When AI Can't Tell You Apart from Your Competitors

Your brand might be ranking just fine and still vanishing from the AI answers that drive pipeline. We unpack cluster collision, the moment LLMs decide your messaging sounds too much like your competitors and summarize you all away. If you lead an agency or own a growth strategy, this one reframes how you think about visibility.

Inside the episode:

Why pattern matching in SEO now works against you in AI search

The data on source rotation, citation, and why being cited isn't being recommended

How to spot where your messaging is interchangeable with competitors

What it takes to build a narrative AI systems retrieve, repeat, and recommend

Why your real competition is the cluster of brands an LLM can collapse you into

The question to bring to your next leadership meeting: is your content making your brand more visible, or more interchangeable?

Transcript

SPEAKER_00 0:13

Usually when we talk about search engines, there's this um this expectation of a map, like a very clear alphabetized index. You type a keyword, you get a list of ten blue links, it's predictable. You know exactly where you stand, right? Page one, page two. It's this clean, almost comforting hierarchy that we're all just so used to. Aaron Ross Powell Right.

SPEAKER_01 0:34

I mean we've spent basically two decades building entire marketing departments around that exact map.

SPEAKER_00 0:40

Oh, absolutely.

SPEAKER_01 0:41

The rules of visibility were just, well, set in stone. You optimize the page, you build the links, and you climb the ladder.

SPEAKER_00 0:46

Trevor Burrus, Jr.: But then you step into the world of AI search, and suddenly that map is drawn on an ECS sketch.

SPEAKER_01 0:52

Yeah.

SPEAKER_00 0:52

We're looking at a visibility landscape that is just terrifyingly unpredictable. And well, that is our mission for this deep dive.

SPEAKER_01 1:01

Exactly.

SPEAKER_00 1:02

Because today we are exploring the very real possibility that doing everything right, according to traditional SEO, might actually make your brand completely invisible in the age of AI search. Trevor Burrus, Jr.

SPEAKER_01 1:15

Which sounds crazy, but it's true.

SPEAKER_00 1:16

Aaron Powell It really is. And we are pulling from a literal mountain research to figure this out. We've got Will Reynolds' latest strategic frameworks on search visibility, a massive Cystrix data set.

SPEAKER_01 1:29

Aaron Ross Powell Oh, that data set is incredible.

SPEAKER_00 1:31

Yeah. Tracking over 80,000 AI prompts. And a really provocative manifesto on why sounding professional is basically a death sentence for your brand.

SPEAKER_01 1:42

Aaron Ross Powell Right. And I think the overarching theme across all this research, you know, is that the fundamental architecture of how information is retrieved has just completely changed.

SPEAKER_02 1:51

Yeah.

SPEAKER_01 1:51

We are moving away from a system that retrieves documents to a system that synthesizes answers.

SPEAKER_02 1:57

Okay. Yes.

SPEAKER_01 1:57

And the really scary part is that the instincts that made a business successful on Google for the last 10 years are the exact instincts setting them up to fail with large language models.

SPEAKER_00 2:07

Trevor Burrus, but let's unpack that. Because on the surface, it sounds super counterintuitive, right? Like if I've been winning on Google, shouldn't I just, you know, keep doing what I'm doing?

SPEAKER_01 2:15

Aaron Powell You'd think so.

SPEAKER_00 2:16

Aaron Powell But the old playbook rewarded one very specific behavior, which was pattern matching. You looked at the top 10 results for a keyword, you matched the intent, you basically used the exact same subheadings and you just tried to do it like 5% better.

SPEAKER_01 2:31

Aaron Powell Yeah. I mean it was pure risk mitigation. Search algorithms back then were essentially just these pattern recognition machines. Right. So if every top ranking page for, say, a software category had a comparison matrix and a pricing tier, well, you built a comparison matrix and a pricing tier.

SPEAKER_00 2:47

Aaron Powell Exactly. You just mirrored them.

SPEAKER_01 2:48

Right. Because the algorithm saw that you look like a recognized authority, so it ranked you. But the byproduct of everyone doing this simultaneously for a decade is this ocean of absolute sameness.

SPEAKER_00 3:01

Yes. And the manifesto we're looking at really calls out the specific language of this sameness. It's all those blurry, safe corporate words.

SPEAKER_01 3:08

Oh, the buzzwords.

SPEAKER_00 3:09

Right, like trusted, innovative, end-to-end, scalable, AI-powered. I mean, seriously, read the home page of almost any B2B company right now, and it's practically a madlibs of these exact terms.

SPEAKER_01 3:23

Aaron Powell It really is. Because marketers well, they use those words to feel safe. They want to sound like the category.

SPEAKER_00 3:29

Sure.

SPEAKER_01 3:30

But in an AI-driven search environment, sounding like the category triggers this phenomenon called cluster collision. Cluster collision.

SPEAKER_00 3:37

Okay.

SPEAKER_01 3:37

Yeah. So when an AI models language, it plots words and concepts in this multidimensional vector space.

SPEAKER_02 3:45

Right.

SPEAKER_01 3:45

If your brand describes itself using the exact same semantic grouping of words as 50 of your competitors, the AI's architecture literally cannot distinguish you. It just groups all of these identical sounding brands into one dense cluster, picks maybe two or three domains as representative of the whole, and then just relegates the rest to interchangeable background voice.

SPEAKER_00 4:09

So I want to visualize this for you listening. Imagine you're at a massive networking event and ten different people walk up and hand you the exact same beige business card, right? With the exact same tagline. You aren't going to remember all 10 individuals.

SPEAKER_01 4:25

No way.

SPEAKER_00 4:26

You're just going to remember that, oh, a bunch of people there do synergistic consulting. If your website reads like a consensus statement, the AI simply treats you like consensus.

SPEAKER_01 4:36

And we really have to look at what that means for the actual buyer's journey. Because the real threat of cluster collision isn't just, you know, losing your ranking position. Okay. It's total invisibility and consideration. Aaron Powell Wait.

SPEAKER_00 4:48

I actually want to push back on that distinction a little because in the old 10 blue links model, being, say, number seven on the page still meant you were on the page.

SPEAKER_01 4:57

Well, technically, sure.

SPEAKER_00 4:58

Like you still got impressions. Even if you were the seventh beige business card, someone might misclick or just scroll down and find you.

SPEAKER_01 5:05

Right. But an AI summary doesn't give you a list of seven, it compresses the entire category.

SPEAKER_00 5:09

Oh, I see.

SPEAKER_01 5:10

So if an AI generates a single paragraph summarizing the best supply chain software and it only names three vendors, the other seven don't get lower impressions. They get zero.

SPEAKER_00 5:20

Wow. Zero.

SPEAKER_01 5:21

Yeah. You lose mind share before the buyer even reaches your website. It's a total loss of visibility that happens so high in the funnel, it's practically invisible in your standard analytics dashboard.

SPEAKER_00 5:33

That is terrifying. Your traffic just slowly evaporates, and you have absolutely no idea why.

SPEAKER_02 5:38

Exactly.

SPEAKER_00 5:39

Which perfectly explains Will Reynolds' concept of zombie content. He points to this really widespread practice of writing scaled, templated listicles.

SPEAKER_01 5:49

Oh, I see this everywhere. Trevor Burrus, Jr.

SPEAKER_00 5:50

Right. The example he gives is a brand churning out an article on the best restaurants in Minnesota purely to capture a high-volume keyword. Right. I mean, no actual human being types something that broad when they're looking for dinner tonight.

SPEAKER_01 6:02

Right.

SPEAKER_00 6:03

But marketers write it anyway, just to feed the algorithm.

SPEAKER_01 6:06

Aaron Powell Well, they write it because the old algorithm ate documents, it constantly needed new pages to index.

SPEAKER_02 6:11

Right.

SPEAKER_01 6:12

But an LLM doesn't want your localized SEO trap, you know? It wants the actual grand truth. So when an AI answers where should I eat in Minneapolis, it isn't going to cite a random B2B vendor's zombie blog post. It's going to synthesize local reviews, map data, and actual culinary consensus.

