism the engines use to decide who to recommend is the same mechanism it has always been. What has changed is the speed at which authentic signals compound and the speed at which artificial ones get caught.
The session’s most operationally specific contribution was the Sal Sciorta case — a plumber in eastern Pennsylvania who rebranded with a Superman cape and wrapped vans, and became a local micro-celebrity. Dennis’s methodology runs paid amplification against that authenticity: Facebook ads, YouTube ads, X remarketing, all targeted at the same eastern Pennsylvania audience that already recognises Sal. The engagement rates are extraordinary because the relevance is extraordinary. The AI engines read those engagement signals — real time on site, ninety-percent video completion, quality comments from people in the actual catchment — and amplify the brand to the next prospect. The Untrained Salesforce, coined by Jason Barnard in 2023 to name the AI assistants recommending brands around the clock without training or instructions, is already selling, right now, for every brand in the market. Dennis’s framing is direct: the brands that have done the authentic-signal work have the salesforce on their side. The brands that have not are operating blind. His worked examples on signal gaming — the Rome restaurant at 30,000 reviews that served terrible food, the gamified Google Maps traffic companies sending real humans on phantom navigation journeys — are the session’s sharpest evidence that the signal layer is the competition, not the content layer. Google’s own DOJ testimony confirmed it: user behaviour signals carry more weight than on-page content.
The personalisation argument closed the session with the most consequential observation about where the market is heading. ChatGPT has accumulated more behavioural intelligence about its users in the last year than fourteen years of personalised search produced. Memory across sessions is now standard. The result is a recommendation surface that rewards reputation more aggressively because it has more signal to grade reputation against. The Seoul Korean barbecue anecdote — Grok recommending a fourth-floor restaurant in a back alley that turned out to be the best Dennis had eaten — is the operator’s evidence that location, location, location has been replaced by signal, signal, signal. Dennis’s closing prediction is sharp: the brands with genuine reputations will do disproportionately better in the AI era. The brands without will be displaced harder and faster than they were under classical search. The very best get pushed up, and everyone else is going to hate AI.
Key quotes
“The stuff you do to rank in Bing is roughly the same stuff you do to rank in Google. But there are a lot of people who want to confuse you and sell services against that. That’s the whole AEO, GEO, E-I-E-I-O thing.” Dennis Yu, on why the label debate is the wrong conversation: the mechanism has not changed, only the marketing around it”
“Real humans, real engagement, real transactions, real time on site. All the hallmarks of a real user.” Dennis Yu, naming the signal substrate AI engines actually read when deciding whether to recommend a brand”
“How is that different, whether you’re trying to please that user or the assistant’s user or the user’s assistant?” Dennis Yu, on the Untrained Salesforce as continuity rather than disruption: the agent is acting on behalf of a human, so the job of earning the recommendation has not changed”
“Google revealed that user behavior signals are actually more important than the content on the page itself.” Dennis Yu, citing the DOJ disclosure that ratified what operators have known for years: signals beat content, every time”
“The very best get pushed up, and everyone else is going to hate AI.” Dennis Yu, closing on the bifurcation the AI era is engineering: reputation compounds harder and faster, and the brands without it are displaced harder and faster”
Further reading
How AI Is Merging Paid and Organic Visibility, part of Jason Barnard at Kalicube’s Search Engine Land series on AI search architecture. The article maps the paid and organic convergence Dennis covers in this session — why the paid and organic boundary is dissolving across AI surfaces, and what that means for how brands think about signal generation across channels.
https://searchengineland.com/ai-paid-organic-visibility-480229
Related session
- Lily Ray on E-E-A-T, N-E-E-A-T-T, and the Credibility Layer of UCD (Session 1.15): the upstream credibility-building work that creates the reputation Dennis’s signal-amplification methodology then compounds across surfaces
Link to preview post
This session was previewed here before the Summit.
Link to speaker post
Links to session clips
The Wagon of 100 Phones That Fooled Google Maps: Dennis Yu on Signals vs Reality
Why AI Agents Selling for You Is Nothing New: Dennis Yu Explains
Transcript for Dennis Yu on Real Signals, Reputation, and Why AEO and GEO Are Just SEO
Jason Barnard: Hi, and we’re continuing the Summit with Dennis. Dennis, welcome. Lovely to see you again.
Dennis Yu: Always good to see you, Jason. I need more red shirts.
Jason Barnard: Yeah. I actually had these made last week in Singapore — new red shirts, tailor-made. The old ones I had made in Thailand seven years ago were beginning to look a bit ropey, so I got some new ones. If anybody’s watching this and thinks I look much smarter than I used to, it’s true.
Dennis Yu: You do.
Jason Barnard: Thank you very much. In both senses of the term — intelligence and dress.
So we’re looking at AEO, GEO, and you’re saying it’s just SEO in new clothes. I like this approach. I coined Answer Engine Optimization back in 2017, and now I’m talking about Assistive Agent Optimization. I’ve just been speaking at Google Marketing Live, where they asked me to explain the Kalicube Process to their enterprise clients, so it looks like I’m on the right track.
What I’m arguing is that Google and Microsoft are essentially the same architecture. I interview a lot of Microsoft engineers, so I’ve got a lot of insight from that. What do you say about the idea that Microsoft and Google are more or less the same thing, and that ChatGPT and Perplexity are joining that as a system?
Dennis Yu: You know, it’s funny — I saw an interview with Demis Hassabis, who runs DeepMind, and some of the other folks from Claude and elsewhere. They all agreed. In fact, there was one with Dario Amodei and Demis Hassabis together, and they both agreed that all the LLMs basically do the same thing and release the same capabilities at the same time. I knew this was true almost thirty years ago, because the stuff we built at Yahoo was the same stuff being built at Google. Then Google would release something, and we would release something right after it.
One of the search engineers I used to run with almost every day, Peter Elahagi, implemented instant type-ahead search results. You know, when you type something and it auto-completes. There was a counterpart at Google who implemented the same thing. So why wouldn’t it be the same today? The stuff you do to rank in Bing is roughly the same stuff you do to rank in Google. But there are a lot of people who want to confuse you and sell services against that. That’s the whole AEO, GEO, E-I-E-I-O thing.
Jason Barnard: Old MacDonald had a farm. What I say is it’s an Algorithmic Trinity: you have search results, you have an LLM, and you have a knowledge graph. The idea that those represent intelligence, information, and validation makes a lot of sense. We’re going down the same route for all of these engines, and as you say, they all do more or less the same thing. They’re serving the same users. They have more or less the same technology. They’re not going to reinvent the wheel.
Dennis Yu: The stuff you’ve been teaching for years — on how entity signals and trust and all these different markers show that you actually do the things you say you do — ties back to knowledge panels, confidence scores, all of those things. That foundation is always going to be the same. There are just other layers stacked on top. A different way of organizing the same information. Now, because those other layers are on top, people are trying to call it something else.
Jason Barnard: And it’s actually one holistic approach. In my world, that’s Assistive Agent Optimization, because we’re now optimizing for agents, assistants, and search together. I’ve seen that at BlitzMetrics you’ve got a brilliant system for entity management.
Dennis Yu: We’re doing something very similar to what you’re doing, because we’re gathering real signals. We’re not generating stuff, not spinning up blog posts full of nonsense. I think the only slight difference between what we teach and what you teach is that you’re way deeper on the SEO components and the technical mechanics of how things tie together. You have these corroborating loops of proof that you are who you say you are. What we do is say: if the proof is there, why don’t we amplify it? Why don’t we run some ads against it? Not to drive direct leads, but to drive such high engagement that YouTube sees people actually staying to watch ninety percent of a three-minute video. The kind of engagement Facebook sees where the comments are quality comments, from people who are actually in that audience. Not people from some random country, not bots. Real signals.
So what we teach is this: if you actually do what you say you do — say you’re a plumber in eastern Pennsylvania — then we’re going to find other people in eastern Pennsylvania who recognize you. Take Sal Sciorta. He rebranded. He’s got a Superman cape, his vans drive all around town with the Superman thing. He’s like a mini celebrity in his area. If we show pictures of Sal Sciorta going around town in his Superman suit to people in that town, the relevance is already so high that when we run Facebook ads and YouTube ads, boost it on Twitter, and remarket across those channels, the engagement rate tells Google, tells the LLMs, tells X — whatever it is — that this content is legitimate and relevant to that audience. Real humans, real engagement, real transactions, real time on site. All the hallmarks of a real user.
