The impact is already visible in healthcare, legal services, finance, software and education. The old funnel, a linear user-driven journey where discovery happened via search engines, followed by comparison across multiple sites and finally a visit to a brand to make a decision, has vanished.
Now a person types a question into an AI chatbot, such as “What treatment options exist for this condition?”, “What platform helps our team get started quickly?” or “What insurance fits my situation?”, and the AI replies.
That’s it. No website clicks. No website visits. No discovery for your brand. This shift is visible in the numbers. A report from management consultancy Bain shows more than half of searches now end with zero clicks. Forbes calls it “the 60% problem”; six in ten people never reach a website. In the industries where this is already prevalent, the top of the funnel is vanishing, and the more markets this hits — e-commerce, for example — the more the old map stops working.
It’s into this pressure cooker that generative engine optimisation (GEO) arrived. GEO is often described as the AI equivalent of search engine optimisation (SEO), a way for brands to measure their presence in AI generated answers. It’s presented as a calming idea in a world where page one of Google no longer matters, category pages lose influence and brand awareness offers little protection.
GEO is a false sense of control
The idea behind GEO sounds simple: test how your organisation appears in a chatbot such as ChatGPT, track visibility and then improve from there. But that can’t work, because the system itself doesn’t work that way as AI outputs are personal. They shift from one phrasing to another. They change with each model update. They are shaped by retrieval you cannot see, and they vary every time you ask the same question. It’s like claiming you know a whole book from reading just a single page.
One of the first GEO companies, Lorelight, shut down. Not because the tool was broken — it tracked mentions well. Customers left because insights didn’t help them change anything. They had reports, not levers; you can’t build a strategy around something you can’t reliably measure.
Kevin Indig, a search and growth analyst behind the Growth Memo newsletter, captured this problem in one sentence: “You cannot sell certainty, only controlled learning.” GEO tries to sell certainty in a system designed to shift every time you touch it. VC firm MMC’s report on AI discoverability points to the same problem: tracking is easy, but insight without the ability to act is not a strategy.
And the brands that show up most aren’t doing AI tricks.They just have strong fundamentals, like authoritative presence, trusted content and real world credibility — similar to the things that matter in traditional SEO. AI rewards substance, not shortcuts. And GEO could only observe this, not influence it.
So what do you do?
Agentic entities can save your discoverability — GEO can’t
When funnels collapse and clicks disappear, one thing still matters: information that AI can understand and act on.
In practice, this means treating every important piece of content as an entity — something AI can recognise, interpret and compare. A product page, a sizing guide, a treatment overview, an onboarding process, a warranty policy - anything a user might ask about.
And because people no longer search with keywords but with natural questions, comparisons and tasks, these entities must adapt to that behaviour. This is the new SEO: not ranking pages, but preparing your information for how AI interprets intent. If that information isn’t expressed in structured, factual, machine-ready form, it won’t appear in the answer.
The world needs to shift toward agentic entities: content that updates itself, adapts to context and stays accurate even as models evolve. They give AI tools what they need to explain you, compare you and recommend you.
Where GEO only watches what AI outputs, agentic entities shape what the AI decides, offer the foundation for creating information AI can trust, interpret and act on and give organisations the leverage they need to stay visible in a world where answers come first and links come last.
The old web was built for humans. The next one must be built for machines, and we’re already underway.