SPEAKER_00 6:28

Trevor Burrus, Jr. So if creating this safe beige zombie content basically guarantees we get thrown into a generic bucket, how do the AI platforms actually decide who gets to represent that bucket to the user? Like who gets pulled from the cluster?

SPEAKER_01 6:43

Yeah, that's the big question.

SPEAKER_00 6:45

And that is exactly what the Cystrix data set out to answer. And I gotta say, the mechanics of what they found are wild.

SPEAKER_01 6:52

Absolutely fascinating.

SPEAKER_00 6:53

They analyzed over 82,000 prompts across 17 weeks, testing three different AI platforms across six countries, and they were measuring something they call citation drift.

SPEAKER_01 7:05

Right. So citation drift is essentially the measurement of how volatile the sources are in an AI's response week over week. Like, do the exact same websites get cited for the same prompt, or does the machine constantly swap them out?

SPEAKER_02 7:17

Yeah.

SPEAKER_01 7:17

And the data totally proves that treating AI search as this one monolithic thing is a fatal flaw. The platforms operate on fundamentally different architectures.

SPEAKER_00 7:27

I want to dive straight into Google AI overviews first because these numbers just broke my brain. Systrix calls this environment the closed circle. In 53% of the prompts they tested, the sources cited didn't change a single time over the entire 17-week period.

SPEAKER_01 7:44

Just completely locked in.

SPEAKER_00 7:45

Yeah. Out of roughly 11 domains cited, eight stay permanently anchored. But here is the massive catch. Even though the AI is pulling from the exact same sources every single week, it rewrites the actual text of the answer 87% of the time.

SPEAKER_01 8:01

Yeah, and you have to think about the mechanism behind that. It's using a process called retrieval augmented generation or ANG.

SPEAKER_00 8:08

Right again, right.

SPEAKER_01 8:09

The system fetches the documents, those stable eight domains, and locks them into its context window. But the language generation model, the part that actually writes the sentences, is probabilistic.

SPEAKER_00 8:20

So it's basically guessing the next word.

SPEAKER_01 8:22

Exactly. It's essentially a student rewriting an essay every single week, but they are only allowed to use the exact same three books from the library.

SPEAKER_00 8:29

Wow. So the AI is essentially pledged allegiance to those specific websites.

SPEAKER_01 8:33

Pretty much.

SPEAKER_00 8:34

It doesn't matter how the user phrases the prompt or how the AI hallucinates the prose. The bibliography is totally fixed.

SPEAKER_01 8:43

Yeah, it tells us that for Google AI overviews, the barrier to entry is just incredibly high.

SPEAKER_02 8:48

I bet.

SPEAKER_01 8:49

If you aren't in that initial snapshot of trusted domains, well, you don't get in, but the reward for breaking through is basically permanent residency.

SPEAKER_00 8:59

But hold on. If we look at Google AI mode, which is, you know, a different feature within the same ecosystem, we see a completely different behavior. Totally different. Cystrix calls this the carousel. In AI mode, 56% of the sources rotate every single week. It is not a closed circle at all.

SPEAKER_01 9:16

No, but it's not entirely random either. The data shows the structural split.

SPEAKER_02 9:21

Okay.

SPEAKER_01 9:21

About 86.5% of the prompts in AI mode have a stable core, usually one to five domains that just stay put. Got it. But the other 89% of the individual links make up the peripheral carousel, and those are the ones swapping out every week.

SPEAKER_00 9:35

So the strategy completely changes here. If I'm optimizing for AI mode, I can't just like celebrate that my brand showed up in an answer on Tuesday.

SPEAKER_01 9:45

Right. You can't pop the champagne yet.

SPEAKER_00 9:46

Because I might just be visiting on the carousel. By Thursday, the AI might swap me out for a competitor.

SPEAKER_01 9:52

And tracking that requires a total shift in how we measure success. You're no longer managing a static ranking, you are managing a continuous presence metric.