Jason Barnard: Absolutely, one hundred percent. I’ve been saying your Brand SERP is Google’s opinion of the world’s opinion of you, and the AI Résumé is now AI’s opinion of the world’s opinion of you. If you’re walking the walk — as you just described — its opinion of you is going to be: this person engages with their audience, their audience loves them. They’re trying to mirror the world. I call it the mirror principle. They’re trying to mirror human behavior. Walk the walk and you’re going to win.
My next question is the Untrained Salesforce. Have you heard me talk about that at all?
Dennis Yu: Briefly.
Jason Barnard: The idea that AI systems are selling for somebody twenty-four hours a day, seven days a week — so why not train them to sell for you? Is that just SEO in new clothes?
Dennis Yu: You just have another audience you’re training. The underlying proof and the ingredients are the same. Think of it like food. You can take that same pound of meat and turn it into cheeseburgers, turn it into a grilled steak, turn it into a beef stew. Many different preparations from the same raw ingredients. So now that there are assistive agents shopping and recommending on our behalf, it’s like I have an executive secretary and you have an executive secretary, and they talk to each other. If our services are good, if we’ve got a great reputation, if we’ve done all the things that show we deserve to be recommended — isn’t that the same thing we’ve been saying all along for Google? What Matt Cutts would say: actually do the stuff that makes it so you deserve to rank number one for that query. You actually have the best answer. You actually have the best reputation. How is that different, whether you’re trying to please that user or the assistant’s user or the user’s assistant?
Jason Barnard: And the assistant is the lens between you and the potential client on the other side.
Dennis Yu: The last fifteen years I’ve been using a service called Fancy Hands — an assistant service where they book things for you, call and sing happy birthday, that kind of thing. People who need an executive secretary to handle all sorts of tasks. I’ve done the same kind of thing where I deputize someone to go do something on my behalf. I have a housekeeper, and she goes shopping at Costco, picks up friends from the airport, runs errands like going to the post office. She is fully entrusted with my credit card to do all these things on my behalf. Does that mean that Costco, the post office, or the restaurants where she’s picking up food have to somehow adjust? Maybe a little. But if they deliver great service, she’s going to come back. She’s acting on my behalf. I don’t see a significant difference between asking Claude or Grok or any other agent to do something on my behalf versus a well-trained human assistant.
Jason Barnard: Right. They are working on our behalf. I spend my life — and I know you’re in the same rabbit hole — talking to my AI assistant. I spend too much time talking to Claude.
Dennis Yu: I literally have, if you look at my WhisperFlow, something like 120,000 words spoken in the last two and a half weeks.
Jason Barnard: I was about to say that. I’ve just started using WhisperFlow, and now it really feels like I’m talking to Claude. It’s getting quite compelling.
Dennis Yu: Yeah.
Jason Barnard: This is a new demand that I’m using Claude in ways I never had before. So with AI assistive engines and AI assistive agents — is supply creating its own demand? Is this simply adding on top of search?
Dennis Yu: It depends on the category. If you have a toothache or your toilet is overflowing and you need an emergency plumber, that demand is determined by how many times toilets overflow and how many times people crack a tooth. You can’t create more of that kind of emergency demand. But what I do see is that because it’s easier to make recommendations — just like social media; I learned magic from TikTok videos and actually bought the items, which is wild — in that case it’s generating demand, because the more I consume, the more opportunity there is to buy.
For example, I was in China recently, and I’d seen other people go to the Avatar Mountains. That’s actually a real place in China. I was doing research through social media, and I was also talking to Grok and Claude, saying I’m going to be at Mount Huashan, Tianmen Mountain, and Zhangjiajie. What else should I do? Claude said: if you’re in Zhangjiajie, do the Yellow Dragon Cave, eat at this restaurant because that area is known for a specific triple meat dish unique to that village, and there’s this Chinese acrobatic show — like Cirque du Soleil but in China — that has forty thousand reviews and you absolutely need to see it. Initially I had one intent: hike the Avatar Mountains. But if it wasn’t for the AI recommending those other things, I wouldn’t have done the Yellow Dragon Cave, I wouldn’t have gone to the acrobatics, and I wouldn’t have eaten at that restaurant. Is that generative demand? I think it’s like an upsell. You came in for one thing, and the assistive agent said: would you like fries with that?
Jason Barnard: That’s a really good way of putting it. And there are two things going on: the upsell you just described, and the shifting of demand from one restaurant to another. The restaurant you would have eaten at has just lost the business because it wasn’t recommended by the AI.
Dennis Yu: Real estate used to run on location, location, location. It cost more to be in the city center because that’s where the tourists were, where the foot traffic was. Anyone a couple of streets away had less demand. But now — I was in Seoul recently, doing the typical tourist things, and I said: I want the very best Korean barbecue. Instead of going to the main street with all the tourist trap restaurants, I asked Grok, because I trust it for this kind of thing. It said: the place you need is in this back alley, up on the fourth floor. Dan and I walked through a dark alley thinking this seems sketchy, maybe someone gamed the agent. We walked up through what looked like some kind of industrial space. And there on the fourth floor was this restaurant. It was amazing. You would never have known it existed. The only way you could have found it was through digital means.
I think this creates a real equalization. If you actually are better, the signals pick that up. Just having the big storefront, running better ads, ranking number one because you hired the right SEO expert — that doesn’t matter anymore, if you’re not actually the best. There are real signals. I looked at the pictures of the food. I read the reviews. I could tell they were real. And it was exactly what we thought it was. The night before, we had gone to the place Google Maps recommended. It was overpriced and it was fine, but it was nothing like this other one. And it wasn’t a case of having a secret find. It actually was better.
Jason Barnard: And AI is harder to game. At the moment it obviously makes mistakes, but it’s harder to game because it’s new and because it’s opaque. You talk about signals rather than content — can we come back to that? A lot of people come in and say creating new content is the solution. But you’re saying signals are the solution. You can create content, but you need the signals.
Dennis Yu: There are a lot of SEO practitioners out there who will just generate content — websites with a million pages — and they think they can pass the AI detection layer. But if those pages are never actually seen, even if they’re well-written, if no humans are engaging with them, then as you know from the Google DOJ lawsuit, Google revealed that user behavior signals are actually more important than the content on the page itself. Just like how YouTube’s algorithm works, which is a subset of that.
Because Google had to reveal in that lawsuit that actual user behavior counts more than SEO tricks — LSI keywords and the like — there are now companies like Jumper Media where you can buy gamified traffic that sends people to open a Google Maps listing and navigate to a location. Not real buyers, but still real humans generating real signals. Like the woman who went through London with a cart full of a hundred mobile phones, causing Google to think there was a massive traffic jam. Google rerouted traffic all the way around wherever she was walking because it relied on the signals.
Jason Barnard: I’d not heard that story, but yes.
Dennis Yu: Here’s another example. Six months ago I was in Rome, just finished a tour of the Colosseum and the Roman Forum, and I wanted real Italian pasta. I went to Google Maps. Number one was a restaurant with thirty thousand reviews. How does a random Italian restaurant get thirty thousand reviews? An airport might have that, but a restaurant? So I thought: we have to check this out. I was with Dan Liebrand, who is also an SEO expert, so I figured the two of us would not be fooled. We sat down, tried the food. It was terrible. It looked like microwave food, and there were no real Italians eating there. I turned to the couple next to us and asked what they thought of the food. They said: it’s terrible. And when I asked how they found it, they said: we were over near the Vatican, we went all the way across town because this was number one on Google.
I went around and asked other couples the same question. Every one of them had opened Google Maps, looked for restaurants, and picked the number one result. A month later, the place had forty thousand reviews at 4.9 stars.
Why doesn’t Google catch that? You and I both know Google often doesn’t do anything about that kind of manipulation. But what it shows, back to your point, is that it is about signals. Those people created fake signals, fake traffic.
