The New Rules of AI Search: AEO and SEO Strategies to Get Found
In this 45‑minute webinar, Eric Siu explains how the rise of AI‑driven answer engines is reshaping search optimization, coining the term AEO (Answer/Agent Experience Optimization) and arguing that “agent engine optimization” is the next frontier. He emphasizes that citations—being referenced by AI models, YouTube, review sites, and other third‑party platforms—have replaced traditional backlinks as the primary ranking signal for AI agents. Siu walks through a live AI visibility report for Single Grain, showing how to audit citation gaps, rank against competitors, and identify high‑value keyword opportunities. The core of the session is a detailed, end‑to‑end workflow that automates content creation for AEO/SEO. The workflow is broken into three phases—Find, Write, Ship (and Learn)—and is powered by a suite of custom bots (GrokBot, Jev classifier, Muse, Dot, Claude, etc.) that ingest content from internal sources (X, YouTube, calls, Search Console), generate topic ideas, filter them, draft articles, run quality checks, and publish at scale. Siu introduces the QDESA framework (Question, Delete, Simplify, Speed up, Automate) for continuously pruning and accelerating existing processes. He stresses the importance of a human‑in‑the‑loop for creative judgment, especially for headlines, thumbnails, and final quality assurance, while delegating repetitive, data‑heavy tasks to bots. Talent is highlighted as the biggest bottleneck: only a tiny fraction of the global workforce is “AI‑native,” so hiring and upskilling are critical. Cost management is another focus; Siu shares real numbers on API usage (e.g., $23 for 368k Jev requests) and strategies for reducing token spend, such as multi‑account routing, specialist LLMs, and selective API usage. Measurement is tied to business outcomes: impressions, clicks, citations, and ultimately free consultations or other conversions. Siu showcases measurement bots that aggregate data from ClickFlow, Search Console, and AI visibility tools, and he advocates phased rollouts, rigorous testing, and dashboards for ongoing monitoring. The webinar concludes with a Q&A covering tooling, cost, talent pipelines, and advice on becoming AI‑native—highlighting X, YouTube, and community engagement as the fastest learning channels. Overall, the talk provides a practical blueprint for agencies and marketers to transition from traditional SEO to AI‑centric AEO, leveraging bots, data, and a lean talent model to scale content, cut costs, and drive measurable ROI.
Chapters
We're going to be talking about the new rules of AI search and there's a lot to go through. We're going to be talking through some of our workflows. We're going to be sharing some of these workflows with you guys as well. But we're going to wait for a moment too. While we wait for some people to come in, tell us where you're coming from right now. And we'll wait like a minute or so before we get started. But the AEO SEO game is changing every single day. You can call it AEO, you can call it answer engine optimization. I actually think it's agent engine optimization now. Literally earlier, maybe an hour ago, I used Muse to buy flowers for someone's funeral, right? And it handled the entire transaction end to end. It even translated the language into another language. And then the flower store called me afterwards to confirm it, right? So a lot of— to be the thing that's recommended, that flower shop was recommended number 1. How do you show up for the agents as number 1, right? Well, You're going to have to do better with SEO. You're going to have to do better with AEO, and then you're going to have to do better with agent engine optimization as well. So we'll get started in about 30 seconds or so, but we got people coming in from Cyprus. We got Haley coming in from West Palm Beach. We got Shakar coming in from Mexico City. You got Minnesota as well. So Apollo dialing in from Cyprus, as long as you have your screen on, because there's going to be a lot of screen shares. happening. I'm going to be screen sharing quite a bit here. So what I'm going to do here is you guys can ask questions as we're going along. We'll keep this interactive. Robert, hello from Ohio. And yeah, let's get started. So this is what we're going to cover. So I'm going to switch screen shares every now and then, but these are the new rules of AI search. As you can see these little characters over here, we haven't even defined what these characters are, who they are yet. But it's okay because you have an OpenAI wizard here and then you have a Jev warrior here and then you have a Codex rogue here. Let's just call it a rogue. And then you have a, let's call this one a magician over here, right? And then you have a muse like oracle here, right? We haven't quite named these yet, but we'll figure it out at some point, right? So these are the new rules of AI search, AEO and SEO strategies to get found. And I think Many of you that have been playing the AEO SEO game know that citations are the name of the game now, right? It's no different than how links were the name of the game before. They still are the name of the game right now. You might think link building is dead and it's not quite dead right now, but you have to make sure that you still understand citations, meaning that, oh, you're getting cited on YouTube. Maybe you're getting cited on review sites and making sure that the game that you're playing now is you're getting listed on third-party websites. Not just that you're spamming a bunch of listicles with your own stuff, like best 10 digital marketing agencies. That's not going to work as well. It's pretty easy for answer engines to detect that. So yes, you have to figure out and take a look at how you're listing compared to other companies. And there's a bunch of tools that you can use out there. I mean, you can use ClickFlow to kind of see how you're citated or cited, not citated, cited. And there's a bunch of different APIs that you can use as well, such as Data for SEO. So a whole host of ways to play the game here in order to kind of track how you're going. And there's other tools out there, which I think charge an arm and a leg honestly for what this is. And honestly, most of the SEO AEO tools today are pretty disappointing because they're just showing you how you're cited. And with ClickFlow, you should be able to check for free, at least use it a little bit, right? And there's actually a new citation tool that we have, which I'll show you in a little bit, which you guys can leave a comment too if you want it, right? Maybe you guys get, early access to it. So Max, maybe we can give them early access to this thing because we're pretty close to having it ready. So by the end of this, the promise— so we're doing this a little different. We have slides today and thanks to Max for the slides, right? So I will also caveat this by saying this is basically my first time looking at these slides. So we will go in raw style here and we'll see how it goes. All right, guys. So we're going to show you a citation gap audit, the thing that I'm going to show you in a little bit. Okay. We're going to show you the different checks that we run. We're going to show you the workflows that we have, the agents that we have, the bots that we can have. Steal our bots, guys. Steal our bots. And we're going to show you how we go about scoring this stuff as well. So this is a real life process right now, guys. And we'll go from there. Okay? So I'm going to jump around here and I might just show you the AI citation gap first, the citation gap audit. So this is a— if we want to see how ChatGPT is citing you, we want to see how other language models are citing you. What I'm going to show you right now actually is our AI visibility report. Okay? So this is an AI visibility report that we built. And this is for Single Grain. This is for our website, right? And so what this does is it shows you the number of times that we're cited. So number 3 of 15 AI answers that name you. So 20% of answers. And it shows that we are number 6 among 16 brands that AI names in these 15 answers. And then by the way, this is completely free. Okay. So Max, we'll find a way to get access to everyone for this. Actually, you can see it in the URL right here, Max. So we'll just send people the URL and they can play around with this, right? We'll just see what people put in. And then this isn't even live yet and you guys can play around with this. Okay. So then it says that our biggest gap from a keyword standpoint, it's revenue marketing agencies for SaaS companies. And so AI Names Directive, Kalungi, I'm powered by search, but not you. I'm like, okay, well, that's fine. And then you actually have to hit review my AI visibility gaps and it shows you how you're scoring over here. And so I didn't ask it to do this, right? So it'll show you, oh, you're not named in ChatGPT over here. Okay. AI overview, you're named number 3 for best revenue marketing agencies. It shows who's above us, Google AI mode. And guys, by the way, we just built this on Monday. Today's Wednesday, right? So there's no reason not to be able to ship things very quickly. If anybody on your team tells you, Oh, it's going to take 2 weeks or 1 week to get a design concept out or a prototype out for any of this type of stuff. Nope, you can't accept it. It's too slow. And if you accept it, and if you accept that slowness throughout your entire organization, if you don't set the standard and you don't drive that standard every single day, the entire organization is going to slow down. So it's on you as a leader watching this right now. And even if you're not a leader at the company right now, you will drive the energy. And if that company can't give you the energy, by the way, There's amazing companies out there that would love to hire someone who's AI native just like you, like us. We would love to hire more AI native people. You know why? Because only 0.2% of people, only 20 million people in the world work like this. Okay? We are interested in hiring those 20 million people. So singlegrain.com/careers if you're interested in that. So anyway, so I'm showing you this AI visibility report and part of the reason I'm showing you this too, I'll give you kind of a, there's like multi-pronged attack with this. Okay? So one, yes, absolutely. You could look at, Best revenue marketing agencies, revenue marketing agencies for SaaS companies, revenue marketing agencies for lead generation. I didn't put this in. I just put in singlegrain.com and this is our tool that popped this out. So it shows where we're ranking over here. And then you can look at the gap and you can decide how you want to attack this. We even give you a little bonus here. There's a homepage first screen audit over here where they'll audit your homepage, tell you what you should be fixing, and then you can decide whatever you want to fix. Okay? I'll even tell you the solution as we go through this video a little more. But it's like, oh, you should fix this popup over here. You should offer a name-free step such as an audit, demo, or free trial, which is actually what this is right here, which is what I'm showing you guys over here. And then get name for these. And then it's like, okay, well, what would be helpful here is if you tell me how I could get name, who can I reach out to from these query fanouts where you can see these other sites or other listicles where it's listing other people. How do we get listed there? How do you reach out? And which agents do you use to do that? And I'll go through that in a moment with you guys. But we built this basically as a lead magnet so we can drive paid media traffic to this. We could have this on our homepage as well. Agents can find this as well. And then this is, to me, it's just like, okay, how do we create more free things like this? And how do we create one for ad creative scaling, looking at competitor ad gaps as well? Another webinar, but I'm just telling you guys right now, one, you can use this to see what your gap is. 2, ClickFlow, you can go a little deeper into it to look at what's happening, what's decaying on your website, your AI visibility overall versus this is more of a high-level report. And then you can just continue to make things like this to drive more leads for your business. I'm assuming that mainly if you're in B2B, but obviously you can make a lot of different tools like this even if you're D2C as well. That's one piece of it. I also want to show you some of the output as well as I go through these slides more because the output that we're seeing right now from the content that we're creating, this is an example. This is our workflow. Our AEO SEO bot is creating workflows that look like this. It's like, oh, OpenAI Dots versus Grokbot versus MetaMuse, how to choose the best team agent. This came from one of the YouTube videos that I did recently. It's like, okay, you scroll through this and there's a table of contents over here. And pretty good. There's these images over here. There's internal links, there's external links as well. There's embeds, there's tables, embeds to the original video over here, which is this one. And then it tells you what you can do over here. There's little callouts as well. So I think this is really well done by the team. Team did a really good job with this. There's FAQ at the bottom. The whole idea here, by the way, is I'm going to show you guys the bot in a moment, But the whole idea is that you create a bot like this and then you hand it off to your team and your team who's already in the trenches, they're gonna know, they're gonna have more context than you have because they're actually on the battlefield and they're working with customers. They're working with customers first and foremost, but they're working with your stuff as well 'cause this is going to our site. I'm always gonna have more trust in what they're doing versus what I might know because I'm running around at free safety within my organization fixing whatever the biggest bottleneck is. And that is your job if you're a founder in your company, that is your job to look at what the biggest constraint is and unlock that constraint and then move on to the next problem, move on to the next problem, inject your energy, inject your energy, move on to the next problem, move on to the next problem. And that's the game every single day. And the whole point of business is that you're a problem-solving machine, right? And there's no shortage of problems to solve. You're never done solving problems with your business, okay? So, because if you are, then you're done growing, okay? Which is not fun. So this is another example over here, okay? Here's SeedDance, how to scale more ad creative. This is my first time seeing this image over here using ChatGPT 2.0. It's a cute one. This is based on a webinar that I did with Cody Schneider over here. It actually calls out some of the screenshots from it as well. This is pretty cool. This is my first time seeing this one. I'm like, this to me is— we're refining this every single day. What I'm going to show you as well is I'm going to show you a workflow that we use To describe how this works. And I'm going to show you the bot in a moment too. But the whole idea here is that this allows your humans to focus on human work, which should be more strategic if I'm being honest. And the robot work should go to robots. And right now the problem is we humans have been so used to doing the robot things, the things that robots can do now, that there's a lot of inertia. We're so used to doing certain things a certain way. And the old way, I should say. And I remember recently, a few months ago, we had a recruiter on the team and she made a comment saying that, oh, I have 150 emails to review. How am I supposed to go through them this weekend? I didn't have the time. And then I didn't even have to say anything, but there was maybe 3 other people in the room looking at her. And it was just a look of concern. It's like, because you can't say that type of stuff because you just let the robots do it. You don't need to wait 3 days to review something by hand, right? Just doesn't work. You have the robots do it. Or if you need the robots to review a bunch of— review all my content, for example, look for the content spikes and then draft it and then run it through an evaluation to determine that, hey, is this duplicate content already? Have we done something out there? Or any of the statistics that we're citing in the article, are these in fact verified statistics? Or can we cross-reference it to the original piece of content? Can we embed that piece of content in here? And then can we also create some images as well? Can we make sure we have some tables? Can we make sure we have some external links as well? Can we pull some external statistics as well? These things can all be combined in here, but there's no one person that can do all this stuff. In fact, the old way of doing this is having it be almost like a factory line where you have multiple people doing it, but every single person that you add adds a layer of complexity and adds delays to the process, right? And so there's a great book out there called The Algorithm right now, which this guy used to be president at Tesla. And it talks about the algorithm that you run. And I would encourage you all to think not just from an AEO SEO standpoint right now, but for all the processes that you have, QDESA, right? Question, delete, simplify, speed up, automate. Okay? So Max, if you can put that acronym into chat, QDESA, right? Again, question this process. What I'm showing you on the screen right now, why should a human have to spend 2 to 3 to 4 hours to write an article? Why should it take that long? What is it exactly? Is it the research? Is it looking for all the content spikes? What is it exactly? Why does it take that long? Oh, well then you start to break out all the processes from before. It's like, well, that doesn't need to be done that way anymore. Does a human still need to take 1 hour to 2 hours to edit a piece of content? Well, no, because you have evals that the original editor can create. The editor has that taste and they can put evals in there. Eval are the definition of success that you pass to the AI, and then the AI knows what good Looks like. And then you can just continue to tweak the process. And then when you look at the process, okay, first you question everything because I'm willing to bet 95% of the process that you have right now are pretty stupid, right? I think that's true for us, right? Probably true for you. So question, then you delete. What can you delete here? Why do we need to have a human in this process over here? Can we just have the robot over here? Can we have someone work on something else, right? There's that. And so then you simplify, then it's like, okay, well then you simplify and then you speed up and only then do you automate. By the way, Max, I noticed that LinkedIn and Instagram aren't working. You didn't test it. So anyway, so all I will say is that's another piece too, right? Testing is a really important thing. When you put a process together, and this is me giving Max some shit, but like, and you guys can't see him, but Max is a great, wonderful producer for this This webinar. But when you're doing a process, everything is a workflow, right? So when we're running this webinar, for example, why are we not testing LinkedIn? Why are we not testing Instagram if we can? Why are we not testing to make sure that these are working first and foremost? And then just so the webinar presenter doesn't have to spot that issue during the webinar, right? But this goes for every single workflow, whether it's AEO, SEO as well. You have to make sure that you're first testing it on yourself first, right? Right? So it's a phased rollout. You don't want to just say, hey, all the stuff that Eric's talking about right now, let's just roll it out. And no, it's like, first you test it on yourself. So we're testing it on our properties. Then once we feel good about it, once we start to see results for the next 7 days, next 14 days, next 28 days, next 60 days or so, then we say, okay, we're ready to roll this out for some clients. And only then do we feel really good. Do we go balls to the walls and then all in on it, right? But if we can't, if we can't figure it out for ourselves first, it doesn't deserve to scale. So phased rollout. For a lot of stuff I'm talking about, and I'm not just even talking about the AEO SEO stuff. So there's that. All right. So I'm going to move to, I'm going to show you guys the workflow chart over here that I have for you. Okay. So I have, let's see, let's pull this up. I'm going to change my share for a moment. Let's do this. All right. So this is a— Heath, to answer your question, this is a GrokBot over here. All right. Ahead of this webinar, I said, hey, Grokbot, can you make me a state chart? Not even just a flowchart, a state chart to show how things are running with this AEO SEO bot right now. I just sent this to the ContentOS team earlier today. What it does, we have 3 parts of this process, which I'm going to show you the bot in a moment, and then we're just going to continue through this webinar. 3 parts to this. By the way, this used to be mostly done by a human. 2, 3 years ago. Not anymore. Okay. So first part you find, okay. So find, write, ship, and learn. Those are the 3 parts of the process. So you can see for this lane over here, it's we ingest Eric's content sources. Okay. This is just for us. So it might ingest from my X account. It might ingest from YouTube. It might ingest from sales calls, customer calls, internal meetings, Search Console. There are tons of content spikes. And I would also say that those of you that are recording your calls internally within your organization, Obviously you're not going to publish everything out there, but your best content actually comes from when you're just talking to your team. And because you're not thinking about it, you're just, you're living you, you're being authentic. And that oftentimes is going to be the best content. And then the next step here is mining 40 topic ideas. So 9:15 AM, we'll make sure that we're looking for the best content ideas. Not only that, we have a quality guard for this. Okay? So we'll look to see that nothing is a duplicate. Nothing looks like a template. Nothing is— we're not kind of missing— Mixing up the sources. And then what we do, so by the way, if it fails, we will rebuild the rows. Okay? So this process is going to rebuild the rows. And then after that, we run JEV. So JEV is a classifier. And those of you that have seen my YouTube videos, go check out my YouTube channel, watch my JEV videos on why I think it's unfair. It's an unfair advantage for work or why it's like God mode with AI because you can classify things very quickly for very cheap. Right? So JEV, We'll filter out our content for, does this cover content for our buyer? And only if it's clear, it moves on. And so this prevents, let's say our editor Haley, for example, from her having to do all this stuff. She doesn't have to come up with the 40 ideas. That used to be one person. This ingesting all of my content sources, that used to be one person. That'd probably take hours. Coming up with 40 topic ideas, that would take you