What you need to know

• Google’s US$600bn search ad model is being disrupted as users shift from typing queries to asking AI-powered LLMs (large language models) for direct answers, cutting out the traditional search-results-and-clicks economy.
• But a customer behavioral shift also mean an ad model shake-up. Instead of searching, users now expect suggestions. That undermines keyword bidding, SEO, and traffic funnels, all core to Google’s dominance and to how brands have planned digital campaigns for decades.
• Per Benedict Evans: “Trillion-dollar optimisation industry built around answering the wrong question.” LLMs change the nature of the question itself, and with it, the rules of engagement for marketers, platforms, and publishers.
• WPP and others are building AI-native infrastructure to automate creative and content production at scale, bypassing the need for traditional keyword- or query-based planning.
• Intent is collapsing inward. AI allows consumers to ask open-ended, previously unsearchable questions like “What should I take on a picnic?”—creating a new, suggestion-led funnel and redistributing brand visibility dynamics.
• Platforms in scramble mode with Meta, Amazon, TikTok, and Google now embedding generative AI into their ecosystems to retain control. But with lower barriers to entry and capital as the only moat, strategic defensibility is fragile.
• Brands must shift from campaign-thinking to system-building: training AI models, integrating automation, and adapting to new interface layers where visibility is mediated by machines.
• As LLMs become the new interface layer, Evans likens the shift to the dawn of the internet or AWS—those who control the interface may control the next era of value capture

A quiet revolution is underway – not on the surface of the web, but in the interface itself. Generative AI is hijacking user behaviour, replacing typed queries with direct answers, and dismantling the mechanics that made Google the most powerful company in advertising history. No blue links. No clicks. No bidding wars. Just one answer – and it's not coming from Search.

Benedict Evans, keynote speaker at last week's Meta Festival in Sydney, says it is a platform shift. For traditional search, it may be an extinction event.

Google’s search engine has long sat atop the digital ad economy like a cash-guzzling colossus, monetising every question with a neat stack of clickable, bid-laden links. That’s a US$600 billion business model, and it’s under siege.

Evans argued that an important platform shift is underway. And this time, the user isn't searching. They’re asking. “Instead of looking at the fridge and saying, ‘Well, this is what I’ve got, what can I make with this?’ and going to Google,” he explained, “you take a picture of your fridge and give it to an LLM."

That’s not a search. That’s a suggestion. And it cuts traditional Google search, and the entire paid search economy, out of the equation. Google of course is also a pioneer in LLM search so it's not exactly panicking – yet. 

Behavioural Pivot

For two decades, digital search has been an intent goldmine. A user types a query. Advertisers bid for visibility. Publishers optimise for rankings. But LLMs flip this.

"Find me a chocolate chip cookie recipe,” Evans said, “and the problem, of course, is there's a million chocolate chip cookie recipes on the internet, because that’s how search works. You just go to the LLM and say, 'Give me the recipe.'"

No links. No clicks. No bidding war. This seemingly benign shift in user behaviour, away from queries toward direct Q&A, threatens the core mechanics of the search ad model.

“If I just go to an LLM and ask it, ‘How should I decorate my apartment?’ or ‘What lamp should I buy?’ then you get a whole different bunch of answers,” Evans warned. “We’re going to get a whole bunch of new questions, and those are going to change different industries in different ways.”

Platform or parasite?

Google has been here before. Its own featured snippets already skimmed traffic from publishers. But LLMs go one better, or worse, depending on your perspective, they answer the question completely.

There’s a trillion-dollar optimisation industry built around answering the wrong question, said Evans. “And all of that is now going to get reconfigured.”

Alphabet still generates over half its revenue from search advertising. A drop in search queries, even by a few percentage points, would ripple through media budgets globally. But the dilemma facing Google is its business is still structurally aligned with the click economy, even as users drift toward conversational commerce and zero-click resolution.

Budget drift 

Marketers aren’t waiting for the bottom to fall out.

“There are AI questions that are about bottom line innovation. But then there are [questions] about top line innovation, and that's always how new technologies get deployed to begin with. You try and absorb it.” said Evans. “Big technology companies make it a feature. Google and Microsoft spray it all over their products. Every text box on the internet gets an LLM.”

That retooling is already underway. WPP is building out an AI-native creative infrastructure to generate ad variants, translations, and asset libraries at scale. “What would you do if you had 50 interns? Or a thousand?” Evans asked. “There’s a whole layer of automation that’s going to sweep through the cost of marketing.”

And it’s not just about cost.

This new wave of interfaces lets brands leapfrog the query altogether. “You want to ask, ‘What would I buy to take on a picnic?’ That’s not a high-intent query you could have really asked on Amazon or Walmart before,” said Evans. “An LLM lets you ask those kinds of questions.”

The effect: intent is collapsing inward. Search is giving way to suggestion. Brands are adjusting accordingly.

Business models wobble

The race to adapt is well underway. Meta is investing heavily in AI-driven discovery and recommendation. TikTok already operates on that model. Amazon is folding generative interfaces directly into shopping journeys. Google is threading Gemini through Workspace, Docs, and core search, desperate to keep users inside its ecosystem.

But Evans warns: hardware and infra investment may not guarantee strategic control.

“If you’ve got $100 billion and a bunch of video chips, you can have a frontier model,” he said. “And a half billion dollars isn’t nothing, but it points to the fact that there are no fundamental barriers to entry here.”

He’s blunt about where the moat currently lies: “The only moat we do have is access to capital.”

Strategic Implications

So what now for brands?

First, understand that the funnel is being atomised. Visibility will no longer depend on keyword bids or organic rankings, but on whether your brand appears in the AI’s response.

Second, it’s not about optimisation. It’s about training. “You find something that people do in email or Excel or Salesforce,” said Evans, “and you turn that into a company.” The same applies to marketing: Everything that was previously manual, repetitive, or statistical can now be absorbed into automated agents or co-pilot systems.

Third, think like a system, not a campaign. “There are AI questions that are for Accenture,” Evans said. “There are AI questions that are for Bain, BCG, McKinsey… and then there are innovations about top-line innovation.”

Evans draws a neat parallel with AWS: “It feels like this might be the new AWS. This might be the new layer of abstraction.”

The prompt is the interface

To Evans, this is more than just a marketing problem. It’s a computing paradigm shift.

“Traditional software is deterministic,” he said. “It always does the same thing. Machine learning, which started working a decade ago, solved problems that were hard to explain to a computer, like whether this customer sounds angry or if this is a picture of a dog.”

And now? “LLMs do things that are easy to explain to an intern.”

The result is that millions of knowledge work tasks, marketing included, are being absorbed into a new, suggestive interface layer. And the interface is where control lives.

Back in 1995, said Evans, people thought you’d make money selling web browsers. “It wasn’t clear that the internet would win. It wasn’t clear where the value capture would be.”

Today, that uncertainty is back, but this time with stakes an order of magnitude higher.