In-Ear Insights: Competitive GEO - Generative Engine Optimization FAQ

In this episode of In-Ear Insights, the Trust Insights podcast, Katie and Chris discuss the evolving landscape of Generative Engine Optimization and how it reshapes modern marketing. You’ll learn why traditional metrics like share of voice and domain authority fail to provide accurate insights within AI models. You’ll discover how to optimize your content strategy for semantic relevance to ensure search engines recognize your brand. You’ll gain practical knowledge about preparing your digital presence for the future rise of autonomous AI agents. You’ll uncover the truth behind the “alligator chart” and what it reveals about your current search performance.

00:00 – Introduction
03:15 – Debunking AI share of voice and domain authority
07:45 – Why listicles outperform single-vendor sites
12:20 – Managing AI crawlers and web governance
16:40 – Tracking presence in Gemini versus Claude
21:10 – The future of agentic SEO and LLMs.txt
25:30 – Call to action

Watch the full episode now to future-proof your digital presence and master the next phase of search.

Watch the video here:

In-Ear Insights: Competitive GEO - Generative Engine Optimization FAQ

Watch this video on YouTube

Can’t see anything? Watch it on YouTube here.

Listen to the audio here:

Download the MP3 audio here.

[podcastsponsor]

Machine-Generated Transcript

What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for listening to the episode.

Christopher S. Penn:
In this week’s In-Ear Insights, we’re talking about GEO 201. We recently launched our new course on competitive GEO and had a webinar about it. As is the case with everything, there were more questions than we had time for. Katie, you wanted to go through your choice of these to figure out what else people need to know that we didn’t have time for.

Katie Robbert:
I’m really excited about this new course because it takes GEO to the next level. GEO 101 was really the explainer and the foundation of good GEO. In 201, the next natural question is: what is the guy down the street doing? We walk through that competitive analysis and give you the tools to create a scorecard of what it looks like for you against your competitors. Some questions keep coming up about GEO, especially regarding competitive analysis. The first question everyone is asking is how do you actually measure share of voice in tools like ChatGPT, Gemini, or Claude when the answer is different every single time you run the prompt? This is a 101 and 201 question. It’s the question that persists and will not stop being asked until someone gets a satisfactory answer.

Christopher S. Penn:
They’re not going to get one because the answer is you can’t. Period. End of story. These are probabilistic tools by definition. They are probability-based. We did this in the 201 webinar—I showed a prompt asking for dinner recommendations in Boston, and five seconds apart, in new chats, I got totally different answers. This is the part that drives me up a wall. All these software companies saying we can measure your brand in ChatGPT are full of it. You cannot. There is no way to do it. I apologize that this question keeps coming up. The answer is not satisfactory, but you can’t answer it.

Katie Robbert:
I think that’s something people need to keep hearing. If you can’t measure share of voice, what can you measure? That’s what we go through in the 101 and 201 courses. Anyone who knows us knows that we are all about measuring the work you’re doing, because, Chris, you say data without decision is just distraction. We want to make sure you’re able to make decisions. If you’re trying to find out your share of voice, that might be the wrong way to be thinking about it. The next question is: let’s say a competitor with way less domain authority is somehow dominating Perplexity or Gemini recommendations. What’s actually happening under the hood? We need to start with: is domain authority even still relevant?

Christopher S. Penn:
Domain authority has never been a great measure because service providers have always had to basically guess what a site’s domain authority is. Google does not have such a measure. If you go back to the summer of 2024 when Google’s Content Warehouse API leaked, there were over 200 different ranking signals in Google at the time. Domain authority is a proxy that came up with Moz around 2010 as a proxy for how likely it is you will show up in search results. This goes back to GEO 101: Does the model know you exist? What’s the likelihood that the model will bring you up in a traditional web search? And once you’re in the consideration set, how likely is it that the model will actually use your content as part of the answer and surface you? When someone does a search in Perplexity, ChatGPT, Gemini, or Claude, you pass through all three of these gates. The change today for marketers is that any one of those gates can stop you, but you don’t necessarily know which one. That’s what we cover in GEO 201. When we look at the measures in our templates and worksheets, you can see where the stopping gate might be. If you don’t measure the things that are proxies for those gates, you have no way of knowing where you’re falling out; you just know that you’re not in the results.