SPEAKER_00 10:01

Continuous presence. Okay. And if you think Google's carousel is volatile, the data they found on ChatGPT is just staggering.

SPEAKER_01 10:09

Oh, it's wild.

SPEAKER_00 10:09

Systrix found a 74% weekly churn rate. The median prompt in ChatGPT doesn't have a single domain that stays for all 17 weeks, is total fluctuation. Unbelievable. But there's a specific stat here that I cannot stop thinking about. For German language searches, 68% of ChatGPT's sources are actually in English.

SPEAKER_01 10:30

Yeah, and this is a really critical insight into how LLMs are trained versus how traditional search indexes are built. Oh so? Well, Google's a search engine first, right? It maps the local web. But ChatGPT is a language model heavily biased by its training corpus. And the vast majority of the internet's high authority institutional documentation is just in English.

SPEAKER_00 10:53

So think about what that means if you are a global brand listening to this.

SPEAKER_01 10:57

It's massive.

SPEAKER_00 10:58

You could be pouring millions of dollars into localized, perfectly translated German SEO content. But when a German user actually asks Chat GPT a question about your industry, the machine ignores your localized site entirely because its architecture inherently trusts an English white paper more.

SPEAKER_01 11:16

Exactly. Which really highlights the absurdity of trying to have one unified AI search strategy.

SPEAKER_00 11:22

Yeah, it doesn't work.

SPEAKER_01 11:22

A strategy that anchors you in Google AI mode might rely heavily on native language shop pages, but that same strategy completely fails on Tat GPT. They're diametrically opposed.

SPEAKER_00 11:33

Okay. So knowing that a stable core does exist, at least within Google's architecture, we need to figure out what type of content actually earns that permanent seat, like who survives the carousel and who gets tossed off.

SPEAKER_01 11:43

Yeah, the Systrix data establishes a very clear hierarchy of survival here.

SPEAKER_00 11:47

Okay, what's at the top?

SPEAKER_01 11:48

The ultimate winner with a 24% core rate is YouTube. Video content is just incredibly sticky in generative responses.

SPEAKER_00 11:58

Which, you know, on the surface feels kind of obvious. I mean, Google owns YouTube. Of course they're going to preference their own property.

SPEAKER_01 12:06

Aaron Powell Well, I'd actually push back on calling it just a property preference.

SPEAKER_00 12:10

Oh really?

SPEAKER_01 12:10

Yeah. Think about the data structure of a YouTube video. It's not just a video file, it comes with a massive timestamp transcript of unstructured text. Aaron Powell Oh.

SPEAKER_00 12:21

Think about that.

SPEAKER_01 12:21

Trevor Burrus, Jr.: Right. And LLMs love parsing dense conversational text to extract semantic meaning. It just provides a richness that a standard, you know, 500-word blog post simply doesn't have.

SPEAKER_00 12:33

Aaron Powell Okay. That makes total sense. It's multimedia doing double duty.

SPEAKER_01 12:36

Exactly.

SPEAKER_00 12:37

Aaron Ross Powell Now, on the flip side, the ultimate losers in this data set are editorial news sites. They have a dismal 1.4% core rate. It's rough. They basically get quoted once for a breaking topic, and then they completely vanish from the AI's response the following week.

SPEAKER_01 12:51

Aaron Powell Yeah, because large language models are structurally terrified of transient data.

SPEAKER_00 12:56

Aaron Powell Terrified of it.

SPEAKER_01 12:57

Right. The models know they are prone to hallucination, so their temperature for volatile, rapidly changing news is tuned to just discard it quickly.

SPEAKER_00 13:06

Aaron Powell That makes sense. They want safe answers.

SPEAKER_01 13:08

Exactly. They want to anchor to stable, evergreen knowledge. That's why product pages, technical documentation, and shop pages completely dominate the stable core.

SPEAKER_00 13:18

And this volatility applies even when the user is searching specifically for you. The data looked at branded queries, specifically a prompt for BBC female newsreaders.

SPEAKER_02 13:29

Right.