hours. Having a quality guard as well, that's someone to just audit. That's a different set of skills. Having another classifier filter on top of it, that's another, like these are multi, these whole things alone would take days of work, right? And I say days because it would take hours and hours to compound this stuff. And then you have a search check, is the intent there? Is the demand there? Is there overlap with our own posts, right? And then if there is overlap, we drop that, okay? So that's only the first part of this right here, the find piece. I'm telling you guys right now, you got to look at your most important process for us because we do AEO SEO for our clients and we do paid media. We are scaling ad creatives for our clients as well. If you aren't thinking about the highest leverage processes that you're doing right now, the ones that are maybe driving the most revenue for you and you're not redoing them, you're in trouble, right? And thank God you're on this webinar because we're just going to nerd out on this stuff every single week, right? Almost every single week. But for me, I'll just tell you, from a founder standpoint, yes, I helped generate this for my team, but now I've handed this off because this isn't my responsibility. I just wanted to get it going and then hand it off to the team. That's one piece. But the second piece is I'm realizing that based on the amount that we might be able to close from a pipeline standpoint, talent is the number one constraint. Because if there's only 20 million people in the world that work like this, 0.2% of people, it's hard to find these people. And so I need to make sure that our talent workflows are good, but we run it through the same type of process here where it's like, we just look at everything, we question it. It's like, man, then when you look at the process, it's like, Whoa, this is like 10 or 20 people doing a job and it's being done a lot faster. And you know what? No human wants to do these jobs anyway. So that's great. Anyway, next step here. So I know it says Eric for the right section, it says Eric picks topics. It's actually the editor that picks the topics now, right? And then she will decide or he will decide inside of Slack because we put these Grokbots into Slack. They will decide what to ship, what not to ship, right? Because GrokBot for Teams is pretty good. And look, Muse is going to come out with their Teams version. They're already working on it. We're going to see an explosion in agents, right? I was watching an Elon video yesterday with him talking about how GrokBot is the fastest thing he's ever seen grow, right? And I believe that. He believes that most of the SI or superintelligence compute, they'll be able to take a lot of that compute power because SpaceX has a lot of compute, right? GrokBot can eat up a lot of it. That's what he believes. Anyway, we draft it in WordPress. We use WPvibe, which is a great MCP for our CMS. By the way, there's a PHP issue that came out. Those of you that are on PHP and Node, you guys should check that out. There's a big security vulnerability that came out today. You guys should check that out, hand it off to your team. No worries. I'm here to help in all ways. Next step, rules. We have a rules checker. Does it follow the rules? Is it failing any of the checks? Maybe over 100 checkers. And if it fails on the critical ones, then it will make sure that it fails and then the editor will check it, right? And then in this case, Haily might review it and then she'll say, okay, let's go ahead and publish these, right? And so she'll review a couple of these. She's not going to— if we're going to publish 100 in a day, which I think in the very near future, guys, just keep in mind, when we interview people, oh yeah, Google really doesn't like it when you publish too many. I'm like, okay, so how many posts should we publish per week? Probably 2. We're in 2026 and you're telling me we should publish 2 pieces of content? It doesn't work like that anymore, right? If you think about it, Google has changed their guidelines to say, hey, as long as your content is valuable, as long as your content is valuable, we don't care if it's AI, right? Obviously don't push out a bunch of slop, but there is already a proliferation in webpages. There's already a proliferation in agents crawling your website. Why the F would you not publish more content, especially if you can ingest more content at rapid pace, right? So for us right now, we're just going for 20 a week. But my full expectation, because we have some of the people from my team on this call right now, my full expectation, guys, is that we jack it up to 40 to 60 to 80 to 100 to probably even more than that, if I'm being very honest with you. And then there's going to be some bottlenecks with talent because I still believe that you need a human in the loop here. But you can't— there's just going to be more, guys. There's going to be more abundance. There's going to be more content. It's going to be harder to stand out. But the cool thing is if you're ingesting from up here, the top left, you have an original point of view. And that original point of view, no AI is going to be able to copy that. It's you just coming from the heart and you're going to have your own point of view on everything. And that's why a workflow like this is special. Okay? Next step here. So then we ship it on our site, live on Single Grain, which I just showed you 2 examples. And then we'll check if it's indexing. If it's not indexing, we'll ping it to an indexer. And then we'll measure. Okay, this part's important, right? Don't just build. A lot of people have a build disease right now. Let's just build, build, build, build, build, build, build, right? You want to build, you want to measure, you want to iterate. And I don't know how that's a new concept in 2026, but it blows my mind right now that people just build, just throw a bunch of slop out there. And maybe that's why Bain thinks that 94%, they did a survey, 94%, only 6% of people find value in AI. And then it's like only 2% of consumers are paying for AI right now. Well, yeah, no shit. Of course. Sorry, pardon my language, but of course most people have inertia. But if you're on this webinar right now, you don't have inertia. You cannot afford to have inertia. You must push forward. You must be the one that is carrying that torch forward, right? I feel like I'm like a motivational speaker sometime with this stuff, but this is how I feel. People are just moving too slow right now. Again, no crap that 94% of people believe that AI is not really driving ROI or only 6% of people believe it. Because people are lazy, people are slow, and I don't understand it. I don't want to understand it. You shouldn't want to understand it either. Okay. So anyway, all that to say is we'll measure the impressions, the clicks, the citations, right? And we'll measure all that through day 28, and then we'll measure the winners. Okay. And then we'll feed the winners back into, should we repurpose these or should we make a permutation of this post? Okay. And Based on the Seedance one on scaling ads to tens of thousands. Okay, should we make another one? Should we make one on Seedance versus all these other ad creative tools? Or should we make another agentic comparison? Or should we go deeper on how instead of OpenAI versus Muse, Dots versus Muse versus Instinct, should we just have go deeper dive on those? We probably should. And in fact, me doing this webinar right now, it's going to find content spikes. And maybe I might go to my harness and say, hey, I actually want to do the deeper dive on OpenAI Dots, on Muse, on Instinct. And I want to compile all the videos that Eric has done on these, all the videos Eric has done on Jev. And I want you to compile it into a long form for X. And I want to do it for the blog as well. But do you see how this is working? Because with me now, a lot of times I'm just sitting on my couch. I'm opening up GrokBot Voice. I'm just talking to it. I'm opening up my OpenAI Dots. I'm just talking to it. In fact, when I drive to see my dad, 45-minute drive, In LA here, I don't even listen to podcasts when I drive anymore. I just call my Dot and I just talk to it for 45 minutes, that entire drive. That's what you guys should be doing. You guys are going on a walk, talk to your Dot, talk to your Grok bot, talk to your AI. That's what you should be doing. All right. So final thing on this flowchart over here or the state chart is a decay refresh. So if you are seeing that your content is starting to decay, guess what? You don't have to see it anymore. You have a bot that is managing the decays. You have a bot that's managing your competitor citation gap as well. And it's calling out what content your competitors are making that you're not making, that you should be attacking, that you're not attacking, right? Don't got to worry about that anymore, right? That used to be another role. There used to be roles dedicated to full-time content updaters, right? Now you just hire AI-native people who are very hungry with this stuff and have good experience and then boom, off to the races. In fact, I'll say this. When I showed this workflow to this ContentOS workflow to the team, some of the responses were, yeah, I feel like I don't have to work. I don't have to just be this role anymore. I can be so much more than that. And that is the goal of this stuff. How do you free up your team? How do you get your team to see the light? And once they see the light, they will unlock themselves. But it's just a matter of sometimes it's As long as your team has the will, then you can train up the skill piece. But if they don't have the will, it's not much you can do there. You can't train will, you just can't. Skill is more trainable. All right. So anyway, I will continue on here and then Paulo, yes, I will share some details that will come in a moment as we go through this, but I will stop my share on this side. So I will go back to Max's wonderful slides after I get a sip of water. And by the way, guys, keep your questions coming because I'm going to— I'll probably get to your questions pretty soon, I would imagine. So, okay. So let's take a look at this. Okay. So this is one of the bots that we have. And Max, the QR code we're going to share, I'm assuming it has access to the bots. Is that a thumbs up? Okay, that's fine. Okay, cool. Yeah, we'll give you guys access to these bots as well. And by the way, I'll tell you something cool too. Thank you, Eric. My video is crispy. It's going to rain over here today in LA. So anyway, so we have this bot here called Team GEO Audit. Okay, we're going to share all these bots with you guys. And you know what? We're going to add the access to connect to Single Brain Gateway in there. I haven't even talked about that. I don't need to promote with that. But just to give you guys some examples here, I mean, this is what a query fanout might look like on the side, right? You can hit show all. This is like with the Google AI mode, but you can see what sources are citing. And you can see that, okay, this page, there's all these B2B pages over here. This is funny because this shows October 16th max. This actually still reinforces the fact that even if you make listicles on your own website covering your own agency or your own services or your own product, It still works because the fact that Google's AI mode is citing these over here is pretty amazing. That's what I'll call out. But guys, if you wanted to say, hey, I want to find out what the citation gap is on my website, you can literally open a Muse, a Dot, or a Grok bot and just ask it to do it for you. And then that's why we have this audit bot over here. It used to be a year ago or 2 years ago, it's like, oh yeah, you're going to have to download this software over here. Guys, you just talk to your harness and you just make what you need when you need it. That's it. For whatever reason, we all still think in the past. I have to give a little grace there too because we're on October 21st and it wasn't until the last week of December where everything hard takeoff for AI happened, last week of December. So it hasn't even been a year yet, guys. But you know what? I have high expectations and I have high expectations for you. So I hold you to a high standard and I hold you in high regard because you are here right now to learn. I'm going to jump around over here. Let's see. Let's jump to these bots over here. Earlier we talked about this. These are 3 bots that we use inside of GrokBot, which maybe Max will just email to people afterwards, people that attended this webinar, maybe even just all the people that registered. But we're going to give you guys our Team AEO SEO bot, which is what allowed that workflow that I showed you earlier. We're going to give you access to that. Ain't going to charge you anything for that. And then Team AEO SEO measurement bots. We made a separate bot to measure how things are performing over time. Team GEO audit as well. We'll measure citation gaps and then look at how— actually prepare briefs for you too on what content we should be creating. And then the reason we have these in team is because You using these on your own, that's cool, but it's much higher leverage if you have your AI agents working with your team inside of Slack or inside of Microsoft Teams. Because if you gain all the leverage and you hog all the leverage, you gatekeep all the leverage for yourself, you can't possibly compound as quickly as you would want to as an organization. And so it's on you to drive that forward. So as I mentioned earlier, you have all these ingestors over here and then it'll I'll push it to bot and then we'll push this content out there. And then the type of content we're pushing out there too is I showed you guys, these are about 1,500 words each. You guys saw the quality. I showed you 2 examples. I am willing to bet a good amount of money that these will perform well. And as we continue to— those of you that are in services, you build these workflows for your clients. You're always going to have a job if you know how to work like this because someone's going to need you to optimize these bots. Someone's going to need you to upgrade these bots. Someone's going to need you to maintain these bots. That's not going to go away. And if you have taste, if you know what's going on in your industry and you understand how AI works, those 2 combined give you ultra leverage. So that's how I think about that. So I showed you the bot already, but the thing we should talk about too is Jev. So Jev is an API that you can use. Oh my God, I don't know how to use API. Guys, just talk to your harness and tell it to use Jev. That's all you need to do. Literally all I do now, I am the laziest person when it comes to this stuff. I will read stuff on X and either I will forward it to my team, hey, this is pretty cool, or now I just forward it to my harness and I just say, hey, how high leverage is this on a scale of 1 to 100? And then can you just do it for me? And oftentimes it'll say it's not high leverage. And sometimes it'll say something that maybe is not so smart and I'll have to correct it and then we can tweak it. But the way I see it is if I see an interesting API, I'm just going to try it because it costs me next to nothing to do it. And Jev, I will tell you this. I looked at the numbers the other day. I think we spent, and I know you guys like to see numbers, so I might just show you a screen. Okay? You guys wanna see the screen? I'll show you the screen. So let me explain what