Katie Robbert:
So we’ve debunked two metrics so far: share of voice and domain authority. Sorry to everyone if you were hanging your hat on those. The next question we got from a couple of participants is: why do AI models keep recommending top 10 listicles from publishers instead of the actual vendor sites? How are real brands supposed to compete against aggregator content?

Christopher S. Penn:
Because aggregator content is more semantically relevant. Semantically relevant means when you type something into today’s search engines, it’s looking at the context of the query. If I say “AI consulting” and Trust Insights is in the consideration set, no matter how long our page is, we’re still just one vendor. If you go to a listicle that has 82 vendors and a paragraph about each, you’re going to cover a much wider net of all the different ways to talk about this space. The more semantically diverse your content, the more likely it is to be seen by these models as helpful. If I’m just talking about Trust Insights all day, that is semantically very narrow. If I write a listicle that contains a wide spread of vendor names, concepts, and things, it will do better. People talk about needing FAQ schema, but you’re missing the point. If those are real questions, they probably have a much wider footprint of unique language. At the heart of all these things are language models. If the language you use is very narrow, it will by default be seen as less useful than a wider language base.

Katie Robbert:
This sounds like GEO is just the next evolution of SEO. It sounds very similar to when you first taught me about SEO—you have your anchor topic and then all the related topics around it. It’s not enough to say “here is Trust Insights, we are the best.” It’s almost like the Hero-Hub-Help framework is still relevant. Do we have content that explores those services from every angle—what is it, why is it important, who is it for, what happens when you don’t do it? You’re just layering another piece of software on top of good SEO.

Christopher S. Penn:
Best practices and specifics really matter. I recently put some content through a semantic engine I wrote, and it said the person writes like AI—they are generally lacking specifics and proper nouns. My LinkedIn posts don’t get flagged because they have so many weird specifics in them that it’s semantically a much wider footprint. If you say Trust Insights does analytics and data science, and we help you do things like moving average convergence divergence indicators or uplift modeling, we’re getting into detailed specifics that cast a wide semantic footprint. Listicles work not because they are listicles, but because they are wide semantic footprints.

Katie Robbert:
It strikes me there’s going to be a challenge for writers trained a certain way. Technical copywriters were trained to be uniform and sound the same for large companies, which is a narrow lens. If you’re trained in your career to write that way, it’s hard to be told you’re doing it wrong. I’ve been dinged for my writing style, like over-explaining in sentences or using em dashes, which AI also does. Everyone assumes you’re writing with AI. If you want to stay competitive, you have to look at your content strategy.

Christopher S. Penn:
Think about what marketers have been told for 30 years: don’t use jargon, make your content accessible. Guess who writes like that? AI. Guess how AI learned to write like that? From marketers.

Katie Robbert:
All right, next question: if a competitor is blocking GPTBot and ClaudeBot at the crawler level, is that an opening for us, or is it working in their favor?

Christopher S. Penn:
We ran into this with Boston Consulting Group, who block a substantial number of AI bots. There are two classes: training bots, which gather information, and user bots, which perform searches to answer questions. I advise clients not to block either. If you’re a content shop that makes money on your content, block the training bots, but don’t block the answer bots. If a competitor is blocking them, that’s an opportunity for you, provided you aren’t blocking them yourself. Some crawlers, like Perplexity, don’t respect those rules anyway. At a bare minimum, talk to your IT team and have some governance around AI and your website instead of just letting one group wing it.

Katie Robbert:
If our brand is 97% present in Gemini but barely 16% in Claude, what does that pattern tell us about how each model is being trained?

Christopher S. Penn:
I want to throttle people who ask questions like this because they’re getting it from an SEO tool that is making it up. Unless you’re spending tens of thousands of dollars reverse-engineering the embeddings of these models, there is no credible evidence that those percentages are grounded in reality. If you want to know if Gemini is recommending you more than Claude, look in your web analytics. We have a guide on the Trust Insights website on how to set up Google Analytics to track this. Google Analytics and Google Search Console recently announced support for tracking generative AI for Google specifically. I mentioned in the 201 webinar that if you are a B2B company and you are not looking in Bing Webmaster Tools at the grounding queries Copilot is using, you’ve missed the boat.

