Search is drying up.
Not search in the literal sense. People still need restaurants, lawyers, software, wine labels, boxing gloves and a thousand other things. Human demand has not disappeared because someone put a chatbot on top of the internet.
What is drying up is the old economic structure of search.
For 20 years, Google gave marketers a wonderfully simple bargain. Consumers reduced their messy needs into a handful of keywords. Google returned a ranked list. Businesses fought to appear at the top of it. Paid search let you buy your way into the auction. SEO let you try to engineer your way into the organic results without paying for every click.
And, for a long time, that was extraordinarily valuable.
The mistake now is assuming that because people are still searching, SEO remains fundamentally the solution…it isn’t.
AI search changes the unit of competition. We are moving from a world in which businesses competed to rank for keywords to one in which brands compete to become recommendable.
That is a much bigger change than most of the SEO industry wants to admit.
Why SEO became so economically powerful
To understand what replaces SEO, you first have to understand why SEO worked so bloody well.
The genius of traditional search was not simply that Google had a lot of users. It was that humans learned to communicate with search engines in an incredibly crude language.
You wanted boxing gloves, so you typed “best boxing gloves”.
You wanted somewhere to buy a car, so you typed “car dealers near me”.
You wanted a booking platform, so you typed “business booking software”.
These phrases were not accurate descriptions of what people really wanted. They were shortcuts.
Consumers learned that giving Google three or four useful words usually produced something good enough. And good enough mattered because clicking through 10 websites was a pain in the arse. Cookie banners. Slow pages. Pop-ups. Bad navigation. Endless generic copy.
So we satisficed.
We did not hunt indefinitely for the perfect answer. We accepted the least-worst suitable answer presented to us, usually somewhere on page one. That created the old joke that page two of Google was the best place to hide a body.
It also created a fantastic economic opportunity.
If everybody looking for a service used roughly the same handful of keywords, and almost nobody went beyond the first page, ranking highly for those phrases concentrated an enormous amount of demand.
That was SEO.
Google preferred the polite version of SEO: make your site crawlable, create useful content and follow the rules. Smart SEOs knew there were signals that moved rankings, and if manipulating those signals produced profitable visibility, businesses were going to pull those levers.
The logic was obvious. If a keyword was valuable enough to pay Google for, it was valuable enough to try to win organically.
That entire system depended on one thing:
The keyword being the starting point.
AI search removes the starting point
AI does not need you to communicate in caveman search language.
You can tell it what you actually want.
That sounds like a small UX improvement.
Economically, it is enormous.
Imagine you are launching a new wine brand. Under the old model, you might search “wine bottle stickers” or “custom wine labels”. The businesses ranking for those terms got the opportunity to sell to you, even if they were only vaguely suited to the actual job.
With AI you can say:
I need labels for a new wine brand. They need to work on a circular bottle, use these colours, be customisable, arrive next day and I would prefer a supplier that can help with the design.
Now the system does not merely need to find websites associated with “wine bottle stickers”.
It needs to understand what good looks like for this particular problem and then identify suppliers that appear capable of solving it.
That change means search changes, and with it, ranking is no longer the goal.
The consumer has more power because the consumer can finally explain the whole problem.
And the brand has less power because it can no longer rely on a broad keyword to funnel thousands of imperfectly matched prospects towards the same result.
This is why saying GEO is “just SEO” misses the point.
SEO optimised for the starting query.
AI search increasingly optimises for the solution.
The new starting point is the category entry point
If keywords stop being the fundamental unit, what replaces them?
Category entry points.
These are the situations, triggers, needs and wants that cause somebody to enter a market.
Long before a consumer opens Google, ChatGPT or anything else, another search engine has already fired up:
Their brain.
Something happens.
They get hungry. A client comes to town. Their booking software becomes unbearable. Their daughter’s birthday approaches. Their labels keep peeling off bottles. Their sales team needs a better way to record meetings.
The need appears first.
Then the consumer looks around. What do they already know? Which brands come to mind? What can they see? Which solutions are physically or mentally available?
Only then do they reach for an external search system.
This matters because businesses do not really sell products.
They solve problems that customers cannot, or do not want to, solve themselves.
Take a seafood restaurant in central London.