Jev is first, and then I'm gonna share my screen to show you what we spent on this so far. So Jev is a classifier. And what that means is you can look at, for example, I can look at my entire LinkedIn connections list. I can say, okay, Who's a good mergers and acquisitions target here? Who's good to potentially recruit based on our recruiting, our careers page? By the way, we're recruiting strong SEO and AEO people, strong paid media people, strong salespeople, singleoring.com/careers. So anyway, we can look at who's ICP fit that's connected to me. Okay. And who should we be reaching out to? What's the different offer? So do I want to go through? I don't know how many connections I have. I think it's like, I don't know, 50,000, 60,000 followers, but then connections-wise is a little less than that. Nobody's going to go through that. So you use Jev and Jev will classify all those people with probably in a few minutes and it'll cost you less than a penny to do that. And it's not going to hallucinate much. And so let me just show you what that costs. Okay. So, and I guys, I want you to type into, I want you to type into chat on 350,000 requests that I ran in Jev to classify things, how much money do you think I spent? Drop it in the chat right now. And Stacy, you can even tell me how much money you spent, um, on JEV so far. But guys, 350,000 requests. How much money did I spend on JEV? You know, is it $1,000? Is it $3,000? Is it a couple hundred dollars? Is it $50? From what Melly said, is it $200? What is it? So Rob says $500. Stacy says $30. Bastian says $500, $35, $15, sub-$20. Okay, great. Let's take a look because I looked at this the other day, but I don't have an update on this. Let me just do this. Let me move over here. Here we go. Let's look at usage. Guys, I went up by almost 20,000 requests. 368,000 requests, we have spent $23 on this. That is crazy. It is really bad if you're not using this right now. Look at this, 556 million tokens. All right. So cheap, so crazy. It absolutely stays. Guys, you just need to start using this stuff. And I'll say one more thing. Jev just added images. And that earlier AI visibility audit that I just showed you, Jev images. You can make any lead magnet that you want. You can do these analyses a lot faster. Sometimes you might have other API costs that you're bringing in there, but For each one that you're running, if it's costing you $0.004, why the F would you not run it hardcore? Especially when it comes to outbound or especially when someone reaches out to you. How do you hit them with value? Constant value. Instead of your sales team saying, oh, just checking in, just following up, just checking in, just following up. How about you send them value, value, value every time? And now you can do it for nothing. So Robert, how How are you spending $250 to $500 a day on Jeff? Am I misunderstanding something? What are you spending that on? I have to know. Please tell us. Please tell us, Robert. $250 to $500 a day. What are you spending? You must be spending on frontier tokens. Okay. Yeah. So I was doing that, but you don't need to do that. I mean, which API, which frontier tokens are you going with? Just drop it in chat. I'll come back to you in a moment. Because I was spending like $15,000 to $20,000 a month for one month. And then I figured out how to optimize for that, which you guys can let me know if you want me to answer that in a moment. So anyway, let's continue here. So I'm going to go back to Max's beautiful slides. So let's go over here and let's move over here. Okay. So what I just showed you earlier. So if we have like 80 ideas coming in, okay. So let's say 52 of the ideas are fresh and then only 5 are drafted, 3 pass. Okay. Basically what Jev does over here is you can set multiple levels. So you can say, okay, one, are the statistics backed up by evidence? Is that just a yes or no? And then also confidence score, on a scale of 0 to 1, what are the odds that we've actually published something like this? And anything that's under 50% odds, then you're probably safe to publish it. Because sometimes you do want variations of a topic, You don't want to just have net new for everything. You want to go deeper. So there's that. And then do you have images in there? Is there AI tell? This one's really good. Is there an AI tell? So the AI tell basically is like, does it sound like AI slop? And we have a no slop, like a humanizer skill. And that's one thing to call out too, because a lot of the bots that we have, sometimes I get overzealous and I upgrade it too much. There does come a point of diminishing returns where you probably need something like a bot improver So we have a bot improver that actually has Claude Code and Codex combined. And the little flowchart I showed you earlier, each week it's required to report on the state chart of the most important bots that we have. So it could be the recruiting bots, the AEO SEO bot to check them. And then it'll check if the Codex will check if the code is good and then Claude Code will check if the output is good. But then you have adversarial review and then you have a bot improver that's constantly improving your bots. If you don't have something like that, Your bots will start to drift. And if you start to add too much stuff to your bot, it will start to drift as well. Also, your Grok bot is running on a virtual machine. Here's the problem with that. Oftentimes you might have a bunch of different tabs open, your Grok bot will. You have to make sure that you are culling the different tabs that you have open so you're freeing up the memory because a lot of these virtual machines don't have a lot of memory. I'm hoping that we can maybe bring our own virtual machines in the future so we can add as much memory as we want to it. as much context as we want to because I think all the agents will be talking to each other very soon. So there's that. And then let's see. Yes, Eric, bring your own virtual machine. Instead of BYOB, BYOVM. Okay. So Max, I don't know if you guys drink beer in college still, but this should be the BYOB now, BYOVM, and you guys should just have LAN parties. But anyway, Let me show this question, this point from Robert. I want to address it real quick. Let's have it go for 20 seconds. Okay, great. Robert says, so it's $250 to $500 a day on Anthropic mostly, but I run on the frontier and open source models. Right now I'm running GrokBot on one screen, Claude on another, ChatGPT on another, MetaMuse on another. Great. Robert, we have a very similar setup if I'm being honest, but I was spending $15,000 to $20,000 a month. on Anthropic API costs. This was earlier in the year, maybe like January or February. That's when I bought my DGX Sparks. Now they're more expensive. I knew they were going to go. I should have bought more, but I knew they were going to go up in price. I would say that if you want to run models locally, maybe you want to make a specialist model. For us as an AI marketing, AI native marketing firm, for me, I'm like, okay, how do we make a marketing LLM? How do we make something that's dedicated for our customers where we just have the best workflows and we have data that's just helping us improve workflows for clients? That's how I see it. That might be helpful because Coinbase recently, they came out with their own. They use Qin, which is from, I think it's from Alibaba, but it's one of those Chinese open source models and they use it. I think maybe they use it to spot fraud, but we're going to see a lot more of these specialist LLMs coming out. That's one thing. But in terms of bringing your token cost down, so I think it was SemiAnalysis, something like that. It was one of these AI research firms. So they did an analysis. This was a few months ago, maybe 2, 3 months ago. But if you were paying