[Newport, 1965]
Bob Dylan walked onto the stage at the Newport Folk Festival with an electric guitar and a backing band.
The audience booed.
Members of the folk movement criticized him for moving away from his folk roots and performing with a rock band.
The music was, by most accounts, extraordinary. And while initial reaction was hostile among parts of Dylan’s folk audience, the retrospective view is almost the inverse: his electric turn is now regarded as one of the defining breakthroughs in popular music.
Why? Well, Dylan was the acoustic folk guy. But what, he’s wielding an electric guitar now!? Category alignment problem.
Hendrix came onto the scene without much historical “positioning” work.
There’s a famous story where Hendrix jammed with Cream in October 1966, shortly after arriving in London, and played “Killing Floor” at a speed and intensity that stunned Eric Clapton. Clapton reportedly walked offstage, and while angrily smoking a cigarette and pacing, asked something along the lines of, “Is he really that fucking good?”
Point is, if people know you for anything, then what they know about you lingers in their memories and expectations.
It’s hard to enter a market, let alone lead a market, and it’s very tricky to navigate or pivot to new markets. It’s why positioning is such a foundational and high leverage exercise.
As AI search summarizes your brand into a somewhat concise answer, the echoes of your brand and its perception are now intermingled with your own website and claims.
The same factors that lead to an incumbent’s advantage (quantity of mentions, ubiquity in influential sources like Reddit/review sites, topical authority, generally high brand awareness) make it more difficult to turn the ship if something changes or is unfavorable.
Authority, SEO Armor, and Cumulative Advantage
I’m not saying this was irrelevant to SEO. You couldn’t get out past your skis too far in SEO, either, lest you incur the wrath of a Helpful Content Update.
I can tell you, however, from personal experience working at HubSpot, that it was trivial to write and rank for high value topics, even outside of your core niche, if you had sufficient domain authority.
This was a huge strategic underpinning of SEO programs. Domain authority compounds over time, and a fifteen year head start meant fifteen years of link equity, topical authority, and indexed pages. This is why companies created link magnets, ran plays like scholarships to get .edu backlinks, distributed badges that link back to your website, etc.
And with all of that authority, if you wanted to reposition, say, from a CRM to an “AI-native CRM,” you updated your homepage, rewrote your solutions pages, and the domain authority carried any new content up the rankings.
We’ve written about these dynamics in-depth, even positioning our original barbell strategy as a way to index on both authority building assets while building out the “money pages” that rank with increased authority and domain power over time. My early guides on content marketing strategy and economics are built on this idea.
While these dynamics didn’t disappear (i.e.flywheels and The Matthew Effect are still the core mechanisms by which organic growth programs scale), the same advantage that helped build SEO privilege (years of links, mentions, authority) make it harder to sway models’ opinions of your brand in AI search.
Take that same brand who is hoping to reposition from a CRM to an “AI-native CRM,” but consider AI search.
You now have thousands of data points that describe your brand one way, and while updating your website messaging is one way to correct this, it’s a drop in the ocean.
A Brittle Monopoly
A recent study on brand bias in LLM recommendations tested how GPT-4o-mini, Claude, and Gemini recommend products when known brands compete against fictional alternatives.
When all products had identical specifications, the known brand was recommended 100% of the time. Every single trial, across all three models, both languages tested, and all product subcategories. The incumbent advantage was huge.
But here’s where it gets interesting.
That monopoly broke with less than a +0.1-star rating advantage for the competitor.
The default seems to have been a tiebreaker, a default heuristic the model falls back on when it has few other signals to generate recommendations. The moment a challenger offered even marginally better specificity, the brand recognition advantage decayed.
This maps to what Bliey and Chatwin argue in their Stanford AEO paper:
“A niche brand producing authoritative, data-rich content for a narrow domain achieves higher vector-space proximity than a generalist incumbent whose content is broad but shallow.
Large brands’ legacy SEO libraries, optimized for high-volume keywords, produce few passages that survive chunking and pairwise comparison for specific long-tail queries. Their content covers topics without addressing the precise attribute intersections that agentic queries target.”
Now, could a large brand couple their default status with highly specific content and pages? Possibly. But challengers have a clean swim lane to win on very specific characteristics and use cases.
Another study, LightSite AI’s analysis of over 1,000 websites confirmed the pattern empirically: smaller, focused competitors can and are outperforming global brands in AI search.
They diagnosed two gaps for large brands – a content gap (enterprise content that’s polished but generic, “written more for slides than for real conversations”) and a technology gap (product data not exposed in structured, machine-readable formats).
Again, some of this may be correlation (i.e. larger brands being slower to adapt to modern search and content strategies), but it, at the very least, opens up the idea that the playing field is open, particularly when it comes to specificity.
Think to your own usage of LLMs. First, your workspace probably knows a lot about you already.
So when I ask for an AEO software recommendation, it already knows I run an agency that works with B2B brands.
And then on the user behavior side, queries tend to be longer and more contextual.
I’m giving incredibly specific requirements, such as pricing and integrations and my use cases. This, in contrast to a short phrase like “best AEO software,” may shift the landscape of recommendations.
A Clarification on Incumbency
In this essay, I’m using the term “incumbent” broadly as “one that occupies a particular position or place.”
Often, this means a large enterprise, but not always. It simply means that your brand has established a position and has built “equity” in that position.
I have seen three examples this week of varying company sizes and industries:
- A legacy analytics software founded in the 1990s that has substantial visibility for legacy positioning, but is struggling to adapt to AI native queries despite launching popular products in the space.
- A fast growing scale-up that wants, more than anything, to go upmarket and attract enterprise customers, yet has a 10x AI visibility gap between broad market queries and enterprise modifiers
- A startup that has pivoted over the past several years and now targets a new persona with a new market position and product.
I’ll tell you the story of the latter since it is being publicly chronicled.
The Barge vs the Speedboat
Mutiny is a great example of the challenges of steering positioning in AI search. You can follow their AEO journey through Matt Ratchford’s LinkedIn posts.
Basically, they built incredible authority in one category (B2B website personalization) when they first launched, and have pivoted over the years to be an agentic AI platform for sales people.
Much of the work Matt describes in the LinkedIn series resembles the tactics you’ve heard about.
As he puts it: “onsite content claims what you are, offsite content builds consensus around that claim, and the two together get you retrieved and cited. Visibility goes up. Simple, right?”
Turns out, not really.
The goal isn’t simply to “move visibility,” it’s to enter a new category with new competitors, and to clean up historical positioning in the process. This requires pruning pages, launching new media and PR campaigns, but also the tedious work of reaching out to the many, many websites that describe them as B2B personalization software and asking them to update the description or remove them from the page.
My previous essay, In AEO Being a Brand Is The Point, still applies here. But the reality is, authority built up in B2B personalization, competing against Optimizely, doesn’t necessarily translate to authority in AI sales platforms, competing with Clay and Apollo.
Matt put it well here:
“Look at Clay. Look at Apollo. Look at the players who own this category. There’s a real step change between them and everyone else.
Here’s what I think actually vaults a company across that gap: compounding marketing and flywheels, over the course of a long period of time. Clear value props that pulling people into the product -> Turning them into champions -> Giving them a reason to share -> Getting a lot of people talking about you out in the wild -> Acquire more users. And so it continues.
The leaders got there because a ton of people know them, use them, and vouch for them, and the visibility is a by-product of that.”
In SEO, your website was the authoritative source for what your company does, with backlinks bolstering your authority and review sites as a secondary proof point in the consideration phase.
In AEO, the model synthesizes across the entire corpus – your site, third-party reviews, listicles, press releases, podcast mentions, G2 profiles, old blog posts, analyst reports. Your page is one data point among thousands. If a thousand other pages describe you as a personalization tool for marketers and your homepage says you’re an AI sales assistant, the model has to reconcile that.
The Snowball (i.e. the Matthew Effect’s Revenge)
I’ve described this mechanism before, but it’s worth repeating: awareness begets awareness. AKA, the rich get richer.
If you’re well known in a category, when a new publisher writes a roundup of best products in that category, you get listed by default. You have to list Profound in AEO software comparisons, for instance.
True of human writers, but new content is increasingly researched and drafted agentically. An AI agent pulling together a brief on the “top tools for X” runs web searches, finds the existing corpus, and synthesizes what’s already out there. If an incumbent has years of exposure in a particular category, the research that forms the brief for new content describes them the same way they were described before. The old narrative compounds.
This is absolutely amazing if the compounding is in your favor, but a runaway train if not.
Suganthan Mohanadasan’s research on ChatGPT’s query fan-out showed that the model compiles a shortlist of brands before it fetches a single page. That shortlist is built from the model’s training data, which is, by definition, the past.
The Speedboat’s Window
The challenger’s advantage, then, is a clean slate, specificity, and a deep understanding of the customer pain points and decision criteria.
This is partially why we index so heavily on voice of customer research to inform our AI visibility measurement as well as subsequent roadmaps and strategy (the bigger reason is that it’s just great marketing practice).
This is, in some ways, an acceleration of what Benji and Devesh from Grow & Convert called “Pain Point SEO,” where you plan around relevancy, specificity, and jobs-to-be-done instead of high volume topics.
When a buyer prompts ChatGPT with “I need a scheduling tool for distributed teams with mostly hourly workers that integrates with our existing HRIS and doesn’t require a six-month implementation,” the challenger that has built focused, data rich content for exactly that use case, through documentation, case studies, product page copy, blog content, and off page proof, has a chance of surfacing.
The Stanford paper predicts this leads to “hyperspecialization,” or small brands doubling down on niches, large companies segmenting into sub-brands. The market composition shifts to a new layer of specialized brands operating between consolidated manufacturers and consumers. Brand equity, instead of being a single powerful asset that lets you expand into adjacent markets, becomes “localized to specific markets.”
It’s, of course, unclear how predictions will play out for many reasons, but the idea holds up against the current day mechanics.
But there’s a caution for challengers too.
Startups are built on the pivot, iterating your way to product-market fit, and pivots are a feature, not a bug.
Except now, if you build up substantial exposure in any given hypothesis before pivoting, you inherit a miniature version of the incumbent’s problem.
The startup that iterates in public builds a legacy faster than it realizes. The very agility that’s supposed to be your advantage can become a liability if you’re not conscious about what you’re leaving behind in the corpus.
Narrative Architecture
A prospect recently asked a great question, “where does AEO stop working?” Basically, what can we do and what can we not do through our work.
Most of it circles back to the foundations.
For instance, we cannot fix product and customer experience issues at scale.
We once had a well-known company (Superbowl ads and all that) ask us to do this, yet they had thousands of negative reviews on Reddit due to poor product quality and customer experience. It’s whack-a-mole to fix the surface issue there, a waste of resources. Fix the product and experience, first.
Similarly, narrative architecture is an art and a science and is upstream of any organic growth work that builds upon it.
Like Dylan, you’ll probably want to evolve. It’s how companies survive and thrive. There’s a way to do it, and several companies have done it well.
One example that came to mind: Robert Greene.
The author worked something like 80 jobs before he wrote The 48 Laws of Power. Restaurant work, assistant gigs, disconnected roles with no obvious throughline.
The way he described it: he saw the backside of every type of organization and observed the same power dynamics playing out in every one. He looked back through history and saw the same patterns in Napoleon and the Medicis. The 80 jobs became part of his story and his positioning. .
HubSpot went from inbound marketing to CRM to front-office software, and each transition was additive, each prior identity nested inside the new one. The story expanded without contradicting itself. Like Russian nesting dolls: freelancer inside agency owner inside entrepreneur inside author.
The competitive implication is that narrative skill (the ability to tell a coherent story that connects where you’ve been to where you’re going) becomes a concrete strategic asset, not a branding nicety.
Companies whose evolution can be narrated as expansion will navigate AEO better than companies whose evolution requires explanation.
As I continually say, good product marketers are very valuable in the AI era.
Visibility != Recommendation
Something JH Scherck noted a while back: just because your brand is mentioned in an AI answer does not mean it is a recommendation.
This is worth watching, not only if you are going through a positioning shift or launching a new product, but also to gauge the accuracy of how you are represented in AI answers.
For instance, we are quite visible in our core service category entry points, but I am constantly identifying descriptions that don’t line up with reality.
In some cases, they are deliberate attempts by competitors to dilute our positioning. Okay, there are actions for that.
In many, many more cases, I have to point the finger back at myself and our company though – we simply hadn’t communicated our capabilities in a given area, so writers and the AI models had nothing to source.
When I discover a strategic discrepancy in our AI outputs, and it is a meaningful gap, I prioritize work that would bolster our proof in that area. Let’s just say, for example, that models don’t gauge us as being strong at technical SEO. We see that and we’re confused, because we run technical SEO projects on Fortune 500 brands’ websites.
But then, upon review, we barely talk about this on our website, none of our case studies mention it, and we don’t have any informational content covering our methods. As such, no off-page sources or review sites mention it either.
Well, what would we expect?
So the question isn’t just visibility, though obviously visibility is important (you don’t want to be invisible). The question is also whether the version of them that’s visible is the version they want buyers to see.
Want more insights like this? Subscribe to Field Notes
