SEO vs GEO is the wrong question
28th July, 2026
AI-mediated discovery has sparked an identity crisis in marketing.
As more buying journeys begin in AI interfaces, the industry has reached for a familiar response: another optimisation discipline, a market of tools and consultants, and endless advice about citations, prompts, Reddit, reviews and “AI visibility”.
So everybody is asking: how do we optimise for AI search, and what do we call that discipline if it isn’t SEO? But the interesting problem isn’t naming the thing – it’s that we’re (still) asking the wrong question.
The important shift is that discovery, comparison and evaluation increasingly happen before somebody reaches your website. Marketing’s job changes as a result.
In my Marketing Manifesto, I argued that marketing’s purpose is to shape what the market discovers, understands and believes. AI raises the stakes. As people delegate more discovery and evaluation to machines, marketing must still earn attention while shaping the information and evidence from which those systems form conclusions.
The consequences become clearer through three connected transitions. Together, they push marketing upstream.
From visibility to legibility
For most of the web’s history, digital marketing has revolved around visibility. We wanted to appear in search results, social feeds, marketplaces, inboxes and news coverage. Once we had attention, we had an opportunity to persuade.
Visibility still matters, but its role is changing. It supplies some of the evidence from which people and machines form an understanding of your business.
Much of this still depends on conventional search infrastructure. For current, commercial and research queries, AI systems often retrieve information from external indexes before constructing an answer. If your content cannot be crawled, indexed and retrieved, it may never enter the evidence set at all. Lily Ray has made this point repeatedly: abandoning SEO for GEO is not sophistication. It is usually a failure to understand the machinery.
In Marketing after the fog clears, I argued that AI reduces the cost of navigating information. That doesn’t eliminate the need for visibility. It changes what visibility is for.
Being mentioned in ten places isn’t inherently useful if those mentions are inconsistent, shallow, outdated or disconnected. Equally, technically perfect structured data won’t help much if nobody talks about your business beyond your own website.
What matters is whether the market, and the machines increasingly representing it, can identify and distinguish your business accurately.
Those systems aren’t simply indexing documents. They increasingly retrieve and reconcile information around entities: businesses, products, people, places and the relationships between them. Your website supplies one source among many.
The resulting representation needs to capture what you sell, who it is for, how the offer fits together, which problems it solves, where you are genuinely strong, and how independent sources describe you.
Every mention, citation, review, support article, community discussion and product page adds evidence. AI systems don’t just retrieve those signals. They reconcile them, compare them and synthesise them into recommendations.
Mike King uses relevance engineering for a broader discipline: treating visibility as an information-retrieval problem rather than a checklist. It spans query fan-out, semantic and passage-level relevance, extractability, source aggregation, simulation and measurement across a probabilistic pipeline.
The value of the term is methodological. It asks teams to form and test hypotheses about how information is retrieved and used, rather than accumulate generic “AI-friendly” tactics.
Technical SEO contributes to that. So does PR. So do product marketing, documentation, community engagement, analyst relations, customer education and every other activity that helps the market form a coherent understanding of what you are.
Visibility earns inclusion. Legibility determines what the market makes of you.
From persuasion to scrutiny
Marketing has traditionally been brilliant at representation. We learned how to position products, shape narratives, refine messaging and influence perception.
Those skills remain valuable, but reality places tighter limits on them.
AI systems don’t simply repeat the story you tell about yourself. They can test it against whatever documentation, reviews, competitors, support forums, pricing, specifications and expert opinion their retrieval systems happen to surface.
That changes the economics of persuasion.
As I explored in When scrutiny becomes cheap, representation loses leverage when comparison becomes effectively free.
Cheap comparison doesn’t guarantee competent scrutiny. These systems can mistake repetition for consensus, cite fabricated claims and confidently reproduce the sludge generated by other machines. Lily Ray’s work on the AI slop loop, and her experiments with gullible answer engines, make that risk difficult to dismiss.
