Ross Hudgens, founder and CEO of Siege Media and author of the new Wiley book Generative Engine Optimization: The Definitive Guide to AI SEO, joins Greg Sterling and David Mihm (Mike Blumenthal is out) to argue over the label, then get practical: which engines are worth tracking, where AI answers get their evidence, and how a business earns the recommendation.

Ross Hudgens' Current State vs. Ideal State

Ross's recurring distinction (15:38, 39:42): what the models cite today is a snapshot, not a strategy. His top "what marketers get wrong" answer is the same idea: they aim for the current state instead of influence.

Takeaways

• Local is where SEO and GEO diverge least. Google has already compiled the review context that models are hunting for. Nationally, the models go off-site to Reddit, affiliates, LinkedIn and YouTube, which Ross argues were once indirect ranking inputs and are now direct ones. A niche D2C brand that could never outrank Amazon can still get recommended through its reviews and listings. (10:08–11:47)

• Drop Perplexity from your default tracker. Many tracking tools default to it, so an unweighted tracker can push you to chase a listing that looks like 25% of the impact but delivers under 1%. For the average local business, Ross sees ChatGPT as the only relevant LLM besides Google's models. (13:08–14:17)

• Claude is the B2B surprise — and it searches through Brave. Ross's own open-ended lead form surfaced Claude "out of nowhere." Since Claude's web search runs on Brave, your Brave rankings are a cheap proxy when full tracking is too expensive. David's read of Similarweb data: ChatGPT has stalled while Gemini and Claude are the only engines still growing. (14:17–15:22, 34:20–35:34) [1]

• Missing third-party citations are an opening, not a destiny. In year one, Ross saw homepages, competitors and similar sites taking ~60% of citations. He reads that as a gap in third-party reviews that the models would fill if the reviews existed. (15:38–17:01)

• YouTube is the most-cited single domain for consumer queries. Siege rarely makes the videos itself. It pays affiliates and influencers to place the client "in context," and Ross suggests local TV stations and publishers, who are "starving for cash," as sponsored-video partners for local firms. David's version for law firms: move the statute-of-limitations content nobody read on the blog onto YouTube. (17:01–20:56)

• Mine your reviews for the landing pages you're missing. Feed the review corpus to Claude, ChatGPT or Gemini alongside the Ahrefs or Semrush MCP and look for ways customers describe you that no page reflects. Petco's example: "picky eater" kept surfacing across dog-food reviews and became its own angle. Use search volume as a proxy for prompt volume to prioritize. (22:48–25:39)

• GEO is CRO. Help the model make the decision the same way you'd help a buyer: put awards, top reviews and hard proof points ("14 years in business," "80 case studies") on the homepage. "Hit LLMs over the face with" data points. (26:26–27:20)

• Versus pages predict AI traffic best. In Siege's analysis, having comparison ("X vs. Y") pages correlated more strongly with site-wide AI traffic than any other content type. You control the message and face little competition for the click. (27:20–28:22) Siege Media study [2]

• Some brand can only be bought. In personal-injury law, brand comes from billboards and TV (Ross cites Thomas J. Henry, whom Near Media sees constantly in its research). In "best car insurance" listicles, a startup can pay its way onto the list for a week, but Geico's brand click-through will push it off. (29:51–31:01)

• Siege won't take clients without existing demand. Its Reddit and affiliate services now require a minimum level of branded search, around 20,000 a month. "Gasoline on a fire that's already burning." Data journalism and digital PR are its main way of building brand from scratch. (31:12–32:07)

• Run a quarterly award hunt. Awards are Ross's fastest brand lever for local: run a Claude workflow every quarter to find "best in [city]" lists, apply, show the wins on your site, and repeat every year. (33:04–34:08)

• Ask open-ended attribution questions, not multiple choice. Multiple-choice answers default to the first option. A low-friction open field surfaces channels you didn't know were working. Ross treats LLM tracking as "a compass rather than a GPS" and avoids synthetic prompt-volume data. Siege uses Peec AI and also rates Profound and Scrunch. (34:20–37:23) [3]

• On Reddit, monitor and redirect — don't fake it. Use the free Reddit Pro to catch recommendation threads. Then ask your known five-star Google reviewers to add a review on Yelp, or to weigh in on the thread if they authentically can. Respond to brand mentions fast. (37:53–39:28)

• On Clutch, go deep, not wide. Diversify reviews, but go all-in on the highest-quality platform for your category. For agencies, Ross says that's Clutch: sponsor the city sub-lists ("San Francisco SEO companies") rather than the national list. (41:10–42:33)

Related Reading

• Generative Engine Optimization: The Definitive Guide to AI SEO — the book's landing page (Wiley, 2026). Amazon · Wiley.
• The Content That Predicts AI Search Traffic Most: Versus Pages — Siege Media's correlation study behind the versus-page claim.
• How Generative Engine Optimization (GEO) Rewrites the Rules of Search — the a16z post that popularized the term "GEO."
• EP 275 — Rankings Up, Calls Down: What AI Visibility Is Really Worth to Local Businesses (Part 2) — Andrew Shotland and Mark Kabana on attribution and self-reported lead sources.
• EP 274 — Why "Position One" Doesn't Exist in Local Anymore (BrightonSEO Recap) — the same BrightonSEO San Diego conference that David references.
• EP 270 — The Hidden Advantage: How Brand Awareness Decides the Click in Legal Search — the user-behavior research Greg describes at 28:22.
• AI Search Stats in 2026 — Similarweb's gen-AI traffic-share tracker.
• Yelp deal pushes local reviews into ChatGPT — Axios on the Yelp–OpenAI partnership Greg mentions.
• The New Rules of AI Search — Crystal Carter's book, mentioned in the intro.
• Tools mentioned: Peec AI, Profound, Scrunch, cloro, Clutch.

Practitioner Notes

• Open-ended attribution is the measurement layer local firms are missing. Ross's "Claude came out of nowhere" moment came from a free-text "how did you hear about us" field. Near Media's monthly lead-source attribution work for law-firm clients faces the same blind spot in GA4. An open-ended intake question is the cheapest way to see LLM-sourced leads, and it pairs naturally with the self-reported lead-source argument in EP 275.
• Ross's "brand must be bought" point in legal is backed by Near Media's own research. Thomas J. Henry's ubiquity is exactly the recognition effect in Near Media's Choosing a Lawyer user study (EP 270). The open question is whether that offline awareness carries over to LLM recommendations the way it carries over to SERP clicks.
• The GBP card in AI Overviews is the exception to "no clicks." David's observation: AIO mentions almost never click through to a local business website, but the newer desktop GBP card, with stars and a photo, does draw clicks. That makes GBP completeness and review velocity a direct AI-visibility lever, not only a local-pack one — worth tracking in GBP research.
• Yelp is back on the review-diversification list. With Yelp's data now licensed into ChatGPT, Ross's "ask your five-star Google reviewers for a Yelp review" tactic stops being optional for local clients that have let Yelp lapse.

Concepts

• Current state vs. ideal state — the gap between what models cite today and what they would cite if better third-party evidence existed. Ross treats the gap as the opportunity.
• Influence over citations — optimizing for the places real people form opinions (and that writers research), on the bet that citations follow, rather than buying placement on low-traffic pages that happen to be cited.
• GEO is CRO — making a page easy for a model to recommend uses the same moves as making it easy for a buyer to choose: awards, reviews and hard proof points up front.
• Versus content — head-to-head comparison pages ("X vs. Y"). Siege's data ties them to site-wide AI traffic.
• Search volume as a proxy for prompt volume — prioritizing with real keyword demand rather than synthetic estimates of what people ask LLMs.
• Compass, not GPS — LLM visibility tracking is directional. It tells you whether you're improving, not your exact position.
• "Gasoline on a fire" — paid amplification channels (Reddit, affiliate) accelerate existing brand demand but can't create it.
• ICP mapping — building pages for each ideal-customer dimension (industries, use cases, roles) to catch long-tail, refined prompts.

Pick your starting point:

00:00 Intro: Ross Hudgens and the GEO book
02:42 GEO vs. AEO vs. AI SEO: the label fight
09:37 How different is GEO from SEO?
11:47 Yelp's comeback in AI answers
13:08 Which AI engines to track (drop Perplexity)
15:38 Current state vs. ideal state for citations
17:01 YouTube and third-party video
22:16 Mining reviews for content
26:26 GEO is CRO
27:20 Versus pages and AI traffic
28:22 Brand: when it has to be bought
33:04 Awards as a brand lever
34:08 Attribution and tracking tools
37:23 Reddit without spamming
39:28 Biggest mistake: chasing citations
44:16 Will LLM clicks grow?

Full Transcript -->

The Near Memo — Ross Hudgens on Generative Engine Optimization

Hosts: Greg Sterling, David Mihm · Guest: Ross Hudgens, founder and CEO, Siege Media

Editor's note: This is a cleaned reading transcript. Stutters, false starts and repeated words have been removed. Clear transcription errors have been corrected, including brand and product names. Short acknowledgments ("yeah," "mm-hmm," "right") are folded into the surrounding turn. Otherwise the wording stays faithful to what was said.

Greg Sterling (00:10)
Hey everybody, welcome back to the Near Memo. With me today is David Mihm, who co-founded Near Media with me and Mike Blumenthal. Mike is out today, and we are very excited to be talking to Ross Hudgens, who is the author of the new book Generative Engine Optimization, or GEO for short — a term we don't like, and we'll talk about that. But it is a good book. Welcome, Ross. Thanks for being here.

Ross Hudgens (00:32)
Yeah, well, I really appreciate you guys having me.

David Mihm (00:35)
Despite the hostile reception, you're excited to be here, huh?

Greg Sterling (00:38)
It's not hostile. It's a challenge. Anyway, I should mention that you are the founder and CEO of Siege Media, which is a digital marketing agency that's been around for a long, long time. How many years, in fact?

Ross Hudgens (00:52)
Fourteen.

Greg Sterling (00:53)
Okay, that's a pretty long time. And who typically are your clients? Just give us a little sense of that.

Ross Hudgens (00:59)
Yeah, generally high-growth startups. We tend to be national. We do have several multi-location brands, and I know that's a lot of your audience on the local side, so that's good context for what I say. But I'll try to tailor it.

Greg Sterling (01:11)
Okay, great. So as I said in the green room before we started, there are tons of people writing GEO, AEO, LLM optimization and AI optimization books. Crystal Carter is out with one, which is supposed to be very good also. There are tons and tons of people out there. What did you want to say that was different, or what made you feel compelled to write this book, given all that?

Ross Hudgens (01:35)
Yeah, in short, it was partially a timing thing. I was pushed to do it. I'm part of a business forum of peers, and they brought up this idea of the Misogi Challenge. It's a Japanese thing: essentially, you undergo a hard challenge once a year. I thought it was a pretty cool concept. I've always wanted to write a book — I was a creative writing major for a moment — and I thought, we're in this AI moment, it's literally perfect timing to do this. Obviously there are challenges with AI in particular, given how fast this whole thing moves. But we in particular have already been doing a lot of things that touch GEO, especially on the national side, which we'll get into for local. I have a background in affiliate marketing and lead gen, that kind of context. We're doing digital PR and Reddit — I'm an active Redditor. We touch a lot of the things that influence it. So it felt like we had some credibility to do that. And a lot of things like the third-party roundups and the first-party roundups, which are still influential for some brands, we had already been doing for clients for several years. So it just happened that we had a good sample size of what works, and it felt like an applicable time to talk about it.

Greg Sterling (02:42)
So let's talk about the term GEO, since I challenged you on that. There's a bunch of terms floating around, and GEO seems to be one of the couple that have taken hold. Obviously there's SEO, which we all know. GEO, AEO. Some people have used the term LLMO, which probably isn't going to stick. AI optimization is the more generic term, which I typically use. What term is the right term? And the second part is: Google says good SEO is good GEO, and other people say no, it's a completely different discipline. People like Mike King and others would really say it's not the same as SEO and you have to do different things for it. So where do you stand on that — the label, and then how similar to SEO or different?

Ross Hudgens (03:32)
Yeah. Starting with the label, my opinion is to use what the customer uses. I chose GEO. Honestly, I like "answer engine optimization" slightly more as a framing, but I did the data research, and as of now — and it's still the case today — "generative engine optimization" is more searched, and we generally hear it more. Answer engine optimization is not far behind.

Greg Sterling (03:57)
So the book title is GEO.

Ross Hudgens (04:01)
Yeah, it's GEO. I literally also put "Definitive Guide" — you can't see it on here, but it's The Definitive Guide to AI SEO, because I've got to get that extra keyword in there, you know?

Greg Sterling (04:11)
Right. That's so funny. So where do you come down on SEO versus GEO? David, go ahead.

David Mihm (04:20)
Well, let me just chat a little bit more about the genesis of choosing this as the title, or choosing this as the acronym. From a purely capitalist perspective, as an agency owner or consultant or whatever, I actually kind of agree with you that it's better to shift the framing to GEO. Especially in local, SEO has almost never had a seat at the table when it comes to marketing priorities at enterprise brands. And local SEO has been one step further removed from that table. So taking the opportunity of this big macro shift to try to get a seat at the table is not a bad idea, and I don't begrudge you the framing.

Although I will say I have not heard this from my clients — and I don't have very many as a consultant, maybe half a dozen regular clients — and we don't hear it at Near Media from our market research customers very much. So I'm wondering: does this partly have to do with your customer base being tech, software or other startup people who are more inclined to be on this new thing? Or do you truly think that across the board — retail, e-commerce, home services, you name the local category — this is what people are talking about?

Ross Hudgens (05:49)
Yeah, that's a great point, David. I'm using a generalized framing of the term, so going to your market and framing for them is more pertinent, for sure. The origin of the framing — or at least the popularized origin — is Andreessen Horowitz writing a post. You guys might have seen it: yet another "end of SEO," which I don't agree with. I forget the exact title; I talk about it a little bit in the book. But they're obviously highly influential in startup and tech spaces, so it's understandable why the community latched onto that to a good degree. Profound is using AEO, I'm sure, because they have the most capital behind them.

David Mihm (06:29)
Well, Profound has a billion reasons to do that.

Ross Hudgens (06:35)
So by all means, use what fits. My customer may not be your customer, so you adjust to the customer base and get that context of what they're saying. Obviously you all do some great surveys, so I'm sure you've got a better pulse than I do.

David Mihm (06:48)
Yeah, and I feel like "AI SEO" is actually more how our customers tend to think about it — using AI, but being visible in the search portion of that journey just feels like a more natural evolution. You've been working with much larger customers than we have for most of our careers. It took small businesses until, what, Greg, probably 2015 or 2016 before they knew what SEO meant and before it became an actual channel for them. I think we're still dealing with the ramifications of that today. So I hesitate to shift to something that's going to take yet another seven or eight years of education to help move the conversation along.

Greg Sterling (07:28)
One comment about that, which is different from the terminology but goes to the underlying awareness. You always used to have to drag small businesses kicking and screaming into digital marketing, for various reasons. For years and years people were saying, "Your competitors are doing this, and if you're not, you're not going to be visible and you're going to lose business," and so on.

In my travels — talking to small businesses, interviewing small businesses and doing some SMB surveys — the degree of adoption of AI was really striking. It was in the culture. They got hold of it and figured out certain things they could do with it. They're not as sophisticated as enterprise users, but they were ahead of the curve on this marketing trend, which is completely unprecedented for them — maybe social, but otherwise unprecedented. It's really interesting. Now I think they've hit a wall in terms of what they can do. As a group they're not a monolith, obviously, but you can figure out certain things intuitively, and then when you try to get into workflows and automation, it becomes much harder for them. But that's just a digression about adoption.

David Mihm (08:49)
Yeah. And Greg, I think this maybe speaks to the split in customer bases. As you said, SMBs are adopting AI, and newer SMBs are probably AI-native, or whatever that term is going to be. So leaning into the AI portion of the optimization that we're all trying to help them with is maybe the right path. From a messaging and positioning standpoint, that may even be something to think about as an agency: what terms are your customers using? That's not something Google Trends is necessarily going to tell you. So Ross, I think you're smart, given who your audience is and that you're trying to speak to the C-suite, to latch onto generative engine optimization as the hook. But I'm not sure it's the right hook for everybody writ large.

Greg Sterling (09:37)
So let's talk a little bit about the substantive differences between SEO and whatever this other thing is that we're calling GEO, AEO, et cetera. As I said a moment ago, there are people who think there's something like an eighty percent overlap between SEO and GEO. There are people who think they're very separate disciplines, with maybe only a ten or twenty percent overlap. And then there's Google, which says it's more or less a one-to-one thing. Where do you come down on that? And if you think it's different, what are the things that are different?

Ross Hudgens (10:08)
Yeah, this conversation so far is a great segue to the answer, in my view. Talking with both of you before we hit record further clarified for me that, of all the industries out there, local has the least difference between SEO and AI SEO or GEO. It sounds like you guys agree.

When we go national, there's more divergence, because there's less of the off-site review context that Google has already compiled on the local side. We were talking about the advantage Google has and the data it has: all that review context is essentially what the models are grasping for on these other sites, and it simply doesn't exist in some instances. So they go off-site — to Reddit, to affiliate sources, to LinkedIn, to YouTube. Those were indirect sources before. They probably contributed to a higher click-through rate and dwell time and all those things, but I don't think they were direct; that was just my opinion. Today I think they're direct in influencing GEO, for certain sites anyway. And that percentage changes by business.

