The SEO industry doesn’t want to change.
This is not surprising. Industries rarely do. But the speed at which search is moving means the people who resist the longest will pay the highest price.
I should declare my position upfront.
I spent years building a following around the value of SEO. I published a book with that title. I am not some outside critic lobbing grenades. I am someone who knows this industry well enough to tell you that its current response to generative AI is, mostly, wrong.
The popular line is that GEO is just SEO with a new label. It is a comfortable position.
It requires no change, no new skills, no uncomfortable questions about the future of your role. It is also false.
This guide to building a GEO strategy will help you to get your head around the future.
Read it. Dwell on it. Absorb it. Share it. Talk about it with your team. Prod it. Poke it.
The goal is to give you options for thinking differently.
What follows is a series of thoughts on GEO, what it is, why it differs, what works, what doesn’t, and why.
Along with a super-quick, practical operating model for SEOs and brands to move GEO in the right direction.
Dive in.
How search actually changed
For twenty years, the commercial logic of search was simple. Rank in Google. Build links. Publish content. Target keywords. Improve technical performance. Do enough of it well enough, and traffic follows.
That game is over.
A buyer today does not type “best SEO agency Manchester” and scroll ten blue links. They open ChatGPT, Gemini, Claude, or Perplexity and describe the actual situation:
“I run a Shopify brand in the UK. Organic traffic has dropped. I need an agency that understands e-commerce SEO and can help recover commercial pages.”
That changes everything about the task.
In traditional search, a buyer compared ten results. In AI search, they receive a shortlist. The model has pre-filtered the market on the buyer’s behalf. The consideration set collapses before the buyer has clicked anything.
That is why GEO matters.
Not because it is a new discipline to put on your LinkedIn headline, but because the commercial mechanism of search has fundamentally changed.
What GEO actually is
Generative Engine Optimisation is not about chasing citations, formatting content for language models, or sprinkling pages with so-called AI-friendly wording.
It is about increasing the probability that your brand is surfaced and recommended when a buyer describes a problem you genuinely solve.
That is a different objective from classic SEO. It requires a different strategy.
The current conversation is off track because people are measuring the wrong thing. They count citations. They celebrate being footnoted by a model as though that were commercial success. It’s not.
A citation is not a recommendation. Being visible inside an LLM is not the same as being available in a buying moment.
What matters is AI availability: the likelihood that a model recommends your brand in a buying situation. Not whether the model has seen you. Whether it can retrieve you, interpret you, and justify selecting you.
A brand can be present in a model’s background knowledge and still not be recommended. It can contribute information while the commercial value goes entirely to a competitor.
If your GEO thinking stops at visibility, you are measuring awareness while missing recommendations entirely.
Recommendation is what drives revenue. Not citation volume.
Why brands are not being recommended
Most brands have two obvious problems and two that are far more important.
The first obvious problem is positioning.
Most websites still speak the language of old SEO. They are filled with informational content, generic service copy, vague claims, and pages that circle a topic without clearly stating what the business does, who it does it for, and why it is better than alternatives. A buyer arrives with a problem. The site responds with category mush. Thousands of words that were added to please an algorithm that no human reads.
The second obvious problem is distribution.
Most brands have not built enough signals outside their own site. They have not earned mentions in the places buyers trust. They have not distributed their proof, perspective, or performance attributes across the web in a way that gives models confidence.
The third problem is more important than either of those. Most brands do not clearly articulate what network scientist Professor Albert-László Barabási calls their fitness signals.
Fitness signals are the attributes and performance characteristics that enable a brand to outperform competitors in a network. Most brands either bury them, imply them, or never state them at all.
A brand may genuinely be better than competitors in meaningful ways. But if those performance traits are not clearly and repeatedly articulated, they are invisible to the model.
The fourth problem is the trap that most SEO content falls into. It creates retrievability without creating recommendation confidence. It helps a model know you exist. It does not help a model infer why you should be chosen.
Topical adjacency is not commercial distinctiveness. A model may find your domain but not trust your brand as the answer. You can appear in AI outputs and still never make the shortlist.
This is why analysing content embeddings and semantic relevance scores is clever but commercially useless. This isn’t what the model needs to recommend you.
GEO is not SEO with a new name
There is overlap. Both care about pages, content, authority, and reputation signals. But the strategic frame is different.
SEO optimised for rankings. GEO optimises for recommendations.
In classic SEO, the interface gave the user a set of options. You fought to occupy a slot. The user still did the filtering.
In AI search, the model plays the active role. It interprets the need, searches across multiple sources, synthesises distributed evidence, and forms a probabilistic judgment about what to present. The user receives a shortlist, not a shelf.
The goal shifts accordingly. It is no longer just to be indexable, crawlable, and relevant to a keyword. The goal is to be legible, credible, and retrievable in relation to a problem.
That is why GEO pulls in disciplines that classic SEO often treats as adjacent: positioning, copywriting, PR, brand building, social proof, category strategy, content marketing, expert commentary, reviews, and third-party validation.
The future of search is not more technical SEO. In many ways it is a return to something older and broader: promotion.
As the interface shifts from blue links to synthetic recommendations, the brands that win are not just the ones with optimised pages. They are the ones whose reputation, proof, and positioning have been distributed widely enough for the model to trust them.
The two forces that drive recommendations
Most generative recommendations are shaped by two things.
The first is constraint fit.
When a buyer provides detailed context, the model needs a brand that closely matches the stated need. If someone asks for a specialist agency for a UK ecommerce brand on Shopify, broad fame alone is not enough. The brand needs precise, extractable signals that it serves that audience, solves that problem, and has proof in that area.
The second is fame.
When the prompt is broad or vague, the model falls back on the brands with the richest and most repeated network of public signals. The better-known, more corroborated, and more widely referenced and distributed brands gain the advantage.
This creates the central tension in GEO strategy.
Specific positioning helps you win constrained prompts. Fame helps you survive broad ones. If your positioning is too vague, you disappear in detailed searches. If your fame is too weak, you disappear in broad ones.
The best GEO strategies build both.
And this is why you can rank on page and still not be recommended. Because you just didn’t display the right signals for the model to match you as the best solution.
Positioning is the foundation
Every GEO strategy starts with positioning. Not mission statements. Not brand values. Not purpose narratives.
Positioning in a GEO context is the extractable answer to four questions:
1. What problem do you solve?
2. Who do you solve it for?
3. Why are you better than the alternatives?
4. What proof do you have?
These are not branding flourishes. They are selection inputs. When a model is working out whether your brand belongs in an answer, these are the signals it needs to recover. If they are clear, repeated, and supported, you become easier to recommend. If they are hidden, generic, or absent, you become difficult to interpret.
Most businesses do have fitness signals. They just bury them.
Their homepages lead with generic category labels. Their service pages list features rather than reasons for selection. Their copy was written to appear relevant to keywords rather than to answer the buyer’s real question: why should I choose you?
That has to change.
A buyer and a model are both trying to reduce uncertainty quickly. If a page immediately states the problem, the audience, the advantage, and the proof, you are far easier to retrieve and recommend.
GEO is an ambiguity reduction exercise
This is one of the most important ideas in the field and the one most people miss entirely.
Models do not recommend brands because they have seen the brand name in isolation. They recommend brands when the web supplies enough mutual information to reduce uncertainty about what that brand is good at. This is information entropy applied commercially.
Your brand needs to co-occur repeatedly with the right problems, use cases, audiences, product traits, proof points, outcomes, and contextual entities. A generic mention is weak. A repeated association between your brand and a specific capability, audience, and commercial result is much stronger.
This is why vague brand building has limited GEO value. Not all mentions are equal. What matters is whether the signal reduces uncertainty in relation to a buying decision.
Fitness signals must not just be stated on your own site. They must be repeated, corroborated, and reflected across the wider web. This is where GEO moves beyond page optimisation and into network construction.
If your site says you are a specialist for e-commerce brands, but no case studies, trade press features, customer reviews, podcasts, or third-party mentions reinforce that claim, the signal is weak.
If the wider web keeps saying the same thing in different ways, the signal becomes much stronger. LLMs do not need exact duplicated phrasing. They need enough overlapping evidence to infer a stable truth.
Fame is not vanity. It is a recommendation advantage.
Large language models cannot directly verify quality the way a human can. They cannot eat the meal, use the software, or experience the service. They infer likely quality from the distributed signal of the web. They look at what has been said, repeated, linked, reviewed, discussed, cited, and associated. The network becomes a proxy for judgment.
This is where fame functions as a commercial advantage, not a soft branding concept. When a model is uncertain, the more a brand is repeatedly seen, mentioned, shortlisted, discussed, and associated with a category or problem space, the easier it becomes to select under uncertainty.
Network theory explains why this compounds.
In competitive networks, new attention flows toward nodes that already have more attention. Mentions attract more mentions. Citations attract more citations. Coverage attracts more coverage.
The brands with richer, more coherent, more corroborated signal networks become easier to retrieve and easier to justify. Excellent brands without distributed signal are difficult for a model to see.
The mechanism for building this is what I call fame engineering.
Fame engineering: systematic, not random mentions
Fame engineering is the deliberate, systematic distribution of your performance attributes across the web in ways that make your brand easier to retrieve, interpret, and recommend.
Not PR for vanity. Not chasing headlines. Not viral content that has nothing to do with what you sell. Repeated, credible co-occurrence between your brand and the traits you want to be selected for.
That requires asking better questions than most marketing teams ask:
– What are the three things we genuinely do better than competitors?
– Where do our buyers encounter information that shapes their view of providers like us?
– Which publications, podcasts, communities, review environments, comparison pages, and knowledge sources are likely to influence model output in our category?
– How do we get our strongest attributes into those environments repeatedly?
For one brand, that means trade press, case studies, award wins, data studies, expert commentary, and founder-led thought leadership. For another, it means product reviews, creator coverage, customer-generated content, community visibility, and list placements.
The channel mix varies. The principle does not.
You are trying to make the web say the same useful, specific things about your brand often enough, clearly enough, and credibly enough that the model keeps recovering the same conclusions.
And the right context matters.
A million views mean nothing if they are not connected to the buying moments that matter commercially. Relevant fame matters more than generic fame.
The question is not “How do we get attention?” It is “How do we become retrievable in the moments where buyers are trying to solve the kinds of problems we are built for?”
Content strategy under GEO
Most content strategy is still trapped in old SEO logic. It asks: what keywords have search volume?
A GEO content strategy asks a different question: what buying situations do we need to become associated with?
Content should not merely attract visitors. It should increase the probability that a model understands your brand as a credible solution to a defined problem. The centre of gravity shifts from topic content to problem content.
Old SEO content often covers topics adjacent to the business but weakly connected to selection. It may bring in traffic and build topical relevance. It rarely makes the brand more recommendable.
GEO content should diagnose the problem clearly, describe the context in which it occurs, explain what a good solution looks like, surface the relevant performance attributes, and connect the problem back to the brand in a commercially legible way.
In practice, the strongest GEO content types are:
Buyer guides. Help buyers evaluate options in your category. Models surface these when users ask how to choose between providers or what criteria matter.
Case studies. Reduce uncertainty by showing problem, intervention, and outcome – especially powerful when tightly aligned to a defined audience or use case.
Problem-centred content. Deep content about the issues your buyers face. Not broad educational filler, but content that maps the actual problem space around your offer.
Authority content. Original data, expert analysis, frameworks, and commentary. Useful because it supports PR, strengthens social distribution, and earns the kind of mentions that compound.
For every content asset, you need it to help solve a user’s problem or build fame for your brand.
– What buying situation is this for?
– What problem is the reader trying to solve?
– What performance attributes make us preferable, and are we talking about them?
– What proof reduces uncertainty for the buyer?
– What added fame does this content build for us?
– Does this content sell us as a solution?
All GEO content should function as content marketing for the business.
Building fame, building preference, building fluency. Not topical keyword fodder. The content should act as advertising, not as a keyword-indexed knowledge base.
Pages built for retrieval and selection
Traditional on-page work does not disappear. It evolves.
A homepage, service page, or product page should make retrieval and selection easy. That means clearly expressing the category, the audience, the problem, the differentiator, the proof, and the context in which the brand is a good fit.
A generic service page is often invisible to constrained prompts because it lacks the detail to match the buyer’s actual situation. A well-positioned page can win because it narrows the ambiguity.
If you work especially well with legal, finance, healthcare, or ecommerce brands, say so. If your product is built for a specific user group, say so. If you solve a particular problem better than the category norm, say so. If your advantage is speed, clarity, experience, or proof, surface it.
What you claim must be true and ideally corroborated. GEO does not reward made-up specificity. It rewards specificity that survives synthesis.
How to measure GEO properly
Most GEO reporting is currently useless. Citations are not the metric.
Track recommendation presence instead, across seven dimensions:
1. Recommendation rate. How often is your brand named across commercially realistic prompts? Real buying situations. Real constraints. Not broad informational queries.
2. Constraint performance. How do you perform in broad prompts versus constrained ones? This reveals whether the issue is weak fame, weak positioning, or both.
3. Attribute recovery. When the model names you, does it recover the right fitness signals? If it mentions you but describes you vaguely or incorrectly, the signal architecture is weak.
4. Source contribution. Which source types appear to influence recommendation in your category? Tier 1 media, trade press, review sites, community forums, comparison sites, podcasts – all play different roles.
5. Competitive adjacency. Which brands are repeatedly co-mentioned with yours? This reveals the shortlist you are actually in, not the one you think you are in.
6. Problem-space coverage. Which buying situations produce your inclusion and which do not? This shows where new content, better positioning, or stronger fame engineering has the highest payoff.
7. Signal production. How many genuinely relevant mentions, proofs, placements, and context-rich references did you generate this month? Activity is not the goal. But signal production precedes recommendation movement.
Measure whether you are getting chosen, where, against whom, and with what traits attached. That is far more useful than knowing you were cited seven times in non-commercial answers.
Traffic still matters too. Not as a vanity metric but as a commercial lens.
What you are getting clicked on, where you are getting clicked, whether brand search is growing – all of it matters. Commercially relevant traffic means commercially relevant pages are being found.
Recommendations lead to traffic. Citations do not.
A practical GEO operating model
Most GEO writing stays theoretical. Here is what the practical version actually looks like, using a specialist accounting software company for recruitment agencies as the example.
Step 1. Define commercial problem spaces. The team lists typical buying scenarios: “An agency owner wants to automate payroll for temporary staff” or “A mid-size agency is switching from spreadsheets due to inaccuracies.”
Step 2. Map pseudo-prompts. Draft realistic prompts a buyer might use: “Best accounting software for recruitment agencies with temp staff.” “Payroll automation tools for UK staffing businesses.”
Step 3. Test recommendation presence. Run these prompts through ChatGPT and Gemini. Note which brands appear, which do not, and what attributes the model attaches to each.
Step 4. Extract true fitness signals. Through analysis, identify the brand’s core strengths: compliance expertise, automated payroll integration, dedicated onboarding for recruiters.
Step 5. Rebuild priority pages. Update homepage and product pages so the positioning is immediate: “Trusted compliance automation for UK recruitment firms.” Support every claim with client results or data.
Step 6. Build problem-led content. Create detailed guides such as “How recruitment agencies can automate payroll and ensure compliance,” with real case studies showing before and after outcomes.
Step 7. Run fame engineering activity. Pitch stories to recruitment industry media, secure podcast appearances, generate customer reviews on relevant platforms, share data studies built around your key proof points.
Step 8. Measure movement by prompt cluster. Three months later, run the same prompts again. Track whether the brand now appears, whether attributes are described correctly, and whether recommendation frequency has shifted.
That is the operating rhythm. Not hacks. Not gimmicks. Not prompt stuffing.
Positioning, proof, and promotion – organised around recommendation.
Traditional SEO still matters. Its role changes.
Technical health still matters. Crawlability still matters. Internal linking still matters. Strong pages still matter. But the emphasis changes.
Keyword research becomes problem-space research.
On-page optimisation becomes positioning work.
Link building becomes part of a wider fame engineering programme.
Content strategy becomes more commercially selective.
Reporting expands from rankings and traffic into recommendation presence and source contribution.
SEO does not die in a GEO world. It gets absorbed into a larger system. SEO will only die as it is currently sold when the rankings cease to be of value.
What dies is content-focused SEO. That ship has sailed.
The SEO who thrives here will not just be a technician. They will be someone who can interpret model behaviour, understand why certain brands get surfaced, map those outcomes back to positioning and signal structures, and direct content, PR, and social activity accordingly.
That is a more strategic role than SEO has typically occupied.
It is also far more valuable. The question is whether the industry will take it.
The new game
GEO is not SEO renamed. It is what happens when recommendation becomes the interface.
In that environment, weak positioning disappears. Generic content loses power. Citation metrics become easy to game and easy to misread.
Fame becomes economically decisive again. Brands with clear fitness signals, distributed proof, and strong co-occurrence across the web become easier to retrieve and easier to trust.
The brands that win will not be chasing every prompt variation or celebrating every LLM footnote. They will be the ones who understand the actual task.
Be easy to interpret. Be easy to justify. Be known for something specific. Distribute the evidence widely enough that the model keeps finding the same conclusion.
That is GEO.
Not content for its own sake. Not visibility for its own sake. Not citation for its own sake.
Recommendation.
That is the new game.
And are you playing it?
Andrew Holland
Originally published on LinkedIn on 25 March 2026.