SPEAKER_00 13:30

Now, obviously the BBC's own domain stays anchored in the core, but the co-citations, the other websites, the AI lists alongside the BBC to provide context rotate out 70% of the time every single week.

SPEAKER_01 13:41

Wow. And furthermore, we really have to distinguish between the domain and the URL.

SPEAKER_00 13:46

Okay, what's the difference there?

SPEAKER_01 13:47

So at the domain level, Google might consistently link to, say, IMDB.

SPEAKER_02 13:51

Yeah.

SPEAKER_01 13:52

But at the URL level, the exact subpage being cited, the drift, is 85% per week. Yeah. The AI swaps out the specific page it references constantly.

SPEAKER_00 14:01

Aaron Powell So if you're reporting to your executive team, you can't just walk into a meeting and say, hey, our new blog post on supply chain logistics is ranking in AI overviews.

SPEAKER_01 14:12

No, it's definitely not.

SPEAKER_00 14:13

Because by the time the meeting ends, that specific URL might be totally gone. You're buying a ticket for a train that kicks you off the very first stop.

SPEAKER_01 14:20

Aaron Powell And this is the paradigm shift of generative engine optimization or GEO.

SPEAKER_00 14:25

GEO, right.

SPEAKER_01 14:26

You are no longer optimizing to rank number one. You're trying to influence the probabilistic factors that a machine uses to construct a response on the fly.

SPEAKER_00 14:34

Aaron Powell Which brings us to the hardest part of this whole equation. We know that AI punishes generic saneness and we know it favors distinct evergreen content. But how do you actually rebuild your strategy to be that distinct source? Like how do you force the AI to anchor to you?

SPEAKER_01 14:49

Aaron Powell This is where Rule Reynolds' concept of being seen, believed, and chosen becomes so critical. Aaron Powell Okay.

SPEAKER_00 14:56

Breakdown. Seen, believed, and chosen.

SPEAKER_01 14:58

Aaron Powell So for years, marketing was only about being seen. You hacked the metadata, you got the traffic, and job done.

SPEAKER_00 15:05

Right. Check the box.

SPEAKER_01 15:06

But in an LLM ecosystem, visibility without belief is basically useless.

SPEAKER_00 15:11

Aaron Powell There's a brilliant example in the research about this, the ethical genes thing.

SPEAKER_01 15:15

Yes, perfect example.

SPEAKER_00 15:17

So a clothing brand used old school SEO tactics to rank number one on traditional Google for the term ethical genes, even though they had no real-world reputation for ethical manufacturing. Trevor Burrus, Jr.

SPEAKER_01 15:28

Right. They just had a really good title tag.

SPEAKER_00 15:30

Exactly. But the moment AI search rolled out, that brand just vanished from the answers.

SPEAKER_01 15:35

Aaron Powell Right, because AI models don't read title tags the way a crawler does.

SPEAKER_00 15:39

They go.

SPEAKER_01 15:40

No, they look for consensus signals. They scrape the broader web to really understand the relationships between entities.

SPEAKER_02 15:46

Okay.

SPEAKER_01 15:46

So if humans on forums, in reviews, and in digital communities don't organically associate your brand with ethical manufacturing, the AI won't cite you no matter how perfectly optimized your website is.

SPEAKER_00 15:57

Wow.

SPEAKER_01 15:58

You simply cannot fake consensus.

SPEAKER_00 16:01

So how do you actually break out of the seameness trap? The manifesto argues you have to replace recycled positioning with original proof.

SPEAKER_01 16:09

Original proof, yes.

SPEAKER_00 16:10

You can't just publish another best practices list that just curates what everyone else is already saying.

SPEAKER_01 16:15

Right.

SPEAKER_00 16:15

You need proprietary data, you need firsthand pattern recognition, you need named frameworks and specific operational insights that, frankly, only your company could possibly know.

SPEAKER_01 16:27

And you have to verify that humans actually care about it. Will Reynolds strongly suggests that marketers go to platforms like Reddit to see if their brand narrative actually holds water.

