Generative Engine Optimisation is sticking as a name, and as always, that is both useful and dangerous.
Useful, because naming something allows us to talk about it seriously. Dangerous, because the moment a term stabilises, the industry rushes to operationalise it before it has properly understood the problem it is meant to describe.
GEO has followed that pattern perfectly. Influencers have arrived. Templates have appeared. Advice has been flattened into checklists about content volume, chunking, FAQs, and structured data.
Most of this advice is not just shallow. It is directionally wrong.
The mistake underpinning almost all current GEO commentary is the assumption that generative engines behave like search engines with a conversational interface.
They do not.
Search engines rank documents. Generative engines assemble solutions. And those two activities rely on fundamentally different signals.
SEO has ranking signals. GEO has fitness signals.
Until that distinction is understood, GEO will continue to be treated as SEO cosplay.
And in this article, I’ll explain it all.
Side note: I’m speaking at Mads Singers amazing event SEO Mastery Summit in March. Grab your tickets as I’ll be delivering my best talk yet.
Why SEO works, and why that logic breaks
Search Engine Optimisation works because performance is bounded and observable.
You can see rankings. You can see page one. You can measure movement over time. You can compare competitors answering the same query.
Because outcomes are visible, search engines can design systems that reward documents based on relative performance.
That is why ranking signals (our industry term is factors) exist.
Headers, page titles, internal links, backlinks, anchor text, topical relevance, and authority are not arbitrary. They are proxies that allow a system to compare documents competing for a limited number of positions.
Even concepts like EEAT, often misunderstood and mystified, ultimately serve this same function. They are algorithmic attempts to approximate trust and quality in a ranking environment where documents must be ordered.
SEO is therefore an optimisation discipline. You optimise a page to win a position.
The moment you remove rankings, queries, and visible competition, that entire logic collapses.
Generative engines operate without bounded performance
Generative engines do not rank pages into positions. They are responding to problems.
Most of the problems they are asked to solve do not have clear, measurable outcomes. There is no universal scoreboard for “best hotel for a quiet anniversary trip” or “best SEO agency for early-stage SaaS.” Success cannot be directly observed. Performance cannot be verified in advance.
This is exactly the situation described by network scientist Professor Albert-László Barabási in Linked and later formalised in The Formula.
Barabási’s core insight is simple and devastating: when success is easy to measure, performance dominates. When success is hard or impossible to measure, networks take over. Popularity, reputation, and visibility begin to outweigh intrinsic quality.
He summarises this brutally. The harder success is to measure, the less performance matters.
Generative engines live almost entirely in this unbounded world.
They must recommend without being able to observe outcomes. They cannot see whether a product actually worked. They cannot measure long-term satisfaction. So instead, they infer quality through the network they traverse through search strategies.
That inference relies on fitness signals.
What fitness signals actually are
Fitness signals are not branding language or marketing spin.
In network science, fitness describes the intrinsic ability of a node to attract attention, links, and adoption over time.
Fitness determines which nodes grow, which stagnate, and which disappear. Crucially, fitness is not the same as visibility.
A node can be highly fit and unknown. It can also be highly visible and weak.
In markets, fitness signals come from two inseparable sources.
First, performance attributes. These are the concrete characteristics of a product or service that determine how well it solves a particular problem.
Second, reputational signals. These are how those performance attributes are validated, repeated, and amplified by the network.
Reputation is not separate from fitness. It is fitness made visible.
This distinction matters because many marketers treat reputation as decoration. In networks, reputation is structural.
A real buying problem, not a search query
This became obvious to me through a very ordinary buying decision.
My daughter boxes. She is thirteen, spars regularly, and wanted pink boxing gloves. That immediately introduces constraints that no keyword captures. The gloves needed to be suitable for sparring, sized for smaller hands, well made, and fastened with Velcro rather than laces. Colour mattered, but it was not the problem.
This is where traditional search starts to feel clumsy. You can search for “pink boxing gloves” and refine endlessly, but what you are really doing is translating a multi-dimensional problem into brittle keyword combinations and hoping the right page appears.
Instead, I asked a generative system to solve the problem directly.
Why the boxing gloves example matters
What surfaced was not driven by keyword usage or page structure. It was driven by fitness signals.
Well-known brands with reputations for quality. Clear statements about glove weight and materials. Explicit references to sparring use. Size guidance for younger fighters. Reviews that mentioned durability, comfort, and wrist support.
The system did not “find pages about pink boxing gloves.” It assembled options that appeared fit for the defined problem.
Importantly, this judgment was made without any ability to measure actual performance. The model could not see my daughter spar. It inferred quality through the network. Descriptions, reviews, brand familiarity, and repeated references across sources.
That is fitness in action.
Interestingly, I ultimately bought from a brand I already knew and trusted. That decision reinforces the same principle. Familiarity and accumulated reputation are powerful fitness signals, even when alternatives are objectively strong.
Performance attributes as core fitness signals
Most product and service pages are poor at communicating performance attributes.
They are sparse. Generic. Written to satisfy templates rather than humans. The reason is economic. Copy costs money. SEO pages are monetised by traffic from rankings, not persuasion. Humans rarely go past page 1 of Google, which is why being there is so powerful.
But generative engines are not looking for relevance matching. They are looking for problem solving.
Performance attributes are the raw material.
Price, speed, durability, convenience, suitability, design, reliability, scale, expertise, described usage and users. These are attributes that matter in context. “Best” is not an attribute. It is a conclusion.
If those attributes are not clearly articulated, fitness cannot be inferred.
Reputation signals are fitness signals
This is the point that must be made explicitly.
Reputation signals are not secondary. They are fitness signals.
Reviews matter because they repeat performance claims. Media mentions matter because they validate relevance. Awards matter because they concentrate attention. Listicles matter because they aggregate choice.
Barabási calls this preferential attachment. Nodes that receive attention are more likely to receive more attention. This is not unfairness. It is how networks scale.
Generative engines rely on this dynamic because they must. When performance cannot be measured, the network becomes the judge.
Fame engineering, not optimisation
This is where the concept I refer to as fame engineering becomes essential.
Generative Engine Optimisation is not really optimisation at all. It is the deliberate engineering of fitness signals and reputational signals so that they propagate through the network.
It is the maximisation of reputation.
But nothing happens until something moves.
Journalists do not hunt stories. Influencers do not discover brands by accident. Visibility must be created. Events, PR, content, social, speaking, reviews, partnerships. These are not tactics. They are transmission mechanisms.
A restaurant with great food and views does not win because it exists. It wins because those attributes are repeatedly demonstrated, discussed, and shared across the network.
That is fame engineering.
Why GEO is not just SEO with a new name
This needs to be stated plainly.
You can rank without fitness. You cannot be recommended without it.
SEO allows mechanical optimisation. GEO does not. It is constrained by reality in a way SEO often is not.
This is also why SEO-first pages often fail in generative responses. They are optimised to match keywords, not to solve problems. The information required to infer fitness is missing, because including it would sometimes harm rankings, or simply the SEO brief was all about keywords.
This tension will not disappear. It will need to be managed.
Of course, fitness exists in the world; it’s just that LLMs are designed to uncover it. These principles have been on display long before LLMs came along.
Harry Potter and the mechanics of success
This example illustrates the mechanics of success better than almost anything else, precisely because it is so often misunderstood.
Harry Potter is now treated as an inevitable cultural phenomenon. A global franchise. A guaranteed success. But at birth, it was almost invisible. When Harry Potter and the Philosopher’s Stone was first published in the UK in June 1997, the initial print run was just 500 copies. Of those, around 300 were sent to libraries, largely for free. That is not market domination. That is obscurity.
What Rowling had at that point was not visibility. She had fitness.
The book itself worked. Children read it and did not put it down. Librarians noticed borrowing queues. Teachers recommended it to parents. Early reviewers described something rare: a story that children actively chose to keep reading. None of this produced immediate sales at scale, but it produced something more important in network terms.
It produced signals.
One reader became another. One review led to another recommendation. One library order led to another library order.
This is the slow, granular process Barabási describes repeatedly. One node at a time. One connection at a time. Fitness expressing itself through accumulation rather than explosion.
For almost a year, nothing dramatic happened. The book did not top charts. It did not dominate headlines. But beneath the surface, preferential attachment was beginning to operate. Each positive signal increased the probability of the next signal. Each mention made the next mention more likely.
When the book was released in the United States in September 1998, it was largely ignored by the mainstream media. It took another year before Harry Potter reached the top of the New York Times bestseller list in August 1999. And once it arrived, it stayed there for nearly eighteen months. Only a structural intervention, the decision to split the list into children’s and adult categories, finally displaced it.
What looks like sudden success was, in reality, delayed network recognition.
The key point is this: the performance of the book did not change. The writing did not improve. The story did not evolve. What changed was the density of reputation signals surrounding it. Fitness remained constant. Visibility caught up.
This is exactly how unbounded success works.
The Robert Galbraith counterfactual
Barabási uses a second story to demonstrate the other side of the same principle, and it is even more instructive.
Years later, J.K. Rowling published crime novels under the pseudonym Robert Galbraith. The books were competent. Some critics argued they were very good. But they failed commercially. They were rejected by publishers. One rejection famously suggested the author join a writing group.
This was not a failure of performance. It was a failure of reputation.
The fitness was there. The network signals were not.
Once Galbraith was revealed to be Rowling, sales spiked instantly. Overnight, the books became bestsellers. Nothing about the text changed. The only variable that changed was the reputation signal injected into the network.
Barabási is clear about what this proves. Success is not just about being good. It is about being recognised as good within a network that cannot directly measure quality.
Why this matters for generative engines
Generative engines behave exactly like this network.
They cannot observe outcomes. They cannot measure satisfaction. They cannot know whether a recommendation worked. So they do what networks always do when success is unbounded.
They look for fitness expressed through reputation.
They count mentions. They detect repetition. They observe which options are referenced across sources, contexts, and domains. They infer quality not from performance itself, but from the density and consistency of signals around that performance.
This is why reputation is not a secondary consideration in GEO. It is not branding polish. It is not optional.
Reputation is fitness made legible to the network.
Fitness must exist, or reputation will not sustain itself. But without reputation, fitness remains invisible.
That is the lesson Harry Potter teaches. And it is the same lesson generative engines are now enforcing at scale.
What this means in practice: copywriting, PR, and engineering fame
If all of this feels uncomfortable, that is because it removes the safety blanket of mechanical execution.
Generative Engine Optimisation is not something you can outsource on the cheap. It is not solved by publishing more blog posts, adding more FAQs, or restructuring content into smaller chunks. Those tactics misunderstand the nature of the problem. GEO is about whether a network recognises you as fit to solve a problem. That recognition has to be earned, displayed, and repeated.
If you claim expertise, you must demonstrate it in public.
Expertise that is not visible to the network does not exist. Large language models do not intuit authority. They detect it through repetition. Quotation. Reference. Citation. Mention.
If you want to be known for your business expertise, you have to become a source. That means being quoted. Being referenced. Being used as an input into other people’s explanations.
This is where PR stops being a “nice to have” and becomes structural.
Every media quote is a reputational signal. Every product review is a reputational signal. Every influencer demonstration, social post, podcast appearance, or case study is another node connecting your brand to a specific set of performance attributes.
One mention at a time. One reference at a time. This is how fitness propagates through a network.
Links still matter, but not as trophies. They matter because they are evidence of attention and doorways to your business.
Copywriting for humans, not for machines
The same logic applies to on-page copy.
Your copy must explain, not merely align. It must describe performance attributes clearly enough that a human, or a model trained on human language, can understand why you are a good solution for a specific problem.
This is where things become genuinely difficult.
Writing copy that persuades humans often conflicts with writing copy that ranks cleanly in search engines. Being explicit about suitability, limitations, use cases, and trade-offs can dilute keyword focus. Descriptive explanations do not always help ranking algorithms that still rely heavily on relevance matching.
SEO is relevance engineering. GEO is fame engineering.
Knowing that trade-off exists is now part of the job.
The answer is not to ignore search engines. It is to manage the tension deliberately. That requires higher optimisation skill, not less. It requires understanding which pages can absorb richer descriptive copy, where reputational signals can compensate for ranking softness, and how PR can support visibility when pure SEO cannot.
This is no longer about page templates. It is about judgment. And you need humans for this.
Building reputation one node at a time
Reputation is not built in campaigns. It is built incrementally.
One review. One quote. One recommendation. One reference.
Barabási is explicit about this. Preferential attachment does not require explosive growth. It requires consistency. Each new signal slightly increases the probability of the next signal. Over time, the curve bends.
This is why patience matters. This is also why denial is dangerous.
Generative engines absorb vast amounts of information quickly, but reputation still compounds at human speed. The businesses that start now will benefit later. The ones waiting for a mechanical shortcut will not.
This is not a service tweak. It is a structural shift.
It is important to be clear about what GEO is not.
It is not a new SKU. It is not a bolt-on service. It is not a content production engine.
It is a shift in how visibility is earned.
Search is already moving in this direction. Google has made it clear that AI-driven results will increasingly incorporate personalisation, memory, and contextual understanding. This mirrors what systems like ChatGPT and Perplexity already do. Results are not just answers. They are filtered through history, preference, and inferred intent.
As memory and personalisation increase, reputation becomes even more important. Being known for something consistently across the network is what allows systems to match you to the right problems, for the right users, at the right time.
Generative optimisation will not be optional. It will become the default environment in which visibility is judged.
The future of SEO careers
This is the part that many in the industry would prefer not to confront.
SEO careers are changing shape. Mechanical execution is being commoditised. Strategy, analysis, and judgment are not.
SEOs are uniquely positioned to sit at the centre of this shift. They understand signals. They understand networks. They can see where performance attributes are unclear, where reputation is thin, and where amplification is failing.
But that only works if they accept what is happening.
Denial is not a strategy. Pretending GEO is “just SEO” is not protection. The role is moving upward, not sideways. From execution to orchestration. From ranking pages to engineering fitness and reputation across the network.
Those who adapt will be invaluable. Those who do not will be bypassed.
The enduring lesson
Generative Engine Optimisation feels new because the interface is new.
The underlying logic is not.
Networks have always worked this way. Reputation has always mattered most where success is hard to measure. Fitness has always needed visibility to express itself. As Barabási’s work shows, these dynamics have governed success for centuries.
Large language models have not invented this. They have simply made it impossible to ignore by traversing the digitised version of human networks at scale.
GEO is not a trend. It is the network revealing itself.
And once you understand that, the path forward is no longer mysterious. It is just demanding.
What GEO actually demands
Generative Engine Optimisation demands that businesses stop optimising for platforms and start engineering their place in the network.
It requires clarity about performance attributes. It requires deliberate reputation building. It requires patience.
Above all, it requires abandoning the illusion that visibility can be engineered without substance.
This is why GEO feels uncomfortable. It shifts value away from mechanical optimisation and toward strategy, copywriting, PR, and fame engineering.
But once you understand fitness signals, GEO stops being mysterious.
It becomes obvious.
Andrew Holland
Originally published on LinkedIn on 23 January 2026.