Stop Optimizing Keywords for ChatGPT (with Shane Tepper)

📬 You like this podcast? You’ll love the newsletter.
Join the weekly 3-2-1 on AI search + marketing: subscribe

Most brands trying to figure out their AI search visibility start with the same move: take their existing keyword list and ask which ones to "optimize for AI" too. Shane Tepper, co-founder of Resonate Labs, thinks that's backwards — and after this conversation, I do too.

This episode is part interview, part real-time reaction to a stretch of AI search news that hit fast and didn't slow down.

In this episode:

• Why starting a GEO strategy with keyword data instead of buyer language queries gets the entire approach backwards — and what query generation looks like when you build it around personas, buying stages, pain points, and features instead
• Google's recently released GEO guidelines, where the framing is defensive rather than informative, and what's missing from "good SEO practices apply to GEO" as a complete strategy
• Why optimizing for AI search isn't one strategy — it's a stack, because each model (ChatGPT, Gemini, Perplexity, Copilot) weighs credibility, authority, and source preference differently
• The agentic browsing audit quietly released through Chrome Canary's Lighthouse tool, what it actually checks for (WebMCP protocol, llms.txt), and why Google is sending mixed signals about whether these even matter for search visibility
• What a 150-query audit actually looks like in practice — and how response data turns into a content strategy, a CMS-routed publishing pipeline, and briefs for earning third-party citations
• Why third-party sources often shape AI answers without ever being cited — and what that means for brands measuring their own visibility

If you're listening to this and thinking I need someone to lead this for me, that's what I do.

I'm an AI search visibility consultant and a fractional content strategist for startups and enterprise brands. If that sounds like the kind of help you're looking for, email me at cassie@cassieclarkmarketing.com. 

Or request your 7-Day AI Search Visibility Audit: https://cassieclarkmarketing.com/ai-search-visibility-audit/

Let’s connect:

LinkedIn → Cassie Clark | AI Search Visibility Consultant
Website → https://cassieclarkmarketing.com

Transcript

SPEAKER_01 0:00

Hey, welcome back to Found in AI. I'm Cassie Clark, a fractional content strategist and an AI search optimization expert, and your host for all things GEO and AEO. Just a quick heads up before we dive into today's episode. There was no Thursday news episode last week. If you caught my episode on good SEOs, good GEO, you know I was technically out of office most of that week. I made a last-minute call Wednesday evening to just fully unplug. My brain could not, it could not go anymore. Which means that I spent the last few days playing catch up on everything that's happened while I was gone. And there's a lot. We've got a Google spam update to dig into, and honestly, who knows what else. I'm still digging through my inbox, and then there's always potential for something else to hit the internet between now and when that episode goes live on Thursday morning. Today's guest is Shane Tepper. He's the co-founder of Resonate Labs. And unlike a lot of guests who talk about one size of GEO, whether that's the Reddit strategy, schema, whatever, Shane's been thinking about this from query generation all the way through content deployment. So this one covers a lot of ground. Let's get into it.

SPEAKER_00 1:14

Hi, I'm Shane Tepper. I am the co-founder of Resonate Labs. We are a company that helps B2B companies understand their visibility status in AI search and then create and deploy the content to help improve their visibility status on AI Search across platforms like ChatGPT, Claude, Gemini, Perplexity.

SPEAKER_01 1:41

Are you looking into Copilot any? Ali asked this because Copilot is quickly becoming one of my favorites.

SPEAKER_00 1:48

Yeah, absolutely. Copilot is super important. A lot of large companies, a lot of enterprise companies have, you know, the Microsoft Suite natively integrated into their uh businesses. So Copilot is a very important engine for us to look at as well. And it's absolutely something we are uh looking forward to to integrating into how our system performs. Yes.

SPEAKER_01 2:10

And I'm not I'm not sure if it's just my internet these days, but it feels like cloud responses are so freaking slow. Copilot is like you hit the button and there's the answer. Have you noticed that?

