How to Rank #1 in AI: 5 steps | The Ultimate Guide on GEO

Speaker A: This episode is brought to you by Accenture. When your advertising operations fall out of sync, everything else follows. Spotify and Accenture are working together to reinvent the rhythm of ad sales, using automation, analytics and smarter workflows to simplify campaign delivery and access better data across the business. The result? Less time spent on operations, more time connecting brands with the moments and fandoms, um, that matter most. Learn more@accenture.com Spotify

Speaker B: when you need to build up your team to handle the growing chaos at work, use Indeed Sponsored Jobs. It gives your job post the boost it needs to be seen and helps reach people with the right skills, certifications, and more. Spend less time searching and more time actually interviewing candidates who check all your boxes. Listeners of this show will get a $75 sponsored job credit@ Indeed.com podcast. That's Indeed.com podcast. Terms and conditions apply. Need a hiring hero? This is a job for Indeed.

Speaker A: Indeed.

Speaker B: Sponsored Jobs the biggest threat to you

Speaker C: and your business right now isn't AI ah replacing you. It's AI ignoring you. Let me share a story that just happened. So a chatbot almost cost Ryan Serhant, real estate agent in New York, a $50 million penthouse deal. What happened is that the buyer asked Chad GPT is 50 million too much? And Chad GPT said yes. So the broker called to pull out at the same time. The seller asked the same chatgpt and it told them to hold out for more. Both sides almost walked over one chatbot answer. And this is happening with your customers every single day before they ever reach your website, your blog, whatever you're promoting. They're asking ChatGPT Cloud of perplexity, what should I buy? Who should I hire? Which company solves this? Who's the best creator in this niche? The AI doesn't hand them 10 links anymore. It hands them three names, maybe five. But if you're not one of them, the sale is over before it starts. And the data shows that 35% of people in the US now start shopping inside an AI assistant instead of a search bar. This is crazy. The conversion is up 693% in a year. And just listen to this number because we actually saw it in my newsletter. People coming from ChatGPT convert at 15.9% when people coming from organic Google convert at 1.76%. It wasn't like that two years ago. It wasn't like that a year ago. It is happening right now and you have to pay attention. So as you know, I run this podcast, Silicon Valley girl. I interview CEOs and leaders and we had some pretty darn amazing guests. We had the CEO of LinkedIn, we had the founder of LinkedIn twice. We had the CEO of GitHub, founder of Perplexity, Mustafa Suleiman from Microsoft. Like pretty good roster, right? And even though we had famous names and all of them are related to AI, when we asked ChatGPT for the best AI, ah, podcast to listen to or like just give me a few a podcast, we were never ever on the list and knowing how people convert when ChatGPT recommends them something and we decided to run a full GEO experiment. GEO stands for Generative Engine Optimization. Some people call it aeo. It's basically the same optimizing whatever you are producing online your content, your website for those chatbots. And uh, we've run this experiment for a few months and based on our queries we became the fifth most visible of the 11 podcast brands the models track and our uh, AI visibility doubled. So of course I have to share everything with you. In this video I'll show you exactly what we changed, which tools we used and what happened to our visibility. I'm going to describe everything step by step so you can copy. I'm going to go in the exact order we did it so that you take the system and make your own business or podcast or whatever you're working on. Show up in the AI age and the real opportunity is that most businesses are not doing this yet. So you have this first mover advantage. 92% of marketers say they plan to optimize for AI search. Only 40% are actually doing that. The gap is your head start. And if you want to build the systems with me, start, subscribe. Because I share everything. I tried myself, what worked, what didn't and the actual numbers. So step number one, find out if AI can even read you. A few months ago I was just testing something simple. I asked a few chatbots to recommend Silicon Valley podcasts and my show didn't come up. Despite the name Silicon Valley Girl. Come on. Not on perplexity, not on ChatGPT, not on Gemini. After 50 CEO interviews and the models acted like the show did not exist. Isn't that wild? All that content was completely invisible. So I did one thing that started this one month long rebuild and it's the first thing I want you to do. My team sent our website URL to Claude with one question. How visible are we in AI search and why aren't we showing up? Claude came back with a very specific list. The HTML was missing the parameters AI crawlers need to read a page. The transcripts or podcasts were hidden behind JavaScript. So the crawlers just saw an empty shell, by the way, by the site. I mean, we just have a website where I host all the information about all my social media. Of course we have YouTube, but we can't really do a lot around its infrastructure. Uh, but we're going to talk about YouTube as well. But basically the website that I have, marinamogilgo co, looked totally fine. To a human, to an AI reading the source code, we were basically, um, blank page. So before you watch the rest of this video, put this on pause, take your URL, paste it into whatever chatbot you're using, and ask it how visible you are in AI search and what's stopping you from showing up. You'll get back a list of fixes specific to your site. And again, any insights on this? Uh, I'm waiting to read your comments because we're still working on this problem. So this list is basically your roadmap, your step number two, you build a home. The crawlers can read. So level one was giving the crawlers something they could actually read. We didn't even have a real podcast site, so we wipe coded one from scratch again, super easy. These days we use Claude code. You describe what you want in plain English and it writes the actual code for you. Then that side has to live somewhere on the Internet. So we put it on Vercel. Vercel is the host. It's the place your website actually sits. Backend is basically the part nobody sees. Ours keeps all raw episode data in one place. Every transcript, every guest name, every summary railway is where that engine runs. So when we add a new episode, its page builds itself from that data, instead of me copying and pasting the same layout into 85 different pages by hand. So now we have this site podcast, marinamagilco Co, where every single episode has its own crawlable URL with a full episode, transcript summary and guest profile. Google indexed that homepage within four weeks. So it's a long term task. It's not going to happen overnight. Then we of course made every page machine readable. And this is the part most people skip. Every page is fully built and just sitting there, all the text already on it before anyone opens it. Nothing has to load or assemble first. So when the AI crawlers come by, the little bots that ChatGPT, Perplexity and Claude send out to read the web, the whole page is right there for them instantly. Nothing to wait for, nothing to click. We built 85 plus episode pages this way. The same setup is used by Huberman Lab and Lex Friedman. Again, it's very important that the full transcript sits right there in plain text without you having to click to open anything. That move adds 8,000 to 15,000 words per episode. The AI can now read and quote back with our name on it. And honestly, if you do nothing else from this video, do that part, there is a real signal behind it. And now step number three, you gotta fix your identity in the machine. So the models don't really decide what your business is or what your social media Persona is just by reading your homepage. They read a few structured databases that label you and then they trust the label. Get the label wrong and you get filed in the wrong. So number one, Wikidata, the machine readable record that tells AI what kind of thing or person you are. So this is something that moved everything. The most Wikidata sits behind Google's Knowledge panel and feeds Gemini. Every entity has a field called instance of which is just what type of thing it is. Mine said a vlogger and a YouTuber, because that's what I was, uh, 10 years ago and 12 years ago when I, when I started Gemini, read that field and called Silicon Valley Girl a vlog. And a vlog never gets pulled into a podcast answer. So we changed the occupation to podcast host, entrepreneur and angel investor, added the podcast as its own entry and translated it into 11 languages. So the same fact holds, no matter which language the model replies in, uh, it took us a few hours. You have to do it as well. This is where the models actually read your category, not just where people listen. This part is for creators who are present on podcasts, but also for founders. You know, podcasting is a big part of your marketing strategy. So when we pulled the sources of the models cited in a category, Apple was about half of them. Half. So we rewrote the show description and the category fields to say the exact phrase, wouldn't Repeat it back. AI, tech and career growth, interviews with CEOs. Same words we put into the wiki. Data. Occupation, Spotify. Same description. Second source for cross checking. By the way, the geo mistake I was making is that I was, um, writing a fresh bio on every platform because it just felt nicer. That was actually a mistake. So now we rewrote my Spotify bio to match Apple and Vikidata word for word on the phrases that matter. When the sources disagree, the models play safe and name someone else. When all three say the same Sentence. It repeats your sentence back the principle under all of it, whatever the AI thinks you are, go to every place it reads your site, Wikidata, Apple, Spotify, and make them say one identical, accurate sentence. This is the cheapest step on the list, and also one of the strongest. Now, step number four is plumbing. The way search and AI tools find you is that they don't magically know your page extension exists. A crawler has to discover each page, read it, and add it to an index. If a page never makes it to that index, no search engine and no AI can ever recommend it. It's invisible by default. So we really needed to get those pages into the indexes. So here's how you can do it. First, Google Search Console. This is like a direct line to Google. You hand it a sitemap, which is one file that lists every page on your site so Google doesn't have to stumble onto one of them at a time. Then it shows you exactly which pages Google found, which ones it couldn't read, and what's broken. We submitted our sitemap, and Google discovered 91 pages it hadn't indexed before. The result from this boring hour worked around 700 clicks a month on the main site that simply weren't there before, plus 91 pages the engines can now find and quote. The lesson here is that before anyone can recommend you, a machine has to find your page and file it away. The next step, of course, you have to measure it. If you don't have the numbers, if you don't know how things are moving, you can't really steer your airplane. We track all of our work through an app called Peak AI P E E C A, I. It checks the three engines we have turned on. Google, Google AI Overview, ChatGPT, and Perplexity. And it tells me for every question whether we show up and who's beating us. By the way. We did our research. We looked at different eight tools. Peak, uh, covered the most engines at the entry price, so we stayed. And also, Peak has this tactic that's worth the whole subscription, in my opinion. It shows you the exact keywords a model searches for internally while it's answering questions. So we just started writing those exact phrases into our episode titles. Yes, that