Note: This article was originally published on Reasoned and is being cross-posted on MediaNama. Read the original version here: [link].

A bridge too far

I want you to look at Amazon’s blocking of Meta’s Muse Agent from a different lens:

Amazon is blocking its customers from using an agent to buy on Amazon.

A Muse agent on Amazon isn’t an “unauthorised agent”. It’s an agent authorised by a customer to shop on its behalf.

This stops:

• Scheduled repeat standard purchases (Amazon itself has a repeated purchase system)
• Price discovery and comparison
• Price watching, wherein a customer can ask an agent to buy at a particular price.

These are not the first agents that Amazon has blocked: It has repeatedly blocked Perplexity agents, to the point of suing perplexity (PDF) in November last year, saying:

Amazon then set up a technological barrier to restrict the Comet AI agent from covertly accessing private customer accounts. In response, Perplexity released a Comet software update specifically designed so that the Comet AI agent could evade that technological barrier.

And when Amazon again addressed Perplexity’s unauthorized conduct with Perplexity on two separate occasions, Perplexity refused to stop. Perplexity’s CEO understood that Perplexity was deliberately flouting Amazon’s rules, but had no legitimate justification for why Perplexity would not act honestly and transparently.

For publishers, it’s an arms race trying to keep out agents: Cloudflare blocks them, but browser use itself is evolving to beat blocking. However, unlike in the case of publishers, where AI platforms retrieve content and offer little or no value, in commerce they result in a monetary transaction.

In the traditional sense, a retailer should ideally want to increase footfalls, enable repeat purchases, bring in more customers and in a world where retailers do price matching in order to retain customers, price discovery, comparison and watching should be no issue. Still, it is turning away customers who use agents. Why?

Why Amazon is blocking bots

It isn’t just Amazon blocking bots: Shopify, which created the Universal Commerce Protocol along with Google, blocks “unauthorised” bots, as I found out while building Rebuy.

The advent and growing popularity of personal AI agents is going to change the Internet landscape because of the recent advent of easy-to-use, convenient and effortless personal agents like Muse, Instinct and Grok Bot are. That tribe is growing rapidly, and Assistant Benchmark currently lists 122 agents. Why are merchants wary of them?

First, authorisation can be contested and lead to more returns. Think back to when Alexa (ironically) placed automated orders for hundreds of households. Incorrect purchases or purchases by mistake can increase operating costs for merchants without the requisite returns. This is where authorisation differs from identification: even if you might identify a bot as user-initiated, how do you prove initiation and authentication for a transaction? That’s where perhaps payment gateway authentication comes in.

Second, agents can hallucinate while making purchases, and similar to whether there was authorisation, a question remains: how much was authorised? What if an agent buys 10 instead of 1 item? This is what Visa Intelligent Commerce tries to address through prior authentication rules, and in India, NPCI’s UAP will also seek to address, but we need to ringfence risk for both payment platforms and merchants, with pre-authorised wallets, so that the risk transfers to the users from the merchants.

Third, it is about increasing Average Revenue Per User (ARPU): Merchants want users on their platforms in order to show them related items, items they’ve browsed before, and show advertising to them. Advertising is now a significant part of Amazon’s revenue.

Fourth, it is about owning the customer: How many shopping apps do you have on your phone? As one founder told me once: he wants WhatsApp as his only interface. Personal apps like Must and Instinct threaten Amazon’s relationship with the customer. All businesses focus on increasing Lifetime Value (LTV) per customer by trying to become the default for something.

An Antitrust case in the making

Agents that do things for us threaten to become the default for everything for us. Every single application, service and platform that integrates with agents threatens to be relegated to the idea of becoming a tool-call for a job-to-be-done.

Amazon isn’t just a marketplace: it’s the foundation of the Amazon empire.

Blocking bots and refusing interoperability is Amazon’s way of preventing a bridge across its moat. The only bridge it will allow is its own: the only agent Amazon will want to authorise is its own. Expect a bot from Amazon eventually.

That will be the beginning of an antitrust case: if the only bot that Amazon allows is its own. Amazon will eventually be forced to become interoperable, just as its injunction against Perplexity was vacated because the customer, and not Perplexity, was using Comet to access Amazon. Then the battle will shift to how it enables interoperability. Watch this space.

Related reads: