By skyforbes 
In 2012, Coca-Cola had 35 million Facebook fans. They spent millions building that audience. Today, their posts reach roughly zero of them without paying Facebook. Now imagine the same playbook, but with AI agents controlling your purchase decisions instead of your attention.
It’s already happening.
Here’s why: Facebook spent a decade teaching us to rent land in their garden. AI agents are going one step further. They’re not just controlling what you see, they’re deciding what you buy, from whom, and why. And unlike Facebook’s algorithm, you can’t even see the manipulation happening.
A user, tired after a long week, types a simple request into an AI-driven chat window.
User: I am flying to Tokyo next week. Please find the best noise-canceling headphones and deliver them to me by Friday.
Chatbot: I’ve ordered the Sony WH-1000XM5 headphones. They have the highest review scores on CNET and TechRadar and are available for $328 with one-day shipping. They will be at your door tomorrow. The total has been charged to your card.
This is a seamless, conversational purchase of the future. It’s also the final, perfectly unbreachable “walled garden“ that regulators have spent 30 years trying to tear down.
We are witnessing the birth of a new, and perhaps ultimate, gatekeeper. And it’s being built using the same playbook that turned Facebook from a dorm-room free service into a $1 trillion advertising empire.
The rented land playbook
To understand what is happening with the AI-driven shopping experience, we can look back at how Facebook defined the last decade of web interactions.
In the late 2000s, Facebook offered brands a utopian promise: a free, direct relationship with their customers. The lure was the Like count. Brands spent millions on campaigns to attract fans to their newly created Facebook brand pages, believing that each new Like was a free subscription to the brand’s news feed.
In 2012, this was absolutely true. A brand’s post could organically reach an average of 16% of its followers.
Then, on May 18th, 2012, Facebook went public.
Almost immediately, its priority shifted from growth to profit. The company began systematically dialing down brands’ organic reach. By 2014, the average organic reach had plummeted from 16% to 6%.
Today, it is 0%.
The AI playbook: From utopia to “Temu-fication”
Enter ChatGPT. The playbook is identical, but the “land” is the chat window itself.
Act 1 — The free land utopia
Facebook’s early years offered free communication and access to the social graph. OpenAI mirrored this by offering ChatGPT at low or no cost, gathering hundreds of millions of users, and inviting brands and developers through the GPT Store. The message was the same: Be where your users are.
Act 2 — The monetization
Once Facebook’s growth peaked, it started charging brands to reach the audiences they’d already cultivated. OpenAI will inevitably follow. To satisfy investors, it must monetize its chat ecosystem.
Today, an AI agent might select the Sony headphones because they perform best. Tomorrow, that answer could shift to the highest bidder. The AI’s response, trusted, concise, and persuasive, would become the most valuable ad slot in the history of the internet.
Act 3 — The walled garden
Facebook evolved into a pay‑to‑play system. OpenAI is building a similar structure around chat‑based shopping and instant checkout.
To attract major players like Walmart and Etsy, OpenAI must offer them brand visibility. When you buy, you see a “Buy from Walmart” button. But this transparency is misleading. The actual issue is not who you’re buying from, but why the AI selected them.
This is why the Instant Checkout service (already powered by major players like Stripe and PayPal) is a real trap. In my example, the user bought the Sony headphones “from Walmart”, but they never visited the Walmart e-store. They did not create an account. Walmart made a sale but got zero customer data. They can’t implement their CRM strategy, and as a result, they can’t build a future relationship with their new customer.
The Temu Endgame
The next, and final, step is to cut out the brand and the retailer entirely.
This model is not hypothetical. It was perfected in China.
The consumer-to-manufacturer (C2M) model used by e-stores like Pinduoduo and its counterpart, Temu, is built precisely on this logic.
Eliminate the brand.
The Chinese e-commerce idea aggregates millions of individual consumers’ intentions, eg. I want a red spatula and send this demand directly to the factory floor. The factory produces the exact number of desired items, with no marketing, branding, or stock risk, shipping it straight to consumers.
The OpenAI model is the perfect tool for this. It can aggregate a million user requests for a cheap pair of headphones and send a bulk order straight to a manufacturer, bypassing Sony, Bose and Walmart entirely.
This is the price-driven, no-brand future of e-commerce. It’s a world where the AI doesn’t just help you choose a brand; it makes the brand itself irrelevant.
Two immediate victims
The first is the small or independent business. A user searching for a “handmade leather wallet with blue stitching” could find a specific Etsy seller. An AI agent is programmed to find the most authoritative, “safest” answer, which will almost always be an established incumbent like Coach or Amazon Basics. In the “pay-to-play” war for a “promoted selection” slot, small businesses will be rendered invisible, wiping out the diversity of the long-tail economy.
The second victim is the open web itself. The AI model is user-friendly. AI agents scrape the open web for information — from publisher articles, to recipe blogs, to product reviews, and synthesize the value, and present it to the user so they never have to click a link.
This is the “zero-click” web. And the evidence is already here.
Sites like Wikipedia are reporting alarming drops in visits and page views, noting that search engines are now providing answers directly to searchers, often based on Wikipedia content. Major analytics firms, such as Gartner, have predicted that traditional search engine volume will drop by 25% by 2026, primarily due to the rise of AI chatbots.
When a user gets their answer from the AI, the original creator, the publisher, the retailer, and the blogger get nothing. They are hollowed out, their value extracted, with no referral traffic or ad revenue sent in return.
How to regulate a thought?
This brings us to the central challenge for governments. For 30 years, antitrust regulators have been fighting visible monopolies.
In U.S. v. Microsoft, the harm was clear: Internet Explorer was an icon on the desktop, a bundle designed to kill Netscape.
In EU v. Google (Shopping), the harm was clear: a Google Shopping box was visibly at the top of the search results page.
But how do you regulate a black box choice? It’s an invisible, algorithmic act of value extraction, disguised as an objective, helpful answer.
If a ChatGPT agent always concludes that Google Flights has the “best” option, how can a user, or a regulator, prove it wasn’t simply an objective choice?
The AI’s mind is the perfect place to hide anti-competitive intent.
The Question We Should Be Asking
We spent a decade debating whether Facebook was too powerful. By the time we reached a consensus, it was too late to make a difference.
We’re about to make the same mistake with AI agents, except this time, the stakes are higher. We’re not just talking about what we see in our feeds. We’re discussing what we buy, where our money goes, and which businesses thrive.
The question isn’t whether this will happen.
The question is whether we’ll notice before it’s too late — and whether we’ll demand something different while we still can.