straightforward. The first time we confirmed it was our, uh, ryan Roslonsky x LinkedIn CEO episode. We titled it to match what model searches for the LinkedIn CEO podcast. And we started winning that answer. Isn't that nice? It's now a repeatable pattern across five different guest questions. Check every title against what people are actually typing into AI. And by the way, this is a long term work that never stops. So this is not final, but let me show you before and after. I think, um, it's just going to inspire you to continue this work. When we started, we showed up in about six, seven of uh, 48 tracked questions. Masters of scale. Reid Hoffman's podcast was in 21. All in of course 19. Diary of a CEO in 15. We were nowhere near them. As of last week, Silicon Valley Girl is the fifth most visible by AI visibility of 11 podcast brands we track, ahead of TVPN, Gary Vee and Mel Robbins of course. Full disclaimer. We're only tracking a few keywords that we think matter to us. We're going to add more very soon and we're going to go low in rankings again. But we are now just seeing how we're rising in the ones that we're tracking, which means there are a few more that we're not tracking. Our, uh, overall AI search visibility doubled over the time we worked on all of this. There are some specific questions that we actually own. And by the way, of course every answer is personalized, but I don't know if you ask AI for podcasts with Yossi Matias, we come up 34% of the time. Often as the only show it names it's head of Google Research, which is amazing, right? Also try Best podcast for women interested in AI and startups. 19% is US perplexity CEO podcast 31%. Although Aravind just did another podcast with 20 VC. I wonder if that messes up the results. But uh, as I said, it's an evolving, ever changing game, but I, uh, love playing it. I also want to be honest about where we still lose the really broad ones. Like of course, best AI podcast 2026. We don't win them. Those get answered from big listicles where competitors get mentioned way more often than we do and we haven't been cited much. We did get named in a Forbes listicle in April and that's kind of a thing that moves the broad questions. So the work is still ongoing. I think my next step, well, first of all, stepping up my game, asking better questions, getting more views, getting more relevant. But also I think I will start working with a publicist. I don't know if you follow me on LinkedIn, but I shared my PR nightmare story that happened a year ago where it didn't pay off at all. But now I have a person who I trust, so, uh, maybe we'll Try that. Anyways, the work is still going. I'll share. If this video does well, I'll do another video on do. And I will definitely keep sharing about our Geo AEO journey in my newsletter called Future Proof. It's free. The link is in the description. So here's your clear roadmap, the same order we did it. 1. You ask AI to read your site and tell you why you're invisible. Do it right now. 2. Give the crawlers something they can actually read. Static HTML real transcripts in the source. We were invisible for one boring technical reason and fixing it was the single highest leverage thing we did. 3. Fix your identity everywhere so it says the same accurate thing. From wiki data to your platform bios to your to your social media, anywhere you show up. 4. Match your titles and headlines to the questions people actually ask the AI because it's the biggest mistake, the way we try to formulate thoughts and the way people actually think. And 5. Pick one tracker and watch it weekly and read the trend over weeks, not days. Don't rush it because a single week doesn't really make a difference. As I said to Google, four weeks to identify us. And here's the honest truth as well. Why I mentioned PR and a publicist. 82% of what AI cites is earned media. Earned media means other people talking about you, not you talking about yourself. An article that mentions you, a podcast listicle you got named in or like top entrepreneurs, a journalist who quoted you, a Reddit thread about you. You don't pay for it and you didn't post it, someone else did. And that's what the AI trusts the most. Because anyone can say they're the best on their own website. If you're like, okay Marina, that was great. Can I have this plan and do it later? Yes, you can. Uh, it's in my Future Proof newsletter. The link is in the description. Subscribe and you'll get the 30 day plan that we've created for ourselves and for you for free. I hope this video was useful. If it was, please thumbs up because uh, I haven't done these tutorials in a while. I want to see if you're liking them and if they're most important, if they're useful and if you're applying them. So please let me know down in the comments if it's useful. Please give this a thumbs up and if you can talk about this podcast on your social media, I'll be so, so grateful. Thank you so much guys. And I hope this information and video was useful.

Speaker D: You know those tiny back to school emergencies that somehow become your problem. That's why I love UberEats. You can order school supplies, snacks and lunchbox essentials for $5 or less. M so when your kid casually drops I don't like peanut butter anymore. Or I need five green highlighters for a project due tomorrow, Uber Eats has you covered. Get everything you need for back to school today from your favorite brands like Aldi and Staples on Uber Eats. Order now ends 97$5 or less before taxes and fees. Select items only. Availability varies. See app for details.

Speaker E: Close your eyes, exhale, feel your body relax and let go of whatever you're carry. Well, I'm letting go of the worry that I wouldn't get my new contacts in time for this class. I got them delivered free from 1-800-contacts. Oh my gosh, they're so fast. And breathe. Oh, sorry. I almost couldn't breathe when I saw the discount they gave me on my first order. Oh, sorry. Namaste.

Speaker C: Visit 1-800-contacts.com today to save on your first order. 1-800-contacts.

Speaker D: Bye.