at the time $200 for a Claude Max subscription versus a Codex subscription from OpenAI, $200, same amount of dollars, the subsidy you were getting from OpenAI was about $14,000. The subsidy you were getting from Quad was maybe $8,000 or so. And I actually saw that reflected. So what I did was I just bought a bunch of different accounts and I would just switch accounts when I needed to. And I got more than enough. So if you're using things a lot more, then maybe you need the API calls, but we don't really use API calls. I just buy myself multiple accounts and I know that I'm going to get a subsidy from it and that's going to cover my costs. Anyway, I'll just leave it at that. I know Hermes will obviously make API calls to frontier models. Maybe you need a model router, maybe you can use RAMP for that, or maybe you can use OpenRouter for that. That's up to you. But you can even use Jev as a router if you want to. That's up to you as well. Anyway, I know we're talking about AEO and SEO here. Let me just go back to that. Sorry, I apologize for the drift. This is what Jev will do. There was a bot draft over here. And then Jev will classify it. But then when you have a human, this is why I think human in the loop is very important with your AEO SEO content. There's just, you can't beat it. Because I'll say this too, on my X profile, so we have Max helping with that, right? And so on the X profile, it's like, why aren't the posts performing that well? Well, it's because we're just using an AI skill for it. But similar on my YouTube channel, whenever I let the AI just handle the packaging, meaning the thumbnail and the headline, It just tanks. It doesn't work. But if I spend 5 to 10 to 15 minutes on the headline and the thumbnails, it performs a lot better. If you look at that, you can just run an analysis on my YouTube channel and you can use Jeff to do that. They should pay me for saying all this about them. But so you have, I believe that when it comes to creative work, non-verifiable work such as creative work, you need to have a human in a loop. And so you can still say, okay, end to end, you can have the human say, okay, just do this and have like 90% of the work done, but you still need that last 5 to 10% of judgment to be there, right? Otherwise it's just not going to do that well. So that's what's been missing, I think, from our AEO SEO workflow. And let's see. Okay. Oh yeah, we have this. Yeah. So this is a QR code that we have. So Max, this is the QR code. I think it's more of the, we showed you all these things, but if you want our help to do these things, You can just scan this QR code because this is what we help with. It's more our AI native marketing team building these agents, building these workflows that live inside of your Slack and your Microsoft Teams. So I'll just leave this up for another 10 seconds as I get a sip of water and then we're gonna continue on through these slides. Man, it's already 43 minutes. It's going fast. All right, great. So here's what I'll say. We talked about the measurement bots as well. So if you publish something, you index it, you have to measure. You should be measuring. Sure, you can measure impressions if it's indexed, clicks and all that, but you have to very quickly get to the point. And the point for us at least is generating more free consultations. So sure, we can get more citations, we can get indexed more, we can get clicks and all that, but none of that ultimately matters if we don't drive more consultations. And so that's why we had this measurement bots. And we know that for the first 1 to 6 weeks, maybe we're looking at impressions and clicks, Maybe not so much free consultations because you're not going to get that many impressions and clicks initially. Same thing with citations, right? But then after that, maybe after the first 6 weeks or so, then you're looking at, are we actually driving more free consultations from this stuff, right? Or whatever your main conversion is. And the cool thing now is whatever workflows you create, you can just say, hey, I want to make sure that this is driving ROI so I can prove to my boss that the work that I'm doing is actually valuable. And you can do that. Previously, it was harder to prove. And the funny thing today, when you talk to a lot of AEO SEO people, when we interview people for SEO AEO roles, yeah, we got our share of voice up from 5% to 20%. What the F does that mean? Right? And so you should just end the interview right there if they say that. And then what I really hate too is like, oh yeah, I drove the traffic up, but I improved the conversions by 200%. Also, what the F does that mean? Off what baseline did you improve it from 1 to 3? What does that mean exactly? Clarity of thought matters a lot more in today's day and age. And I've said this on many webinars before, AI is IA. It's an intelligence amplifier. So you will very quickly see whether someone is very smart or if they're not a fit to work with you. I'll just leave it at that. Okay. I will say that we did have an issue before. So in terms of publishing the 21 posts, so we published 21 live, but in the first week, none of them were getting indexed, right? So there's different tools that you can use. Google Search Console has a ping tool that you can use. IndexNow. There's like a handful of these that have APIs that you can use. But just keep in mind, you don't want to go crazy with APIs because it can start to get very expensive. They start to stack up on you. And you can see with Robert's cost, they started to really stack up, right? So we'll leave it at that. So anyway, you can fork our AEO SEO bot. We'll share it with you. We have it in our Skills Dojo. You can browse our templates as well. So thank you, Max, for sharing that. So skillsdojo.com/templates. Well, also Max, if we can email them afterwards as well, that's going to be good. And if you need help, again, singlebrain.com, or you can just scan that QR code. We can help with all this stuff at the end of the day. I will turn it over for Q&A now. I'll leave the screen share up for a little longer, but we'll cover Q&A for a little bit. And I apologize for talking so quickly. So Max, you got some good questions to put up here? What's up? What's up? Let's see. I can't hear what Max is saying. So let's see. Robert Crabtree, please let us know if you do end up saving money, because I almost want to bet a lot of money that this webinar just saved you a ton. So if you're spending $15 grand, okay, that's $180 grand a year. This webinar just made you $180 grand. Okay. What a great headline. Okay. So let's see. What are we using for tracking anything that's bot native? So Stacy, In terms of what we're tracking in terms of success, I just showed you the AEO SEO measurement bot for this piece. That's what we're using. And that is tied in with ClickFlow. So that's one of our tools. And mainly I just use ClickFlow now for tracking, if I'm being honest. So we can measure Google Search Console so we can see what's decaying in there. And we can also measure AI visibility, right? Because that's connected to another AI or API that can measure What the citation gaps and things are around that. So that's what— then we just combine it into our AEO SEO measurement bot. Now Xander asks, do you have a dashboard to keep track of all these so you can purge? So we do. What I'll say is here, let me kill my screen share. Okay. So do you have a dashboard to keep track of all these so you can purge and merge PRs more efficiently? So we do have an internal repo for all this stuff. For each of these bots, they have their own repos. And then we're able to push things towards it. And the way we work with our clients is we do have, for each team, the media team has a media OS and the content team has a content OS now. And then we