SPEAKER_00 2:22

Yeah, absolutely. No, yeah, Claude is is my favorite LLM to use. It's my my daily driver, uh, so to speak. Um and yeah, uh it has had issues with with compute recently. Um it's it recently signed a deal. Anthropic recently signed a deal with um, I believe, um, I think it was SpaceX to borrow some of their compute to to help with some of these issues that you you just observed with the the slowness of the responses. Um so hopefully, hopefully that gets resolved. I've noticed personally that it has um uh been much faster in recent weeks than it was in the previous several months. Um, but yeah, Copilot you're saying is uh pretty quick, pretty instant response.

SPEAKER_01 3:06

Yeah, someone told me, oh, it's because everyone's using Claude more than they're using the other one. So maybe that's it too. But um I always like to ask anyone on the show, like, when did you really start paying attention to AI search and how to optimize for these engines?

SPEAKER_00 3:22

That's a great question. I was uh recently out of a job uh as a as a content director. Um, was out of a job, our uh our entire marketing team was laid off after an acquisition event. Uh so I was doing some freelance content work and I met up with a um a friend of mine who is a co-founder at a at a startup here in my hometown of Atlanta, um, met up to talk about a freelance content gig. And as we were talking, he mentions that a directive had come down from his CEO that they wanted to be more visible in Chat GPT when their buyers were asking questions about products in their category. And I was like, that's that's interesting. It's like SEO, but for for ChatGPT for generative engines. And I uh went back home and started looking into this and I was like, is this a category? And it turned out yes, absolutely it was a category. It was an emerging industry. Uh, there were a number of companies already trying to solve the problem. Um, and the more I looked into it, the more I realized how much opportunity there was there. Um, there are so many different ways you can approach this. Um, it requires a completely different strategy than a traditional keyword strategy that you would use for traditional search. Um, buyers ask much more highly contextualized questions than you know you would in Google. Google, you're talking about a three to seven word keyword search. Uh any more than that, your responses aren't going to be great. Um, whereas in LLMs, it's typically the opposite. The more context, the more specificity you provide. You know, I'm a CMO at a mid-market B2B SaaS company. I manage a team of 20, and we're looking for a uh an email automation platform that integrates with our with our current CRM, right? You're gonna get much better responses. Um and I thought that that was just a really interesting way in, so to speak. A lot of the platforms, a lot of the GEO platforms, uh SaaS companies today, they start kind of with the keyword data, with the traditional search keyword data, and that's how they develop the queries to optimize around and create and deploy content around. Um, I thought that there were there was a dearth of companies or approaches that start with the buyer language queries to optimize around. So that's that's how we build our practice is around those those natural buyer language queries.

SPEAKER_01 6:01

That's how I thought about it too. Like the keywords are great, but someone on the show many, many months ago really made the point of um think about the buyer persona, think about those questions that they are asking. I'm like, gosh, that makes a lot of sense. Like if you're thinking about the things they are actually typing in, that's what you need to optimize for. So it sounds like it sounds like you've been on this ride for a while now. So just within the last two and a half weeks, I guess, Google has made a lot of noise. I want your initial thoughts on all of this. I have mine, but tell me what yours are.

SPEAKER_00 6:42

Yeah. Um, first of all, I think it's really exciting. Google is acknowledging that you know GEO is a category um that people are paying attention to and and addressing. So that was that was my first my initial response when I was reading their the guidelines that they that they put out or the statement that they put out uh around GEO.

SPEAKER_01 7:03

I'm sorry, hold on, just a minute. My husband is outside and he's the callie.

SPEAKER_00 7:09

Oh yeah, yeah, go ahead.

SPEAKER_01 7:10

So we were talking about Google though. Yeah.

SPEAKER_00 7:13

Yes. So um I was saying that I think it's uh it's a good sign for for our industry that Google is is addressing AI search. Um Google's framing felt a little defensive to me, right? They were essentially saying that that uh GEO, AEO, AI search is SEO, basically, uh, and that the guidelines, their guidelines around SEO best practices, uh, and good SEO hygiene apply to GEO. Um, what I noticed was conspicuously absent was a denial that geo is is a thing. Uh and I will start by saying that um Google's absolutely right, that good SEO hygiene and practices are absolutely the foundation of geo as well. Um, you know, organized organized uh um content, uh proper schema, that all that stuff that uh benefits you in in traditional Google search is absolutely beneficial in uh AI search as well. However, there are uh a number of things that you should be doing for to optimize for GEO that go well beyond uh optimizing for traditional uh SEO. And saying something like optimizing for AI search um is akin to saying optimizing for social media, right? You have different strategies for different platforms, just like you would, you know, your Facebook platform, your Facebook strategy is obviously different than your LinkedIn strategy, it's different than your Instagram strategy. Uh, you the the different LLMs have different, let's call them, preferences for how they extract and surface uh content information to AI searchers. So you don't just optimize for AI search generally. Of course, like like SEO, there's a baseline set of practices that you do apply, but uh each of the individual models uh likes to surface different things, uh, you know, has has slightly different ways of determining credibility, authoritativeness, trustworthiness uh for content. Um they they have different um sources as well that they that they prefer to pull into answers. So you sort of have to uh apply um, you know, you sort of have to stack your strategies when you're when you're creating and and deploying content to optimize for those different platforms.

SPEAKER_01 9:41

So you noticed that the Google Geo guide was uh incomplete too. Is that what I'm hearing?

SPEAKER_00 9:48

Yeah, incomplete or um just uh like I said, like kind of defensive, like saying that you know, we're we're still you know, our guidelines are still relevant, you still need to practice this. Um but yeah, incomplete, incomplete in addressing the requirements for uh, you know, best best uh AI search practices.

SPEAKER_01 10:12

Yeah, I thought that too. Like it's interesting because from their perspective, AI mode, AI overviews, Gemini, it's all still search. But how those searches work is totally different than how traditional search performed in the past. Like that's a good way to say it, performed in the past because it's all changing very quickly. So they had their their GEO guide that dropped. Now they're also talking a lot about the agent first future. Do you have thoughts on that?

SPEAKER_00 10:43

I do. I think we are definitely moving toward a state where commerce, both on the B2B side and on the B2B side, B2C side, will be increasingly mediated in agent-to-agent through agent-to-agent interactions, right? The human completely removed from that. Um, and that will require uh a new set of strategies uh to optimize for. Um I think GEO is is the intermediate step between traditional search and our agent-to-agent mediated future. Um, you know, we're we are optimizing for machine readability, for machine extraction already, um, right, because there's there's a uh um, you know, that that generative process that's that sits in between what the human is asking and and the content that's uh indexed. Um and I think we'll we're just going to for better or for worse, we're going to um continue optimizing our content for for machine extraction. And I think probably the uh the human side of things is going to become less and less relevant. Um and it's just sort of going to become a war of creating content optimized for for agents, for crawlers, for bots.

SPEAKER_01 12:07

Yeah. And I to be honest, like I know I talk about this all the time on the show and like doing the work to figure out how these things work. I don't know how I feel about you know the human being removed entirely from this. Like, are you on the fence about it as much as I am, or are you just like gun home, let's go?

SPEAKER_00 12:24

No, I um, you know, I come from a from a creative background. Um, I was uh you know, an ad copywriter for a long time before I moved in-house um to to doing copy content at uh businesses. Um I've always, you know, at least at the beginning of my career, I, you know, I write you write for humans. You write compelling, uh, emotionally impactful copy. Uh, you know, you're trying to induce some sort of action uh in the in the reader, and this is uh a completely different um, you know, approach. Um, yeah, part of me definitely laments the fact that um, you know, that sort of uh writing is becoming less and less important uh and valued. Um it's one of the major reasons I sort of left, you know, advertising because I saw uh that that uh you know the the the quality of the of the writing was was becoming less valued and it was becoming more about uh you know conversion, which like you know makes sense. We're right, you know, we're writing on behalf of businesses, but um it's funny you mentioned that I was like looking through um an old um life magazine from the 50s the other day, and you see all these ads, you know, for for alcohol brands, for cars, for household products, and the ads have like, you know, there's like a paragraph, a couple paragraphs of like really snappy copy, and the headlines are really nice, and that's just not a thing anymore. And yeah, I think, you know, it's just it's uh it's a shame in some ways, but like, you know, just like the people who rail against AI generally, um, you know, it's it's it's a it's a fruitless thing to do because that is how things work now, um, regardless of of how you feel about it. Um, and we all, you know, are are best served. Well, I don't know about best served, I don't know if that's the best way to say that. But we all are uh we all um you know uh should be should be preparing for this future and and respond to it uh you know in a manner that that that uh is appropriate for how things are are changing and where where things are going.