The pressure on representation comes from the possibility of cross-checking, not any assurance that the cross-check is reliable. These systems still reflect the sources they can access, the weight they give them and the commercial incentives of the platforms mediating the choice.
The cheapest way to improve your marketing increasingly becomes improving your business.
None of this overturns marketing science. Brands still grow through broad reach, mental and physical availability, distinctive assets, and products that are easy to notice, remember and buy. Context, signalling and perceived value still shape choice. AI changes how those cues are encountered, combined and set against other evidence.
That may increase the value of qualities that remain difficult to manufacture at scale: judgement, craft, relationships, reputation, empathy and the experience of dealing with the business. When those qualities leave evidence, machines can help surface them.
From claims to corroboration
Every business claims to be trusted. Every product is innovative. Every service is customer-centric.
Claims have always been cheap. Verifying them was not.
Now, much of it is.
AI-mediated recommendations can draw on reviews, independent publications, community discussions, customer experiences, documentation and public data. They can compare what a business says with what customers and independent sources report.
The market infers trust from that evidence. Marketing’s task is to ensure that its claims can be corroborated by reality, customers and independent sources.
This is also the argument behind Raise the floor. AI doesn’t just reward exceptional moments. It exposes the average experience. The quality of your documentation, the consistency of your messaging, the clarity of your pricing, the tone of your customer support, the completeness of your product information, the health of your technical foundations: all of these become part of the evidence from which recommendations are formed.
Raising the floor becomes at least as important as raising the ceiling.
Durability matters too. A tactic that earns citations for a few months while weakening organic visibility, trust or the underlying site is not a marketing success. It is short-term arbitrage with a flattering dashboard. Ray’s warning that a GEO strategy can destroy the SEO it depends on should be treated as a constraint, not an edge case.
Marketing moves upstream
Marketing remains a department, but its effectiveness increasingly depends on an organisation-wide capability. Product quality, engineering decisions, documentation, support, operations, legal and governance all shape what the market can discover and conclude. Marketing can connect those functions, expose gaps between promise and reality, and help the organisation close them.
Taken together, these shifts pull marketing upstream, towards the products, services and experiences that create perception in the first place.
Where advertising still fits
Once marketing is understood as extending beyond the messages a business controls, advertising becomes easier to distinguish from the whole.
Advertising buys the opportunity to put a chosen message in front of an audience. The wider work of marketing shapes what that audience later concludes from everything else it encounters.
Advertising remains powerful. It can introduce ideas, establish positioning, create demand and, increasingly, buy privileged access to AI interfaces.
But the campaign is only one input. Customers compare experiences, journalists publish reviews, communities discuss trade-offs, and those accumulated signals shape what future customers are told.
Advertising influences what people hear. Marketing shapes what they come to understand and believe.
Which brings us back to SEO and GEO
This is why the SEO versus GEO debate starts from the wrong question.
Understanding how language models retrieve information, weigh evidence and construct recommendations is clearly valuable. Learning how to influence those processes will become an important capability.
But “AI visibility” is a slippery proxy. A system can cite your page without recommending your business, and the same prompt can produce different sources from one run to the next. Ray’s research into self-promotional listicles shows how easily citation and recommendation can diverge. Citation volume matters only insofar as it improves how the business is understood, considered and chosen.
Treating that capability as “the new SEO” still shrinks a fundamental shift in marketing into another optimisation discipline.
The more useful question is which businesses are best equipped to succeed when discovery and evaluation become machine-mediated.
Those will be businesses the market can understand accurately, whose products stand up to comparison and whose claims are consistently supported by evidence.
SEO, GEO, relevance engineering and advertising each influence part of that process. They are instruments, not the organising principle.
The identity crisis comes from confusing those instruments with marketing’s purpose.
Marketing is the work of shaping what the market discovers, understands and believes.
AI hasn’t changed that purpose. It has made the narrower version indefensible.
I was about to write an eMail to a client explaining exactly this.… now I can just send this link. Very decent and in depth analysis – as always. Or in other word: #strong