expect each of our clients to have their own repos and then we can continue to upgrade and compound on that. So then we're all feeding from the same context here. And so the whole idea here is with these bots too, we can customize the scaffolding for each of our clients because they're going to have different needs and they're going to have different preferences as well. Wow, Stacy. Okay. That's amazing. So I didn't know I was a cost saver. This is great. So cut costs from $8K to a steady $3K. I'm assuming that's per month. So that's pretty awesome. That's $60 grand a year. That's awesome. And then Xander, In terms of— okay, great, Stacy. So $60 grand a year. So Xander, how do you keep track of these? And is there a template for each repo? We don't have these templates out there, but I would say that it shouldn't be that hard. Here's the thing. I wanted the team to work like this in the beginning of the year, but if you don't have people on your team that instinctively know how to work like this, where it's like, oh yeah, obviously they should go into repo, it's just not going to happen. So I think that's much more of a talent problem because Now the team just works like this because the people are AI native. And so you're just going to have to rejigger how you go about talent and how you think about talent in general. And then everything's going to flow down because this, I'll say this for another webinar, another time, but my naivete, and one of my friends likes to say that I believe in, I have too much faith in people, which I do. I believe in people. I believe in people in general. I think they can learn this stuff, but You can't train mindset. And so I've said this many times, whether it's on the podcast or whether it's on this webinar, mindset is not something you can train. And unfortunately or fortunately, AI-native people, you can see if they have the mindset or not. So I'll just leave it at that. Yes, exactly. You can lead a horse to water, but you can't make them drink. So let's see. Maybe there are a few other ones over here. API. Okay, got that. Got that. One other thing I'll say too, Is you can use— so different bots have different use cases. And I don't think you should be tied to any of these right now. I think you should be relatively promiscuous with these things. Obviously don't be promiscuous in real life, but be promiscuous with your bots. So you can see Robert here is using Hermes, he's using Muse, he's using GrokBot, he's using Claude, he's using all these different ones. And I have different use cases for each one. So for example, the e-commerce, the agentic commerce thing that just happened using Muse, I made a purchase earlier today. I use Muse just for looking at my Instagram in terms of what I can do from an organic social standpoint and also paid media because it's connected, it's owned by Meta. They're going to have the best connectors in there and they're going to be most seamless. So that's how I think about Muse. But because I pretty heavily use Codex because it's also controlled on my computer. And so I use Dot quite a bit. So I talk to Dot quite a bit. I talk to GrokBot a lot. And in fact, the only thing right now obviously that's running on API costs is GrokBot. GrokBot is eating up— I'll run through usage very quickly because on the team account, you don't get much. Individually, you can get 20x usage, but on the team account, you can only get 5x more. So I'm trying to work with the GrokBot team to buy more usage because the API costs are destroying me right now, but it's just so worth it. It's so worth it to be able to get these things done and to be able to free your team up. And I'm just monitoring usage with my team right now to just see, okay, who needs upgrades, who doesn't? And so that's a whole nother thing. Anyway, I mean, I think at a very high level, just to close things off here, it's like from an AEO SEO standpoint, the workflows that I showed you, you should be questioning all the workflows that you're running right now, make different bots for them, spin them up, and then try to free up your team as much as you can. And only then can they work on the highest leverage things. And then they will thank you for that because They never really wanted to do that stuff in the first place, but they just didn't know any better. So that's what I would say. Now, one more thing I will say, let me just show this. This is from Elena. How to become AI native for beginner? We ask these questions now. We say, how much time are you spending on X per week during the interview process? If you're spending less than 3 hours, we won't even talk to you. And the reason for that is because the action right now is happening on X. Well, oh, X is so toxic. Da da da. Well, that's fine. That's because your algorithm isn't tweaked yet. But if you focus your algorithm to look at the latest AI news, right? So if you're following some people, I'll just name some names over here. Lauren Tan, the engineer at— she used to be at Cursium. Now she's at SpaceX. She is always talking about a lot of stuff from from GrokBot, right? All the new stuff that they're shipping. Matt Palmer would be another guy from GrokBot. Greg Eisenberg shares some good stuff. And then what happens is you start to follow more and more of these stuff. You start to follow more and more of these threads. Your algorithm will be tuned to that. So my first answer would be, if you want to become AI native, you're on X. The second thing is you cannot escape this. You must build. You have no choice but to build. In fact, when we interview people now, we ask for a Loom video, 10 minutes. Show us the coolest thing you're building right now. Just talk through it. We can very clearly see how you're talking through it, the models that you're using, the toolset that you're using, that within 10 minutes that explains a lot more than just a simple interview. And other than that, I don't think you can take a course because any course that you take becomes outdated quickly. But YouTube's a great place. X is a great place. That's where a lot of the action is happening. So that's what I would say. Now, Michelle, I'm going to answer this one. Let's see. So let's put 20 seconds on this one. So let's show— oh, it's not showing it. Okay. Well, okay, there we go. So Michelle says, I use them all. Many times I found their strength and weakness vary depending on the task I have. There is no perfect bot. Exactly. There's no perfect bot. And these things are always leapfrogging each other. Muse Wonder came out recently and it's supposedly at frontier level. I use it with Muse. It's been great. So I'll leave it at that. So anyway, all that to say, guys, if there's any takeaway from today, it's you want to, again, have robots handle robot things and you want to have humans handle human things. And if there's anything from what I showed you from that workflow earlier, that should show you that you can not only free up your valuable humans from 95% of the rote tasks, but you can have the robots do 10 times more. before. Tasks that were unrealistic from before where you couldn't— it was unrealistic for you to classify all the content that you were making before. It was unrealistic to run things through every single check, through a 100-point checklist for every single article. That was unrealistic human work from before. But now that you're paying a fraction of a penny, anything is possible. So we'll leave it at that. I think that's it for today. Thank you all for spending your hour with me. Hope you have a great week and we will see you next week. We'll see you next week. Bye.
More from Eric
A 15-year-old just roasted the entire meeting #corporatelife #workhumor #genz
Most Founders Think Of AI Incorrectly, The Correct Mindset Of Agency Owners, AI Makes Crap Alone