At first glance, the problem it solves is hunger. Fine. But that is an almost useless description of its market.
It might solve a pre-theatre dining problem, a romantic evening, a birthday, a corporate dinner, a tourist’s search for good seafood or a client meeting.
The restaurant itself does not change.
The category entry point does.
Now add the second half of the equation: performance attributes.
What is the restaurant actually good at?
Perhaps it has a private dining room, unusually fresh seafood, a strong wine list, fast pre-theatre service or a location five minutes from several major theatres.
Those attributes are the reasons the restaurant can credibly solve particular problems.
Brand marketers might call them positioning or performance characteristics. Network science might describe similar ideas as fitness signals.
I care less about the terminology than the practical consequence.
Your job is to connect the situations that bring people into the market with the attributes that make your business a good answer.
That intersection is the problem space.
Become recommendable, not merely rankable
This leads to the most important change in search marketing.
Your goal is no longer simply to rank first.
Your goal is to make your brand as easy as possible for an intelligent system to recommend.
That means giving the system evidence.
Go back to the restaurant.
If pre-theatre dining is an important category entry point, a dedicated page makes perfect sense. Show the fixed-price menu. Explain the timings. List the nearby theatres. State the walking distances. Explain when diners should book to make a 7:30pm performance.
If there is a major show running all year nearby, it might even make sense to create an asset specifically for people looking for dinner before that show.
That sounds suspiciously like SEO because some of it is SEO.
Good.
There is no prize for abandoning useful tactics simply because the industry has discovered a new acronym.
Use sensible titles. Make the page easy to extract. Structure the information clearly. Build links if they help. Make sure Google Maps knows you exist.
But the strategic reason for creating the page has changed.
You are not publishing another piece of copy because a keyword tool says 720 people a month type a phrase.
You are encoding a genuine customer situation and presenting evidence that your business can solve it.
Now take a corporate dinner.
If the restaurant has a private room for eight, soundproofing, a screen, video calling, specialist service and flexible menus, put that information somewhere machines and humans can understand it.
That is not “content”.
It is evidence of fit.
Not every category entry point deserves a landing page, either. A couple looking for a romantic evening does not necessarily need a grotesque “romantic seafood restaurant London” SEO page written by someone who has never experienced romance.
A sentence on the homepage may be enough.
Photography of couples may help.
Reviews mentioning anniversaries may help.
Social posts showing the atmosphere may help.
The point is not to create more pages.
The point is to create a clearer distributed understanding of what your brand is good for.
Your website is your home, not your universe
This is where another fashionable prediction needs dealing with: the idea that websites will disappear.
I doubt it.
Brands still need somewhere to hang their hat.
Owning a location in the information space matters in the same way owning a physical location matters. It gives you a canonical place to explain who you are, what you sell, where you operate and why anyone should care.
But ownership is not enough.
Your brand has to exist in other people’s spaces too.
Directories, review sites, YouTube, Instagram, TikTok, trade publications, maps, customer posts, partners and events all become hubs in a wider information ecosystem.
The old web encouraged humans to travel from website to website collecting information.
AI increasingly travels for us.
That is the better mental model for generative search. The system traverses a distributed universe of information, gathers evidence from multiple sources and constructs an answer.
The website therefore becomes one node among many.
This is why a business cannot think about GEO as:
“What do I need to put on my website so ChatGPT mentions me?”
That is SEO thinking wearing an AI hat.
The better question is:
What do I want to be known for, where must that understanding exist, and what evidence makes the claim believable?
You eventually need other people to talk about you
A brand is invisible until it starts talking about itself.
But after a while, talking about yourself has diminishing returns.
You can publish another post. Send another email. Upload another video. Add another page. Eventually you hit the limits of your audience, your budget, the algorithm and the simple fact that everybody in the business has other work to do.
Growth requires other people to talk about you.
That is where visibility starts compounding.
For a restaurant, this is obvious. Reviews matter. Customers photograph meals. Creators visit. Critics write. Events create stories.
For a B2B company the mechanics change, but the principle does not. Publish proprietary data. Run events. Create a genuinely interesting point of view. Produce research worth citing. Give journalists something worth covering.
Make the invisible visible outside your own channels.
This is where traditional channel thinking starts to break down.