Katie Robbert:
I feel like Bing Webmaster Tools has been an underrated tool for a long time.

Christopher S. Penn:
The number one thing Trust Insights is known for in the land of Copilot is your content on Claude.

Katie Robbert:
I actually didn’t know that. That’s pretty cool.

Christopher S. Penn:
Don’t worry about the “I’m presenting this” thing. Look at the data you’re actually given. Katie has been doing a ton of work with Claude, and in a few weeks, for Microsoft Copilot subscribers, a licensed version of Claude is going to be in Copilot. All the content Katie’s done is going to be part of that ecosystem.

Katie Robbert:
Get used to my face. My leadership team only looks at Google Rankings. How does a course like 201 help me change the conversation?

Christopher S. Penn:
SEO is part of GEO—it’s phase two: presence, appearance, relevance. Google Rankings is part of that, and you shouldn’t lose sight of it, especially since generative AI queries are coming to Search Console. But it’s not the whole story. In the 201 course, we break down the three different phases and the metrics for each. You can introduce the presence and relevance portions to your leadership to broaden the conversation.

Katie Robbert:
The 201 course provides a scorecard. The question is: since there’s no search console for ChatGPT or Gemini, how is this scorecard not just guesswork?

Christopher S. Penn:
We’ve been working in AI since 2013; we know the technology that creates these models. We can’t know what someone is typing into a model, but we can know what the model was fed. The scorecards aren’t guesswork because we’re looking at things like media coverage, which is part of the first phase. SEO isn’t guesswork if you’re using good tools, and retrieval isn’t guesswork. We wrote tools to measure this, and the next version of AI-V will have even more measurements to validate that companies are complying with how AI companies want you to behave if you want to be recommended.

Katie Robbert:
If you haven’t checked it out, we have a free version at Trust Insights AI-V where you can test one page and a competitor. One of the questions that came up is: is LLMs.txt actually worth implementing, or is it similar to the Robots.txt hype cycle?

Christopher S. Penn:
Up until May, it was purely hype. Then Google updated their Lighthouse documentation to say they are looking at LLMs.txt, not for ranking, but for agentic tasks. The new Google Lighthouse looks at LLMs.txt and WebMCP to see if a site is set up for agents. Gemini Spark is an agentic platform. GEO 101 and 201 are about humans using AI; GEO 301 will be about AI using AI. Are your properties set up for AI agents to use your website without a human? I’m currently trying to figure out how to implement WebMCP on the Trust Insights website so an agent can programmatically subscribe to our newsletters or YouTube channel.

Katie Robbert:
Definitely check out our past episodes about the legality and logistics of what that means.

Christopher S. Penn:
There are now about five different agentic SEO things you should be doing: WebMCP, LLMs.txt, discoverability, accessibility for agents, and layout stability. There is a whole brand-new frontier coming to SEO.

Katie Robbert:
Many folks are convinced that competitors with Wikipedia pages have an unbeatable lead. Is that real?

Christopher S. Penn:
Wikipedia is in the training data, but it’s relatively small—boiled down, it’s about nine gigabytes of text. That’s a bookshelf of Shakespeare. It’s in there, but it’s not an unbeatable lead. What makes a company do well in GEO is having rabid fans or rabid enemies—people who create mass quantities of content about you without you. If you’re just “fine,” that’s where you’re in trouble.

Katie Robbert:
When your CMO says “we still rank on Google, we’re fine,” what is the one thing you would put in front of them?

Christopher S. Penn:
The alligator. In SEO circles, this describes the chart in Search Console where impressions are going up but clicks are going down. The search box is now AI mode—it’s not the 10 blue links anymore. Your ranking is meaningless in a conversational space. If you look at that chart, everyone has the alligator except for companies that have completely cratered.

Katie Robbert:
GEO 201 is now live. Reach out if you want a version of our webinar replay or to get started with GEO 101. We have a bundle available if you want both courses.

Christopher S. Penn:
If you have thoughts on GEO 101 or 201, pop by our free Slack at trustinsights.ai/analytics-for-marketers, where over 4,700 other marketers are asking these questions every day. Thanks for tuning in.

Need help with your marketing AI and analytics?

Leave a Reply Cancel reply

Pin It on Pinterest