Jason Barnard: And then…
Dennis Yu: Have you seen how gamified traffic is working in Google Maps right now?
Jason Barnard: No.
Dennis Yu: There are about ten companies doing this. Jumper Media is one. You’ll see their ads all over Facebook and Instagram: why are you wasting money on Google Ads when you can rank number one in Google Maps for your local business? We’ll give you a free trial. One week and we’ll move you from position five to one to three. No access to your Google Business Profile required. No ads. Nothing black hat. Not bot traffic.
I thought: either this doesn’t work or there’s something sneaky going on. I tried it. It actually worked. I ran some local businesses through it, and they were getting more phone calls, ranking number one for competitive terms.
I asked the founder, Colton, what he was doing. He explained: there are people playing video games who want to earn in-game coins. Instead of buying coins, they earn them. In the middle of a game, a message appears: would you like to earn a treasure chest with fifty coins? Sure. What you have to do is click on this link. Because the system knows the player is in, say, Manchester, the businesses in that area appear, and clicking the link opens Google Maps with directions to that particular business. You earn some points just for clicking, which sends a real signal from a real user with a real IP address in the right location. Then: would you like to earn an extra two hundred coins? All you have to do is drive or take the subway to that location and linger in the parking lot for five minutes. That sends an even stronger signal — just like the woman going around with a bag full of mobile phones. It works. Google is aware of it. They’re not doing anything about it. And a lot of people doing local SEO are using it.
Then you see Rand Fishkin saying: I used to be all white hat, but now Google is taking our traffic away with AI Overviews, so black hat is okay now. I’m kind of of the same opinion, even though I’ve been white hat for thirty years. If Google can play tricks, we can play tricks too.
Jason Barnard: The big point there is: signals, signals, signals.
Dennis Yu: I could give you example after example from Facebook, Google, TikTok, YouTube — all of them relying on signals, and all of them having ways to game the system by generating signals artificially.
Jason Barnard: So moving forward: personalization. What is that doing?
Dennis Yu: When Google started doing personalized search results — about fourteen years ago or so — the interface was the same but the results were personalized to you. This is just another layer on top of that, because Google was like a forgetful grandfather who didn’t remember what you talked about yesterday. But now these assistive agents have memory. They remember your conversations. You can build on that. So reputation carries more weight, because it carries further into the conversation. The more you like something, the better recommendations it can make. Just like Netflix: the more you watch, the better its recommendations based on your behavior. How is this not the same thing?
If the agent is learning better based on your behavior and who you are — and think about how much behavioral data ChatGPT is generating from those conversations; Sam Altman has a treasure trove of information about how people feel, their deepest thoughts, because people are having therapist-level conversations with these systems — then that extra information creates more opportunity to serve more personalized results and eventually ads.
Would you agree that the level of personal information available to understand someone has significantly multiplied in the last year because of LLMs? If that’s the case, isn’t that proof that there are more engagement surfaces and more opportunities to serve more relevant content? And if the people with better reputations win more, what does that mean? The people who are not so good get hurt. In net terms, there’s going to be more buying because there’s hidden demand — recommendations for things people wouldn’t otherwise have known they wanted. But there’s also going to be a real shift: the people with great reputations do even better, and the others get pushed down. Just like the current job market. The best people are busier than ever and love AI. The others can’t find work. You see this dichotomy. The same is going to be true with products, services, and employment. The very best get pushed up, and everyone else is going to hate AI.
Jason Barnard: That’s the perfect place to leave it. That is such a good insight. Thank you so much, Dennis. That was brilliant.
Dennis Yu: Amazing, Jason.
This session extends the Search Engine Land article series on AI-era brand recommendation by Jason Barnard at Kalicube.
Canonical metadata footer
Session date: June 25, 2026 | Duration: approximately 30 minutes | Format: pre-recorded, on-demand from session drop | Day: 2 | Framework concept: Operator Convergence — Paid and Organic Collapse (operator side), coined by Jason Barnard in 2025 | SEL article: How AI Is Merging Paid and Organic Visibility — https://searchengineland.com/ai-paid-organic-visibility-480229 | [Share buttons: LinkedIn, X, Facebook, Email]