An SEO agency will naturally prescribe SEO.
A PR agency will prescribe PR.
A social agency will prescribe social.
An influencer agency will discover, with remarkable consistency, that the answer involves influencers.
But generative visibility does not care about your agency org chart.
One company might need reviews and product data. Another needs case studies and trade press. Another needs YouTube demonstrations. Another needs partnerships, awards, event coverage and authoritative third-party mentions.
The right mix depends on the category entry points you want to win and the evidence required to prove your fitness for them.
That makes GEO harder than SEO.
And probably more expensive.
AI search changes the economics of growth
There is another consequence that nobody selling a £999-a-month GEO package particularly wants to discuss.
AI makes matching better.
That is excellent for consumers and potentially brutal for businesses that previously benefited from broad, imperfect search demand.
A SaaS company could once rank for “business booking app” and get exposure to almost every kind of company looking for booking software.
Under conversational search, the consumer can specify the exact operational context, company size, integrations, workflow, budget and use case.
The total market does not disappear.
It fragments into more precise problem spaces.
Products that are simple to understand, reproduce and distribute can still serve huge numbers of consumers.
Scalability remains a superpower.
Software demonstrated this for years because code can be reproduced at negligible marginal cost. Music travels because the recording can be distributed infinitely. McDonald’s became McDonald’s because it turned a restaurant experience into something astonishingly reproducible.
Contrast that with a bespoke agency, a hand-built kitchen or a specialist restaurant.
The more labour, artistry, complexity and customisation required to reproduce the product, the harder it becomes to scale.
AI does not repeal that economic reality.
If anything, better matching exposes it.
Simple needs will often produce simple prompts and favour businesses capable of serving enormous markets consistently.
Complex needs will produce more complex prompts and increasingly direct demand towards narrower specialists.
Specialist businesses are not doomed. A craft jewellery maker can be discovered globally. A niche SaaS tool can serve an international vertical.
But being discoverable everywhere is not the same as being infinitely scalable.
That leaves one final lever.
Fame.
Fame is the final hurdle
When you cannot become disproportionately large through reproduction, you can become disproportionately valuable through demand.
That is what brand marketing has always done.
Become known.
Become remembered.
Become associated with something useful.
Become famous enough that consumers search for you directly, mention you to others, review you, cite you, share you and create more evidence that reinforces the brand’s visibility.
The visible become more visible.
There is nothing mystical about this. Publicity, advertising, events, creators, PR and great products can all create awareness.
The tactical route varies.
The strategic objective does not.
I call this fame engineering: deliberately allocating marketing effort to increase the probability that the brand becomes disproportionately known for something commercially useful.
Not famous for the sake of fame.
Famous in relation to the category entry points you want to own and the performance attributes you want remembered.
That fame matters even more in an AI-search world because machines consume the traces it leaves behind: more customers, branded searches, reviews, press, discussion, links and people mentioning the brand in relation to the problems it solves.
Fame creates human mental availability.
And human mental availability increasingly creates machine availability too.
The new search economy
For years, search rewarded businesses that understood how to win the ranking system.
That era produced brilliant businesses, brilliant SEO practitioners and plenty of absolute horseshit too.
But the economic bargain is changing.
Keywords are losing their position as the central organising unit of discovery.
Page one is losing its monopoly over attention.
Consumers can describe their needs with far greater precision.
AI can investigate multiple sources, compare attributes, assess evidence and construct recommendations without requiring the consumer to open 15 tabs.
The winners will therefore be the brands that understand the situations that bring customers into the market, clearly define the attributes that make them a superior solution, distribute evidence of those attributes across the wider information ecosystem and invest sufficiently in becoming known.
Some tactics will look like SEO, some like PR, social, advertising, events, partnerships or old-fashioned customer service.
That is precisely the point.
GEO is not another marketing channel.
It is what happens when search stops being a neat channel and starts becoming an intelligent layer sitting across the entire market.
The companies trying to solve that with another keyword spreadsheet are going to have an interesting few years.
The job is no longer to trick a ranking system into putting you first.
The job is to make the market — human and machine — conclude that you deserve to be chosen.
Stop optimising for position.
Start engineering recommendability.
Andrew Holland
Originally published on LinkedIn on 31 August 2026.