Amazon blocking AI agents causing traffic decline

Amazon Blocked the Agents and Lost 18% of Its Traffic

Amazon just proved the oldest law in business: if you block the agents, the market will simply go somewhere else.

They removed over 600 million product listings from AI-generated results. ChatGPT referral traffic to Amazon dropped 18% almost overnight. Not a rounding error. Not a blip. An 18% hole in the traffic that matters most in the next decade.

The Wall They Built

Amazon looked at the incoming wave of AI agents and did what big platforms always do: they threw up a wall. No more letting those scrapers and summarizers pull clean, structured data from their catalog. They blocked the agents, throttled the APIs, and basically told the future to piss off.

I get it. From their perspective, every AI agent that shops on behalf of a user is a potential disintermediator. Why let some sleek new commerce agent compare your price against Walmart, eBay, and a dozen smaller merchants when you can just starve it of oxygen?

So they chose the garden. They chose control. And for about five minutes, it probably felt smart.

What Actually Happened

The agents didn’t die. They just went around.

While Amazon was busy playing whack-a-mole with AI crawlers, the rest of the internet kept feeding clean product data to the models. Other retailers, marketplaces, and manufacturers didn’t lock down their catalogs. They made themselves legible to the new buyers.

Result? ChatGPT users started getting answers that said “here are five good options” and only one or two of them were on Amazon. The model didn’t lie. It just didn’t have Amazon’s data to work with. So it routed the purchase intent elsewhere.

This isn’t theory. This is happening in real time. The referral traffic drop tells you everything you need to know about where the next trillion in commerce is going to flow.

The Universal Commerce Protocol Won’t Care About Your Walled Garden

This is exactly why the work we’re doing at the Universal Commerce Protocol matters.

The future of commerce isn’t going to be owned by any single platform. It’s going to be a protocol layer that makes every product, every price, every seller, and every fulfillment option instantly legible to any AI agent, anywhere. No scraping. No blocking. No more stupid cat-and-mouse games between paranoid platforms and the agents their own customers are using.

Amazon thinks they’re protecting their moat. They’re actually digging their own.

Every time a major retailer makes their catalog harder for agents to read, they accelerate the shift toward open protocols. The agents are just going to route around them and buy from merchants who play nice with the new reality. That’s not ideology talking. That’s math.

I’ve watched this pattern before. I saw it with Google and the early mobile web. I saw it with Facebook and the open web. The big platforms always try to wall themselves off right before the next layer eats their lunch. They confuse control of the interface with control of the transaction.

The transaction layer is moving up the stack. The agents are becoming the new interface. And the interface that wins is the one that has access to the most complete, cleanest, most real-time commerce data on earth.

Amazon just volunteered to have less of it.

Will’s Take: You can either be the platform that agents prefer to buy from, or you can be the one they learn to work around. Amazon is teaching the entire market which choice is stupid. The winners in agentic commerce won’t be the companies with the biggest catalogs. They’ll be the ones who make their catalogs usable by the new buyers. The rest are just building really expensive museums for their own traffic data. The market is already routing around them, one lost referral at a time.

Related: Configure /.well-known/ucp Discovery Endpoint for AI

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *