Stop renting, start building: GEO is a mirage

Everyone is talking about GEO. Agencies are selling it. Brands are buying it. On its face, the pitch makes sense: the search-based internet that powered the digital economy for decades is giving way to generative AI answers. The pressure to act is real. So is the trap: you’re going to start paying for another lease on someone else’s internet.

GEO is a tactic for a bigger and more important strategy. That strategy is owning your audience. The problem is that this new tactic is generating too much buzz and money for brands to ignore. McKinsey recently released a report arguing that $750B in US revenue will flow through AI-powered search by 2028. If you want your brand in on that revenue, you must invest in so-called generative engine optimization (GEO) or answer engine optimization (AEO). Both are designed to convert user prompts into clear references of a specific brand. And both are important.

Half of consumers already use AI-powered search engines. The shift is real. And optimizing for it isn’t wrong, it’s just not enough. The recommended response from industry leaders and agency directors is the same mistake the industry has been making for fifteen years: renting space from a landlord who can change the terms at any time.

GEO is not a new strategy. It’s a new lease on the same building. And we all know how badly that can go.

HubSpot is the cautionary tale nobody in this industry wants to sit with. They built one of the most sophisticated content operations in the business, dominated SEO for years, and then watched their search traffic fall off a cliff when Google changed how it surfaces results. They were renting. When the landlord renovated, they lost the apartment.

McKinsey’s own data shows that a brand’s owned content comprises only 5 to 10 percent of what AI search actually references. The rest comes from affiliates, user-generated content, publishers, and sources the brand has no control over. So the optimization play — clean up your content, sharpen your headings, structure your data for LLMs — is optimizing a minority stake in the outcome. Data like that kind of gives away the game. Your GEO strategy amounts to tinkering at the margins while the real action happens elsewhere.

The Infrastructure Is More Fragile Than the Pitch Admits

There’s another problem nobody selling GEO is talking about. The web is fighting back. Independent developers have built tools — with names like Nepenthes and Iocaine, after a carnivorous plant and a fictional poison — designed to trap AI crawlers in infinite loops of garbage data, wasting their resources and corrupting their training sets. One developer reported eliminating 94 percent of bot traffic to his site the day he deployed one. Cloudflare and others now sell bot-mitigation and anti-scraping tools to throttle AI crawlers at scale. The resistance has gone commercial.

The signal-to-noise ratio inside LLM training data is getting worse as publishers grow more hostile to being scraped without compensation. Any strategy built entirely on AI systems accurately surfacing your content is betting on a supply chain with an active sabotage problem. That is a risk the GEO consultants are not pricing in.

What Building Looks Like

Discoverability is not the end goal. It’s a byproduct of building something worth finding. There is a distinction between renting attention — optimizing your way into algorithms you don’t control — and building owned media infrastructure that creates direct relationships with audiences.

Red Bull Media House is the most cited example because it’s the most extreme: a beverage company that became a legitimate media operation. The audience relationship it built doesn’t reset every time an algorithm updates. That’s the point of the investment. The brands doing this right are creating content tailored to how people actually search on each platform — what they’re trying to accomplish, what format serves them in that moment — and their GEO and SEO improve as a consequence of being genuinely useful. Useful first, optimized second.

The Agentic Future Already Blinked

McKinsey’s study ends with a prediction that LLMs will eventually act as agents making purchase decisions on behalf of consumers. Most GEO pitches build toward the same vision. So it’s worth examining what happened when OpenAI actually tried to build it.

In September 2025, OpenAI launched Instant Checkout — buy products directly inside ChatGPT, Shopify and Etsy as launch partners. Users flooded in to research products. Almost none of them completed a purchase there. Out of Shopify’s millions of merchants, only a small fraction of eligible merchants went live with the integration. By early 2026, OpenAI pulled the checkout feature back, routing transactions to retailer apps instead.

The agentic shopping future collided with actual human behavior and lost. People used the AI to get smarter about what to buy. Then they went somewhere they trusted to buy it.

That is the whole argument. The brands that will survive AI-mediated discovery are the ones that have already built the kind of trust and direct relationship that makes someone choose them as the destination — not as the result of an optimization score, but because they earned it. No amount of schema markup builds that. No weekly GEO audit cycle builds that. Those are maintenance tasks. Infrastructure is something different.

Stop renting. Start building. The landlords are getting worse, and the leases are getting shorter.

View the full article