Meta launches shopping agent

On September eight, twenty twenty-six, Meta launched Muse, a personal AI agent designed to fill out forms, negotiate, book travel, and make purchases on users' behalf.

Amazon blocks the agent

Twelve days later, Amazon blocked Muse from shopping on its site, arguing that third-party agents must identify themselves and obtain retailer permission before transacting on behalf of customers.

The real shift

The dispute marks an important shift.For over three decades, Amazon determined how customers search, compare, and purchase.

AI moves from finding to representing

AI is moving from helping consumers find products to representing them in the market and increasingly making decisions and acting on their behalf. Companies are already responding to this shift.

Retailers' new problem

Online stores and marketplaces face a new strategic problem: they can own the transaction, supply, and customer engagement while losing control over the decisions that bring customers there.

Traffic grows but control shifts

Traffic from AI sources to U.S.retail sites grew three hundred ninety-three percent year over year in the first quarter of twenty twenty-six, according to Adobe.Visitors stayed thirty-two percent longer and viewed ten percent more pages.

Engagement masks lost influence

A retailer can attract highly engaged customers even as more decisions determining what they consider, compare, and choose are made elsewhere. The retailer risks losing influence over which products enter the customer's consideration set.

Discovery no longer guarantees control

Conversational search improves product discovery, but better discovery is not the same as greater control.A consumer's own AI agent can perform discovery across multiple marketplaces at once, making each marketplace just a source to query.

Competitors converge on same platforms

In October twenty twenty-five, Zillow became the first real estate app inside ChatGPT. Five months later, Realtor.com launched its own ChatGPT app. In travel, Booking.com and Expedia launched inside ChatGPT on the same day.

Network effects lose their power

AI can reduce how much of the network a competitor needs to recreate. An entrant can start with privileged access to one side of a market and use AI to organize the other without owning…

Mercor substitutes depth for breadth

Mercor approached freelancer matching without recreating Upwork's scale. It used AI-based assessments to extract data and create richer candidate profiles, then matched those profiles against employer requirements without building comparable network effects.

Agents know customers across categories

A marketplace understands how someone searches within its categories. A personal agent can know the person across categories and contexts. Muse can recall details across conversations, negotiate, and make purchases on consumers' behalf.

Tripadvisor loses feedback loop

Tripadvisor built more than a billion travel reviews, a formidable strategic asset.AI now weakens both advantages.Its interfaces synthesize review sentiment directly, reducing the need to visit Tripadvisor itself.Tripadvisor often loses visibility into downstream transactions.

Supply lock-in weakens on both sides

A traveler's AI agent can search across Airbnb, Booking.com, Vrbo, hotels, and direct-booking sites.A property-management AI agent can coordinate calendars, pricing, and payments across multiple platforms.This weakens lock-in on both demand and supply sides.

Walmart's own agent doesn't guarantee control

By early twenty twenty-six, engagement with Walmart's Sparky shopping assistant was rising, and customers who purchased through Sparky had baskets about thirty-five percent larger than other customers.

Walmart reclaims checkout control

In October twenty twenty-five, Walmart allowed customers to purchase products inside ChatGPT.In January twenty twenty-six, Walmart adopted Google's purchasing protocol for Gemini.

Marketplace data alone is not enough

In the platform era, data from user activity helped personalize the experience.But a personal agent connected across applications may see what happened beyond initial considerations.

High-frequency feedback loops matter most

Grocery gives Instacart an unusually strong position because purchases are frequent, repetitive, and completed within the platform. Over time, it learns what a household searches for, buys, substitutes, rejects, and prefers across brands and budgets.

Earning the right to represent

Even with strong engagement, network effects, supplier lock-in, proprietary data, and their own AI agents, marketplaces no longer necessarily control the customer relationship.

“In the agent era, owning the customer will increasingly mean earning the right to represent them.”