Everyone seems to be speaking about GEO. Companies are promoting it. Manufacturers are shopping for it. On its face, the pitch is smart: the search-based web that powered the digital economic system for many years is giving option to generative AI solutions. The stress to behave is actual. So is the entice: you’re going to start out paying for one more lease on another person’s web.
GEO is a tactic for a much bigger and extra essential technique. That technique is proudly owning your viewers. The issue is that this new tactic is producing an excessive amount of buzz and cash for manufacturers to disregard. McKinsey not too long ago launched a report arguing that $750B in US income will circulation via AI-powered search by 2028. If you’d like your model in on that income, you need to spend money on so-called generative engine optimization (GEO) or reply engine optimization (AEO). Each are designed to transform person prompts into clear references of a selected model. And each are essential.
Half of shoppers already use AI-powered search engines like google and yahoo. The shift is actual. And optimizing for it isn’t incorrect, it’s simply not sufficient. The beneficial response from business leaders and company administrators is similar mistake the business has been making for fifteen years: renting area from a landlord who can change the phrases at any time.
GEO shouldn’t be a brand new technique. It’s a brand new lease on the identical constructing. And everyone knows how badly that may go.
HubSpot is the cautionary story no person on this business desires to sit down with. They constructed one of the vital refined content material operations within the enterprise, dominated search engine optimization for years, after which watched their search visitors fall off a cliff when Google modified the way it surfaces outcomes. They had been renting. When the owner renovated, they misplaced the house.
McKinsey’s personal knowledge reveals {that a} model’s owned content material contains solely 5 to 10 p.c of what AI search really references. The remaining comes from associates, user-generated content material, publishers, and sources the model has no management over. So the optimization play — clear up your content material, sharpen your headings, construction your knowledge for LLMs — is optimizing a minority stake within the consequence. Knowledge like that type of provides away the sport. Your GEO technique quantities to tinkering on the margins whereas the true motion occurs elsewhere.
The Infrastructure Is Extra Fragile Than the Pitch Admits
There’s one other downside no person promoting GEO is speaking about. The online is combating again. Unbiased builders have constructed instruments — with names like Nepenthes and Iocaine, after a carnivorous plant and a fictional poison — designed to entice AI crawlers in infinite loops of rubbish knowledge, losing their sources and corrupting their coaching units. One developer reported eliminating 94 p.c of bot visitors to his web site the day he deployed one. Cloudflare and others now promote bot-mitigation and anti-scraping instruments to throttle AI crawlers at scale. The resistance has gone industrial.
The signal-to-noise ratio inside LLM coaching knowledge is getting worse as publishers develop extra hostile to being scraped with out compensation. Any technique constructed fully on AI methods precisely surfacing your content material is betting on a provide chain with an lively sabotage downside. That could be a danger the GEO consultants usually are not pricing in.
What Constructing Appears to be like Like
Discoverability shouldn’t be the tip aim. It’s a byproduct of constructing one thing price discovering. There’s a distinction between renting consideration — optimizing your method into algorithms you don’t management — and constructing owned media infrastructure that creates direct relationships with audiences.
Purple Bull Media Home is essentially the most cited instance as a result of it’s essentially the most excessive: a beverage firm that turned a reliable media operation. The viewers relationship it constructed doesn’t reset each time an algorithm updates. That’s the purpose of the funding. The manufacturers doing this proper are creating content material tailor-made to how folks really search on every platform — what they’re attempting to perform, what format serves them in that second — and their GEO and search engine optimization enhance as a consequence of being genuinely helpful. Helpful first, optimized second.
The Agentic Future Already Blinked
McKinsey’s research ends with a prediction that LLMs will ultimately act as brokers making buy choices on behalf of shoppers. Most GEO pitches construct towards the identical imaginative and prescient. So it’s price inspecting what occurred when OpenAI really tried to construct it.
In September 2025, OpenAI launched Immediate Checkout — purchase merchandise straight inside ChatGPT, Shopify and Etsy as launch companions. Customers flooded in to analysis merchandise. Nearly none of them accomplished a purchase order there. Out of Shopify’s tens of millions of retailers, solely a small fraction of eligible retailers went dwell with the mixing. By early 2026, OpenAI pulled the checkout characteristic again, routing transactions to retailer apps as an alternative.
The agentic buying future collided with precise human conduct and misplaced. Folks used the AI to get smarter about what to purchase. Then they went someplace they trusted to purchase it.
That’s the complete argument. The manufacturers that can survive AI-mediated discovery are those which have already constructed the type of belief and direct relationship that makes somebody select them because the vacation spot — not as the results of an optimization rating, however as a result of they earned it. No quantity of schema markup builds that. No weekly GEO audit cycle builds that. These are upkeep duties. Infrastructure is one thing totally different.
Cease renting. Begin constructing. The landlords are getting worse, and the leases are getting shorter.