We were talking about the e-commerce example: you could have a niche D2C e-commerce brand that had no chance to outrank Amazon or Walmart, but it could be recommended in AI search today. If you think about that, they're not really doing any SEO on their own site, but they can contribute through things like reviews and the positioning of their Amazon listings, which maybe you could call SEO. To me that's the clearest example of the divergence between off-site and on-site.

Greg Sterling (11:47)
We talked about this in the green room a little bit, before David joined: Yelp. Yelp has historically been a despised site among marketers and local businesses, and now it has renewed importance, arguably because of the way it's being relied upon to some degree by Gemini and certainly ChatGPT, which has a formal partnership with Yelp and some other sites. That's an example where people focused largely, if not exclusively, on Google reviews and neglected Yelp for the last bunch of years, and now the advice is, "Hey, you've got to look at Yelp again, because it's getting some play in AI."

Marketers can't focus on multiple LLMs that have different sources and different ways of surfacing content, right? All the reports out there, like the Semrush and Ahrefs reports, show there's only so much overlap. What are you doing or recommending in terms of which engines? Google is kind of its own thing, but let's include Gemini in this: Gemini, Perplexity, ChatGPT, Claude, Grok, DeepSeek for some people. And Copilot, which everybody's sort of not paying attention to. Do you just flat-out ignore some of these? What's your hierarchy of attention?

Ross Hudgens (13:08)
Ideally you can ask customers in some way. If you can have your sales reps, or the people answering calls to your business, ask how they heard about you, that context will help inform whether these engines are relevant to you. You can have different lead flows depending on the customer base, of course. We ask how you heard about us on an open-ended form field on our site. You could also do it post-conversion as a multiple choice, if you're concerned about conversion rate. That will tell you whether people are actually mentioning Perplexity or Copilot. Of course, look in GA4 data too where you can. But I'd assume that for the average local business, ChatGPT is the only relevant LLM other than Google's models.

So one of the core recommendations we've been making recently is to actually drop Perplexity from trackers. A lot of them default to Perplexity, and we historically added it, but it's continuing to shrink. If it's not weighted, it could make you go chase some listing that seemingly has twenty-five percent of the impact but has less than one percent of the actual impact. So aligning that modeling with the models people actually use will be fruitful, hopefully.

David Mihm (14:17)
Yeah. I thought one of the better presentations at Brighton — the strangely named BrightonSEO in San Diego, which we were all at a couple of weeks ago — was Baruch Toledano from Similarweb. He had a great chart looking only at LLMs, separate and apart from what's happening in search. Essentially, ChatGPT has kind of flatlined at around 35%, and Gemini and Claude are the only two that are growing. So it's basically those three from a market-share standpoint — with the caveat that maybe you have a bunch of rabid Perplexity fans, or you sell to huge customers who only allow their employees to use Microsoft products on the corporate network. But from a broad consumer standpoint, it feels to me like it's ChatGPT, with Claude hot on its heels, and Gemini — we should all care about what Google's doing in this space.

Ross Hudgens (15:04)
Agreed. Claude is showing up significantly for us, for B2B clients especially. So if you're in B2B, tracking it could be feasible, although it's so expensive. As of now — or the last I heard — they're using Brave Search, so you could look at your rankings there as a proxy pretty easily as well.

Greg Sterling (15:22)
You just alluded to Claude for B2B. Are you making distinctions based on the client's vertical, or on whether they're B2B or B2C, in terms of what you're optimizing for? Not the content, but the engine.

Ross Hudgens (15:38)
Or in terms of the strategy — is that what you're saying, Greg?

Greg Sterling
Yes.

Ross Hudgens
It starts with getting the source URLs out of that context, and that will inform what's listed there. Obviously there are ebbs and flows: ChatGPT isn't referencing Reddit as much recently. It's still referencing it, but I think it's just increasing the overall set of sites it's pulling data from, so Reddit is getting diluted. But I haven't thought as much about it model by model or network by network — ChatGPT versus Claude — as compared to the aggregate view.

So if we start with that being correct, the secondary thing, building off that a little, Greg, is that I do think there's an ideal state. You get the source URLs and the source domains as they are right now. But something I realized in year one of doing this: I'd look and say, "It's only the homepage," or "It's only competitor sites," or "It's only sites like ours that have 60% of the citations." That's probably due to an absence of third-party citations. Just because something is the current state doesn't mean it's the ideal state for the models. The natural follow-on is to think, "I need to do exactly what this says." But my opinion is that these models want third-party reviews, and we can encourage more of those in various ways. If we see an absence of them, that may be an opportunity, rather than a destiny that says we need to focus only on our own site.

David Mihm (17:01)
I wanted to pick up on something you mentioned — I think it was in your Brighton presentation itself, and it's something I've seen elsewhere now that I'm paying attention to it. You brought up that YouTube is the number one most-cited single domain for consumer queries, and I think it was two or three for B2B queries, if I recall. I feel like that's a red flag — "red flag" is the wrong word, but it's a "hey, you guys need to be paying attention to this." Video for SMBs, for multi-location folks and even for my enterprise clients who are somewhere in the middle has felt like a sorely underinvested channel for years and years. And it's something everybody can do now. Anybody with an iPhone and an AI tool can put together a really slick video educating customers about what they're selling, or talking about what's happening in their market. So I wonder: in my experience, video typically sits with the social team at the enterprise level. How do you get into the video content discussion with your clients?

Ross Hudgens (18:07)
So it's actually a lot of third parties. What we're seeing is a lot of third parties doing those videos. It's often an affiliate motion or an influencer motion, where we're paying these people to describe — hopefully organically — the situations where our client belongs. The exception is middle-funnel product comparison content. It could be lab-grown diamond versus raw diamond, or whatever — that kind of framing.

Greg Sterling (18:34)
Blood diamond.

Ross Hudgens (18:38)
I don't think brands are putting that on their own websites, but those kinds of deep pre-purchase comparison searches do show up. One of the framings we like is: what percentage of the time are brands being mentioned in those LLM prompts? Those can be fruitful.

On the local side, it's interesting. I heard you talk about that, David, and my mind immediately went to whether the value equation works — would you want a video for every city-plus-service combination, given the raw cost of making each one? But to your point, you can do it scrappily; you could work with an influencer. And I know you work with a lot of lawyers who are probably primarily in one geography. They could probably afford a single reference like that.

Greg Sterling (19:17)
Well, those guys —

David Mihm (19:19)
Yeah, and just —

Greg Sterling (19:20)
I was just going to say, on the lawyers: those guys have a lot of video, because they're doing TV or connected TV. That's incredibly influential in terms of building awareness and building brand, which we can talk about in a minute. But go ahead, back to you, David.

David Mihm (19:33)
I was just going to say there's no shortage of locally relevant educational content that lawyers can use to insert themselves into this middle-of-funnel, AI-driven conversation. What's the statute of limitations in my state? What are seven things to look out for from an insurance company, and why shouldn't you take their first offer? There are a lot of opportunities. There's a lot of historical blog content that people wrote and, frankly, nobody ever read — I've worked for lawyers for a long time. I think that style of content will actually do much better on YouTube, and it can help reinforce that firm being cited in an AI Overview for these kinds of queries. They're not exactly informational queries, but they're mid-funnel queries where people aren't necessarily in the market for an attorney just yet.

Ross Hudgens (20:25)
And a lot of these local publishers are really starving for cash, right? They already have the means to create video — they have the local TV station — and I know there are already sponsored placements happening in those contexts. So a smart local firm could work with, I think I've seen a Fox LA or whatever, to do that. They're probably not as sophisticated about search as an influencer would be, but if you're sophisticated, or you're an agency owner who can nudge those people, that could be pretty powerful.

Greg Sterling (20:56)
One of the things I'm seeing on TikTok a lot lately — and I have to call out that the algorithm shows you what you look at, right? You look at something and then it shows you a ton more of that kind of thing. But I'm seeing a lot of influencer-created hotel videos, most of which are probably paid: "Come with me as I look at this hotel." I just see a ton of those, and many of them are boutique hotels. They're not part of big chains. It's not "come to the Hilton Waikoloa Village" or something; it's an individual hotel, which I think is pretty interesting. That kind of content can obviously be repurposed, cut, edited, what have you. It may not be exactly what you're talking about, but it can be uploaded to YouTube, and that's an easy thing that's happening quite frequently.

Ross Hudgens (21:48)
Yeah, that variation is definitely something to ask for if you work with someone on a YouTube video. I see clips like that being surfaced in citations from Instagram a lot. I haven't seen TikTok as much in LLM citations yet, but that doesn't mean it can't go that route. I think we'll increasingly see more social — Facebook as well, of course. So making sure they're repurposing that everywhere is just a no-brainer ask to maximize your dollars.

Greg Sterling (22:16)
So we're already talking about content. Let's talk a little more broadly about it. How do you do content now, in this world? In the Google world, especially for local, it's GBP, reviews, local landing pages and a handful of other things. Pretty structured, pretty straightforward — maybe difficult to execute, but conceptually simple. How do you think about content now, and where do you get content ideas? I think that's a major question mark for many people.

Ross Hudgens (22:48)
One of the most LLM-centric things I've been saying is to use your review context to inspire content creation. So many local businesses have a wealth of review volume, and now we can feed that to Claude or ChatGPT or Gemini and say, "Cross-reference this with, say, the Ahrefs MCP or the Semrush MCP, and look for data within it." Is there a certain way people describe something they do with your business that you don't currently have reflected as a landing page? That should be surfaceable through co-citation and the intelligence of these models working over the review corpus. Then that review data can also be used on the landing page itself — you've got enough content simply from hand-selecting a few of those reviews.

An example I give: Petco has, of course, dog food pages — wet dog food, say. A variation that surfaced from the reviews was that a lot of people said, "This food really resonated for my dog; they're a picky eater." Several dog foods had "picky eater" in their reviews. That just bubbled up in the review data that LLMs helped cross-reference and accelerate. Of course you don't have to have an LLM for this, but it definitely helps at enough volume to take advantage of that UGC.

That also leans into the idea of the long tail. People are doing these refined searches on these models, so let's make it easy for the models to surface that we do these things, by having landing pages that map to them. Of course, you don't want to saturate and dilute the rest of the site's equity by doing that scammily. That's where review volume is awesome: if you have it, you're effectively saying by nature that you have a quality page, because you have enough context.

Greg Sterling (24:38)
So would you use the volume as an indicator — focus on the top five or top ten things and leave the rest of the long tail alone? Look at where people are concentrating their questions and focus on that? In a world of limited resources — we're not talking about everything being AI slop, just about making choices — I think that's very often the hard thing: what do I write about, and how much of it? Go ahead.

Ross Hudgens (25:07)
Yeah, agreed. It's whatever the data says. If you have a lot of search volume associated with something, that's probably a good proxy that there's LLM prompt volume for it. Our core recommendation is to use search volume as a proxy for prompt volume. If you have that context, it's pretty helpful. And if you get enough reviews screaming at you that people care about a specific thing — search volume data can be inaccurate sometimes, we all know that — that may be instructive. And of course you use Search Console data for hints, too.

Greg Sterling (25:39)
What else are you doing for your clients, or recommending they do, beyond tracking where leads come from and the review analysis, to help them understand what their customers want or are looking for? As you say, the queries get very, very specific. This is the thing that's interesting about LLMs that was never really true of Google — it was sort of true of Google, but it's different: you can start with "here's my problem, I don't know what I need to solve it," and very quickly get to a recommendation, within a single session. How do they address that? What do you tell them about how to deal with this new behavior in LLMs and what it means for content?

Ross Hudgens (26:26)
The most applicable thing there — well, there are a few things. One I've been saying recently, and actually just posted today, is that GEO is CRO, or CRO is GEO. You're trying to help LLMs make the decision to recommend you. When they're in that deep-funnel consideration stage, comparing one lawyer to another lawyer, don't make it difficult for them — and in turn, don't make it difficult for users. If you don't have traditional CRO best practices in place, the most common workflow I've been recommending recently is: apply for awards, get awards, feature them on the homepage, and feature your great reviews on the homepage. These LLMs will source them anyway, but make them very easy to find. I've also seen award blocks with things like "14 years in business" or "80 case studies" — those are the kinds of data points you can hit LLMs over the head with, and I think they'll make them recommend you a lot more often.

The second variable is versus content. It's probably most applicable to lawyers, maybe not to restaurants. We did a data analysis — yes, mostly of software clients, plus fintech, e-commerce and other high-growth companies, within our client base. The data point that correlated most with AI traffic — not AI traffic to that article, but overall AI traffic to the site — was whether they had versus articles. It was a Spearman correlation of around 0.56. I'm not a mathematician; you can find it on our site. That made sense to me after the fact; I always knew they were important. Think about it: one, you're controlling the message. Two, you're not really competing with many other people for that same click, so of course you'll probably get more clicks to those pages as well. But if you're influencing these LLMs with context about your brand and not letting other people own that conversation, that's potentially highly influential for getting recommended and picked in that down-funnel context. So that's a highly influential lever we've been pushing.

Greg Sterling (28:22)
So what about brand? Let's talk about that for a second. In our work, especially with lawyers but in other categories as well, brand plays a kind of hidden role. Everybody has been talking about building your brand since the Google antitrust trial in 2024, the API leak, NavBoost and all of that. Where do you think brand fits in this?

I mentioned to you earlier that we do user behavior research: we watch people search and see how they interact with the pages. I don't know what percentage of the time it is, but it's significant: people hit a page in a legal context where they don't really know anybody. They're looking at the LSAs, the local pack, a bunch of ads and some organic content. They scan the page and say, "I've seen these guys. I've seen their commercial, I've heard their ad, I know these guys." That seems to be a very powerful driver of clicks, and very often of ultimate conversions. And it's largely invisible to the companies themselves; they just see analytics showing whether they were clicked or not. What's your sense of the role brand plays in LLM visibility, and what do you recommend about how to build it? There are a lot of tactics, but what do you say to, say, a new software company or a startup? Where does brand fit in your tactical advice to them?

Ross Hudgens (29:51)
Huge, undoubtedly. It depends on the market. In some markets brand almost has to be accomplished through paid, and I think of legal that way. Here in Austin, my drive is just billboard after billboard from lawyers — Thomas J. Henry, I don't know if you guys know him —

David Mihm (30:07)
We've seen his ads and his site plenty in our research.

Ross Hudgens (30:10)
Yeah, he's ubiquitous. He's spending some cash. I don't know if I love him — maybe I can introduce some negative sentiment — but I don't want to be sued by him. Anyway, highly influential.

The other reference point I like to give, from affiliate, is a term like "best car insurance." There are all these listings informing LLMs that have a list of, say, five car insurance brands. You can't be a new startup on that list and expect to win over Geico or Progressive, simply because of the sheer brand awareness. You could pay to play for maybe a week, but they have better conversion rates and better brand click-through rates, and that model is going to shift the whole conversation in a way that forces you out. Honestly, you probably won't even get considered by the affiliate. It's pretty challenging. And I forget exactly what vertical you mentioned, Greg — sorry if I skipped that.

Greg Sterling (31:01)
You mean as an example? I was just using B2B software as an example, since you referenced it before. How do you talk to them about brand and how to build it?

Ross Hudgens (31:12)
A few different ways. One, we do a lot of data journalism and digital PR research reports — obviously pretty prevalent right now. Building data that can be referenced and promoted to the press is a great way of doing that, so if they're not doing it, that's our biggest component of the story. We're also increasingly saying, yes, we can help you with that through Reddit and through affiliate, which are still paid channels and brand builders — but of course you need to have a brand. So a lot of our services now have qualifiers. If you don't have search volume of, say, twenty thousand a month on Reddit for your brand — your branded terms — we actually won't entertain you at all. I think a lot of services are like gasoline on a fire that's already burning. It's hard to start a fire anew, at least for a lot of these clients. So hopefully you can help accelerate it rather than start from scratch.

Greg Sterling (32:07)
One of the interesting things about local is that you can do a lot of guerrilla offline activity to build awareness and generate brand searches that way, and also to be recognized. It really depends on who you are. We've talked about local link building for a long, long time — you want your links on local sites that have authority with Google. But there's another side to that, which is showing up at local events. The example given to me recently was plumbers who show up at a farmers market. They get a lot of foot traffic and build awareness of their brand, and that translates into a certain number of online queries. There's a lot local businesses can do to build awareness — not just small businesses — and traditional media is one of them, which is often neglected. That translates into some of the stuff we're talking about. It's a holistic proposition, but there are things people don't think about that can be done to jumpstart it.

Ross Hudgens (33:04)
Awards. Awards are the simplest, fastest route, and to me they'd carry over pretty nicely. You could run a quarterly workflow in Claude that monitors for awards you can apply for. We have "best in Austin" lists here — best restaurants, et cetera. Get those awards, feature them on your website, and repeat that cycle yearly: you want to be the best in 2026, 2027. That can be the brand building. That's how restaurants build brand for me: I look at those lists, and now I'm aware there's a hot new coffee shop I've got to try.

Greg Sterling (33:39)
Many of those are pay-to-play, or there's a fee to join. So I guess that's just what you have to do.

Ross Hudgens (33:46)
Yeah, unfortunately. I understand the cost of putting these together. Hopefully people aren't paying to be called the best, but if they're paying to be considered, I understand some threshold on that. And with affiliate, I like to say there's at least a correlation with quality when people are paying for PPC — you can't do that forever if you're not good. For most businesses, you have to be at least pretty solid to keep that up.

Greg Sterling (34:08)
What are the right metrics to track on all this? You've got all your strategies and tactics. How do you communicate to clients that this is working, or give them a sense that they're moving in the right direction?

Ross Hudgens (34:20)
Attribution is key. We talked about asking on a call, asking on the form, ideally doing all of it — asking on calls, on forms and post-conversion. It depends on how much you think it will hurt your conversion rate; you can make it optional if it's pre-conversion. That will give you full confidence that, hey, 20% of my leads are coming from LLMs — maybe that's high for most people, but if it's in that range, you can now quantify the value of your efforts there. LLM data in GA4 is quite challenged, so it's currently going to be very difficult to get that out. To my knowledge — you all could probably add context on the local side.

Asking the customer and using open-ended self-attribution is preferred over multiple choice. The reason is that with multiple choice people will often just pick the first option; they forget what they did. If you challenge them slightly, without high friction, they're more likely to tell you something new, and they may surprise you with what actually had an impact. Maybe you showed up on a list you didn't even realize you were on. For us, we saw Claude come out of nowhere, and we would never have known that — at least not at the same lead volume — if we hadn't had that open-ended form.

Greg Sterling (35:34)
What tools do you trust? You said GA4 doesn't provide accurate information about LLM sourcing. What tools are you using, or think are worth anything? There's a big debate about synthetic data — are they really showing you what people are doing, or is it all just modeled? What are your thoughts?

Ross Hudgens (35:53)
The analogy I like to give is that LLM tracking is like a compass rather than a GPS. Are we going in the right direction? We can't claim precision by any means, but the larger the sample size — if you're running these prompts daily — the more I see a correlation between sales and that visibility improving, if it's done well. We're picking prompts that have search volume as a proxy, and they're bottom-funnel prompts. As we continue to improve that visibility, I've seen a very strong correlation. Of course we need to ask customers to know it with confidence, but the link is pretty clear from what I've seen. We don't use prompt volume. I know there are tools that offer it, but that's not something we're currently recommending, for the reason you mentioned, Greg — it's synthetic.

Greg Sterling (36:40)
Can you name any tools you think are particularly good or worthwhile?

Ross Hudgens (36:44)
Yeah, sure. We use Peec; we like Peec. We've just recently been looking at cloro.dev. Apparently it's an open-source version that powers a lot of these tools, so you could spin up your own version. They don't want to market themselves super loudly, because they power a lot of these tools — or at least are one of their data sources. But you could potentially spin up your own. I think Profound's good too. Transparently, we went with Peec because, at that time, early on, they were just more expensive, and with Peec we didn't have to pass the cost on to our clients. Scrunch, I know, is good. But we're in with Peec. It has a good UI, and I think it's a pretty strong tool.

Greg Sterling (37:23)
I was going to ask you about Reddit. Reddit is a difficult problem, I think, for a lot of marketers. It's used, although not as much as it was, by LLMs. I don't know what's happening on Google right now — it shows up in these forum packs. What's your general position on Reddit, and how should people approach it? Because if you go in and spam, it's obviously going to be removed completely or have no impact. What do you tell somebody considering Reddit?

Ross Hudgens (37:53)
On the local side, my main suggestion — which also dovetails with our other conversations — is to monitor. Use Reddit Pro, which is a free tool: input the keywords that matter to you and monitor for mentions of those terms. Those will be recommendation threads. Don't jump in there and fake-respond. But what you can do in those situations is go to the people who gave you five stars — the cream of the crop, five-star reviewers you know personally — and send them an email that says, "Really appreciate the review you already gave me on Google. We'd love it if you'd give us a review on Yelp," and potentially point them to the Reddit thread as well. That Reddit thread lets them jump in immediately. If you're monitoring actively and getting enough volume, you can make that secondary ask for someone to authentically give a review there. They may not be able to, and if so, they can just go to Yelp instead and review you there, which is potentially productive for you. Of course, find-and-replace with whatever channel you think is most influential. But that's the most organic approach to me, for a city-relevant thread.

The second piece, as a brand, is to respond to your brand comments super fast. We've been thinking about reputation management on these platforms: there's simply massive leverage in getting negative reviews pulled down and responding quickly. Maybe people won't review you negatively just because you're so active there. Flipping reviews — all of that is true on Reddit too.

Greg Sterling (39:28)
Okay, a couple of final questions and then we'll let you go. This has been a great interview, with a lot of great advice for folks. What do you think marketers are getting wrong right now about AI visibility? What's a fundamental mistake or two you see people making?

Ross Hudgens (39:42)
The one that comes to mind first is aiming for the current state versus influence. There are AI citations, and a lot of people are just trying to get citations in the current state of the environment. I think if we all optimize for where real influence is, we're going to end up with the citations. The problem is, if you go to some AI spam site that has a hundred citations in ChatGPT but not one person visits it, that's probably short-lived. It will go away at some point, and the influence will leave with it.

But if you go where people are actually talking about you and aim to influence that — the Reddit example is a good one, where they're talking about you on Reddit even if it isn't showing up directly in ChatGPT right now — it's still very influential on Google. I'm seeing Reddit everywhere on Google. People who are thinking about you are going to research that, and it's going to have an impact. And vice versa: if someone creates their own listicle, they're looking at Reddit for that research, and that could influence how they write their own article about you. So go where the influence is, rather than just where the citations are, and I think you'll end up with citations as a byproduct.

Greg Sterling (40:52)
We've got a lot of agencies of different sizes that listen to the podcast, many of whom work with local businesses or multi-location brands. What's your strongest piece of advice for folks trying to figure out their AI visibility strategy while still doing all the stuff they do on Google?

Ross Hudgens (41:10)
Invest in reviews, for sure. It depends on the brand, but Clutch is highly influential for us at the national level, and I know they have location-specific lists. SEO in particular is hyper-competitive, and it tends to be more localized. They have an SEO companies list. You probably shouldn't sponsor that one, but you should potentially sponsor the sub-locations that are relevant to you — I think that's probably very influential. If you're more likely trying to optimize for "San Francisco SEO companies" or "LA SEO companies," that kind of thing, that would be big.

One of the core things I think about, and recommend in the book, is going relatively all-in on the platforms you have faith in. There are many places you can ask for a review, and I know I just said to get reviews in a few places — I think that's still smart for diversification — but I've always liked Clutch, simply because I think it's the highest-quality one, and I think quality will win out over time. DesignRush is sort of there for agencies, but Clutch is really powerful.

I talked about all the trust-signal elements on your site. Then long tail: I think vertical pages make a ton of sense. If you don't have vertical pages, you should. I call it ICP mapping: do that whole map and have use cases, verticals, potentially roles — but industries, I think, would be really applicable for a lot of your audience. It's probably obvious, but yeah.

Greg Sterling (42:33)
What about clients — advice for agencies dealing with their clients while doing this optimization work? Most people are already pretty far into this, but what would you say?

Ross Hudgens (42:46)
Getting visibility into their leads, and really encouraging and pushing the client on it, because otherwise we can't speak to this impact. If clicks decline, attribution declines, and it's quite challenging for us as agencies to point to our value. So that's a leading indicator. I'm speaking from experience here, and maybe a lot of your listeners have already felt that pain. That switch went from nice-to-have to "we're putting this front and center in our decks," because we have to have it to know how we're doing. Otherwise, you're going to get churned.

Greg Sterling (43:23)
Final question from me — David may have a question or two. Look, we're doing this a year from now; we reconnect a year from now. What do you think will still be true, and what won't be, in terms of AI optimization?

Ross Hudgens (43:40)
Hopefully that influence point is a decent prognostication: going where the actual customers and the influence are will lead to higher visibility a year from now. It may not be higher visibility a month from now, but if you're playing the one-year game, hopefully that's a pretty good bet for you. It's hard to say exactly. I think there will be fewer clicks. There will most certainly be more decisions made directly in LLMs, so optimizing for that world is a smart thing. Going where we expect things to be, rather than chasing the current state, is probably the smartest move for people — and maybe that's not a prediction. Yeah, David.

David Mihm (44:16)
I'll ask a very specific, directly related question. Do you think actual clicks from LLMs will increase, decrease or stay the same over the next twelve months? The most common pattern we see, even with AI Overviews today, is not someone seeing a business in the AI Overview and clicking the link to its website. It's someone seeing the business and then clicking a link further down the page that mentions the same business the AI Overview mentioned. So do you think direct conversions from LLMs are going to increase, decrease or stay the same? Or will we see a continuation of what happens now, where ChatGPT might recommend a business and then somebody goes to the browser and types in the website, or does a brand search on Google for that business?

Ross Hudgens (45:00)
I'm optimistic. I don't think it will fully solve the attribution problem, but I'm hopeful about the incentives in both of their business models. With ChatGPT's ad model, they've got to figure out how to get people to click things, and we already know Google has that incentive. I know they're testing different things. I don't know if you've seen it in your own data, David, but businesses get linked to and still aren't clicked, despite that.

David Mihm (45:25)
That's ninety percent of the time, anecdotally, when somebody's mentioned. Now, if it's their Google Business Profile that's inserted into the AI Overview — which is a relatively new interface on desktop in the last couple of months — that's a totally different story. The review stars and the photo get people's attention, and they click that. But we're still not seeing direct click-through from an AI Overview to a local business website in most cases. It's much more of an indirect influence as part of the wider search journey.

Greg Sterling (45:54)
Yeah, it leads to a brand search.

Ross Hudgens (45:56)
That makes sense in local. I could see that being a little more neutral to negative, like you mentioned, because there's some probability — or feeling — that you could just call this person and never visit the website. That will probably only increase.

Greg Sterling (46:09)
All right. On that note, Ross Hudgens, thanks very much for all your time, your great advice and the book —

David Mihm (46:14)
And for being such a good sport.

Ross Hudgens (46:16)
Of course. GEO for life, you know?

Greg Sterling (46:20)
The book is Generative Engine Optimization. Where can people buy it if they're interested?

Ross Hudgens (46:25)
Amazon. You can search "GEO book Ross." Hopefully we'll be ranking for "AI SEO book" soon — you'll contribute. But this is not an affiliate motion.

Greg Sterling (46:32)
We'll put a link in the show notes also. And we get no affiliate revenue from this, just to clarify.

David Mihm (46:38)
This was not a paid placement.

Greg Sterling (46:43)
I should mention there's a bunch of things we didn't get to, but one major, major topic we didn't get to — and maybe we'll talk about it at some point in the future — is agentic search: what it means for people and their data, and how real it's going to be. It's a huge topic that we'll have to come back to.

David Mihm (46:59)
That's going to be Ross's second book.

Ross Hudgens (47:02)
It was actually my future section. I wrote that there's going to be more and more agentic activity. It's definitely interesting that the initial e-commerce tests were kind of failures, from what I saw. I haven't seen an update on that recently, but the fact that it converted so poorly is interesting. But for a super-commoditized thing, like buying a movie ticket, I'm happy to use an agentic experience.

Greg Sterling (47:23)
Yeah. People will use the easiest and most efficient tool that doesn't do something really bad with their data —

David Mihm (47:32)
— or order them ninety thousand rolls of toilet paper from Amazon.

Greg Sterling (47:38)
That's right. All right, Ross, thanks very much, and good luck with everything. I know you've got a massive tour of other podcasts and places you're going to be, so thanks for putting us on the list. And thanks, everybody, for listening. We'll see you next time.

Ross Hudgens (47:52)
I appreciate you having me. It was a lot of fun.