SPEAKER_01 14:43

Yeah. So I didn't think that I mean Google has been saying for months like the future is agentic, and I get that from like the AI search standpoint, but the gentic commerce piece of this is the part that I thought like oh I think that'll be forever away. I don't think it is now because I think we're moving quicker to agentic commerce than than I thought we were. And I like like you said, like the GEO part is the piece that kind of connects it all together. So if you're having conversations with any of your clients, like how are you explaining that? Like, is anyone asking about addendic commerce yet and how to be found by the AI bot so that something is done?

SPEAKER_00 15:26

Um no one has really asked about agent-to-agent commerce yet and how we're addressing it. I think they are they are because AI mediated, human AI mediated uh search uh and commerce is still something that most businesses are wrapping their heads around, right? That's where that's where they are. We as people who work in the industry and are reading, you know, work, you know, for the past year plus, I've been consuming everything I can about uh you know about what's happening in this space. Um marketers, uh even, you know, even high-level marketers at companies, I think they're mostly thinking about now, right? They're thinking about this quarter, they're thinking about this fiscal year, um, they're thinking about this current problem. So um, no, that is not something that has come up in in uh conversations with with prospects, with leads yet, but it's absolutely something that is being discussed in the the broader GEO community.

SPEAKER_01 16:23

Right, because um the identific browsing audit was just released with um Lighthouse last week, maybe? I'm a little bit off on my days. Um, but it's one of those things where if you start your audit and Google doesn't actually tell you whether or not, have you looked at that yet? Let me start this.

SPEAKER_00 16:42

No, no, in fact, I think to say, could you send that to me after our conversation?

SPEAKER_01 16:46

Yeah, so and for the listeners, you have to have Chrome Canary to be able to go in there and see this audit. Um, but you find it in the developer tools and then you click on it, and whatever website you're looking at, it'll give you um a general score, not out of a hundred, but just kind of like a readiness score. If your website passes one, well, they have three different checks that they're looking for, and it'll say, Oh, your website is one out of three, so you're not yet ready. Um, but it's looking for things like their web MCP protocol, which is kind of a proposed standard at this point, and it's not an actual thing. It's also looking for that llms.txt file that Google has been saying you don't need for search visibility, but evidently you need fordentic browsing. So is that news to that's news to you, right? Yeah, this just happened within the last today's search day, so within the last week, I guess. But um, yeah, so that's that's the thing that they're pushing for. And I I find it frustrating that it feels like Google is talking out of both sides of their mouth here. Like for AI search optimization, you don't need the LLMS.txt file, it's all still a quote SEO, do whatever. But then you have the agent browsing side that's different. They're like giving different guidance to how you get found if you're a commerce, which still applies to B2B or B2C, whichever one. Um, I'm finding that frustrating. I don't know about you.

SPEAKER_00 18:22

No, yeah, it seems like, yeah, like you said, they're they're trying to, you know, I guess taking this scatter shot approach to make, you know, just have have all your have all your bases covered. However, it's all still SEO.

SPEAKER_01 18:35

So yeah, it's confusing. But then at the same time, like Google is not the only AI engine that we have to worry about because we also have Chat GPT and Claude and all of the other ones. Um yeah, so anyway, that that's where I'm sitting. So I noticed that when you do an audit, you're looking at 150 different queries. Can you tell us what that's about?

SPEAKER_00 19:01

Sure. Yeah, absolutely. Uh we have a broad set of queries that we generate to optimize for. They are informed by uh several dimensions. Um, the first one being um personas, right? For for B2B, uh you you typically have a buying committee, it's called, right, within a within a business. It's not just one person making the decision, it's multiple people, multiple roles within an organization that have multiple uh or that have varying concerns. Um, you know, uh like the CFO is gonna be concerned about ROI, you know, the head of IT is gonna be more concerned about how it integrates with their existing tech stack. Um, so you have to have queries for all of these personas covering all these personas. There's also the buying stages, right? From problem identification all the way down to comparison and decision making. You want to have um queries that represent all phases of those um buying stages or of the buying journey, rather. Um you have pain points, right? Different, as I just mentioned, different personas are gonna be servicing different pains that they're experiencing. Uh, and then you have the the features that the different platforms have, right? They're gonna be asking about different features based on their particular um use case. So those are sort of the dimensions, that's the matrix uh across which we generate these these these queries to basically cover you know any any potential question or at least question category that uh a buyer might ask while comparing options in a in a um in a for for for a particular for a particular category of solutions.

SPEAKER_01 20:42

So pretend I don't pretend we, I say this as an audience, have never heard anything about this before. What are you doing after you have all of these questions? Like how are you using that for your content?

SPEAKER_00 20:54

Yeah, so we take we take those those queries, we run them through uh our models. Um right now we're doing Chat GPT Perplexity, Gemini, and Claude. Uh, you get a ton of response data, and then we have built some software that parses all of this, uh, all of this response data and under so we understand um how our our clients are performing against their competitors uh for different personas across the different stages, uh, when buyers are servicing different uh pains that they're experiencing when they're sort of when they're looking at particular feature features. Um and you know, we look at the citation sources, right? At that point in time, what are the sources that the models are are are pulling in to synthesize their answers? Um, and we sort of give them using all those different elements, slice and dice, and give them uh you know a view into their status quo visibility situation. Um, and then what we do is based on all of that information, based on how we're seeing that um the model. Are you know answering these buyer queries? Um, we develop a content strategy. So, you know, uh you are losing against competitor X on this particular feature or addressing this particular pain point. So uh create content, create and deploy content uh around, you know, comparative content around um uh that that particular area of focus. Uh our system creates content that is optimized for extraction as well. Um and um and then you know, we we integrate uh we create a system that integrates with our clients' uh CMSs. Um, you know, each client has a particular review process as well. So if it needs to go to their legal team, if it needs to go to their content team, we can uh route that content to them. Um and then and then they uh you know either either they trust us to directly upload it to their CMS or they they upload it to their CMS. We also have um generate briefs that make recommendations for third-party um attribution, right? So that's a that's a very important thing across LLNs. It's not just about what you say about your own brand. Uh it's about you know being validated by external third parties as well. So if um you know we're seeing that a particular industry journal, for instance, is being referenced uh a lot in the response data, we will be like, we'll generate a brief that says, you know, make a make a pitch to this publication about this topic uh and try and get get published on on their um you know their website as well. So it's kind of like the combination of the on-domain and the off-domain stuff. So it's pretty, it's a pretty comprehensive system uh that you know, like I said, takes into account uh all of those different dimensions that I enumerated earlier um to create this this action plan for how to how to remediate your your visibility situation.

SPEAKER_01 24:05

Okay, so let's land this plane. We've got Google's DEO guidelines, the Gentec browsing audit, and Shane's breakdown of how a 150-query audit actually turns into a content strategy. If you take only one thing from this episode and I get this, go look at how you're generating the questions you're optimizing for. Most teams are still starting from keyword data, and while that's not bad, it's kind of missing part of the picture. Shane's point and one that I strongly agree with is that your buyer language and keyword data are not the same thing. So this week, take some time, pull up your top three competitor comparison pages or your most cited support docs or your most cited page, whatever that is, and then ask yourself are these answering questions like what someone would actually type into Chat GPT or what they would type into Google five years ago? If the answer is ancient Google, then that's where we need to start and we need to start updating those pages for AI search optimization. That doesn't mean just yank all your keywords out, but it does mean think about how it answers the buyer question. That's it for this week. I'm Cassie Clark. If you want help writing that kind of audit on your own brand, you know where to find me at Cassie Clark Marketing.com. There's also more information in the show notes. Until next time, stay visible.