SPEAKER_02 16:36

Reddit.

SPEAKER_01 16:37

Yeah. LLMs are aggressively scraping Reddit right now because they are starved for authentic human opinion. Oh, that makes a lot of sense. If the sentiment about your brand there is negative or worse, non existent, your AI visibility will plummet. Real brand affinity drives revenue, not just algorithmic visibility. In fact, direct traffic converts 1.5 times better than SEO, and social converts five times better.

SPEAKER_00 17:02

Aaron Powell Let me play devil's advocate here for a second because this all sounds great in theory, but let's be real about the corporate environment.

SPEAKER_01 17:08

Okay, let's hear it.

SPEAKER_00 17:09

Will Reynolds asks if we are willing to sacrifice a little bit of our visibility game to be more believable. But if I'm a marketing director and my CEO is breathing down my neck for quarterly traffic growth, how on earth do I justify stepping back from the easy, scalable, zombie content that juices those top line numbers?

SPEAKER_01 17:33

Aaron Powell Look, that is the daily reality for most teams. It really is. But you have to force a conversation about business outcomes versus vanity metrics. Okay. If your visibility is skyrocketing because you're churning out generic listicles, but your pipeline and your revenue are completely flat, what are you actually achieving? Trevor Burrus, Jr.

SPEAKER_00 17:49

Right. Nothing.

SPEAKER_01 17:50

Exactly. You are optimizing for an illusion. You have to ask your leadership team a very uncomfortable question. Trevor Burrus Which is if an AI had to describe why our brand matters in one single sentence, what would it say?

SPEAKER_00 18:02

Oh man. That is a brutal question.

SPEAKER_01 18:04

Aaron Powell It really is.

SPEAKER_00 18:05

Because if you haven't given the AI the original proof to answer that, it's just going to say they provide scalable, end-to-end synergistic solutions.

SPEAKER_01 18:13

Aaron Powell Exactly. And if that's the answer, you don't have an SEO problem. You have a foundational positioning problem.

SPEAKER_02 18:19

Yeah.

SPEAKER_01 18:19

If your information architecture is just answering broad queries without deepening the actual meaning of your brand, the AI has absolutely no reason to pull you from the cluster.

SPEAKER_00 18:30

Okay, let's distill all of this down for everyone listening. First, the old playbook of pattern matching is dead. Blending in leads straight to cluster collision, where AI just treats you as interchangeable noise.

SPEAKER_01 18:41

Spot on.

SPEAKER_00 18:42

Second, AI citation drift is very real and it's highly platform dependent. Google AI overviews lock in a stable core of sources while hallucinating the text, while ChatGPT wildly churns its sources with a huge bias toward English documentation.

SPEAKER_02 18:56

Yep.

SPEAKER_00 18:56

And third, to survive this volatility, you need original proof. You have to be believed by humans before you can be chosen by a machine.

SPEAKER_01 19:04

That's the core of it. And as we wrap up, I want to leave you with a concept that goes a bit beyond just optimizing your own site.

SPEAKER_00 19:11

Okay, I'm ready.

SPEAKER_01 19:12

We've talked a lot about how AI collapses generic brands into a single mental bucket. But what happens if an AI system hallucinates a completely false narrative about your industry category? Oh wow. Right? If all of your competitors are pumping out generic, interchangeable zombie content, they might actually accidentally reinforce that hallucination through sheer volume.

SPEAKER_00 19:34

Just by feeding it the same bad info.

SPEAKER_01 19:36

Exactly. So the real question is how do you prove your distinct truth to a machine that has already made up its mind based on a consensus of noise?

SPEAKER_00 19:45

That is a terrifying thought, honestly. But it's exactly the kind of problem we need to be anticipating right now.

SPEAKER_02 19:50

Absolutely.

SPEAKER_00 19:51

Well, thank you for joining us on this deep dive. My challenge to you listening today go pull up your own website copy right now, read your homepage. Does it pass the sameness test? Or are you just handing out another beige business card to an Etch a Sketch algorithm? We'll catch you next time.