The next major online-shopping battle may not be between two retailers.
It may be between a retailer and an artificial-intelligence agent acting on behalf of the customer.
Amazon has failed to obtain an immediate US court order preventing Perplexity from operating an AI-powered shopping tool on its platform.
The decision allows Perplexity’s agentic shopping technology to continue operating while the broader legal dispute proceeds. Amazon argues that the system accesses its marketplace in ways that violate its rules and potentially threaten security, customer trust and the shopping experience. Perplexity presents the tool as a consumer assistant capable of researching products and completing purchases for users. �
Reuters
The case is significant because it could influence whether powerful online platforms may decide which external AI agents are permitted to interact with their services.
What Does Perplexity’s Shopping Agent Do?
A traditional shopping assistant may recommend products and provide links.
An agentic shopping system can go further.
Depending on the tools and permissions available, it may:
Search for products
Compare specifications
Read customer reviews
Evaluate prices
Add products to a basket
Enter delivery information
Navigate checkout
Complete a purchase with the user’s approval
The consumer gives the system a goal rather than manually visiting every product page.
For example:
“Find a reliable 32-inch smart television under £200, compare independent reviews and buy the best option after showing me the final choice.”
The AI may complete much of the journey on the customer’s behalf.
That convenience threatens one of the central sources of power in e-commerce: control over how shoppers move through the marketplace.
Why Amazon Objects
Amazon has spent decades designing its interface, recommendation systems, advertising placements and checkout process.
An external AI agent may bypass parts of that experience.
Instead of a human shopper viewing sponsored listings, product banners and Amazon’s own recommendations, the agent may collect information directly and present only its selected options.
That could affect:
Advertising revenue
Product visibility
Customer-data collection
Recommendation systems
Fraud prevention
Seller competition
Brand control
Platform security
Amazon also argues that automated access can create technical and security risks when third-party systems interact with accounts or checkout processes without an approved partnership. �
Reuters
The dispute is therefore not simply about one shopping feature.
It is about whether Amazon owns only the marketplace—or also controls the digital path every customer must take through it.
Why the Judge Refused the Immediate Ban
The court did not necessarily declare Perplexity the final legal winner.
It declined to grant Amazon the immediate emergency restriction it requested.
That means the underlying questions concerning access, platform terms, automation and potential consumer harm remain unresolved.
The ruling is still important because preliminary bans can determine whether a young technology survives long enough to reach a full trial.
Had the court blocked the agent immediately, the decision could have discouraged other companies from building tools that navigate dominant digital platforms.
Why the Story Went Viral
The case makes the idea of an AI agent feel practical rather than futuristic.
People have spent years asking when AI will stop merely talking and begin doing things.
Shopping is one of the clearest examples.
The system does not only explain which television is best.
It attempts to purchase the television.
That changes the commercial relationship.
Amazon traditionally competes for the human shopper’s attention.
In an agentic economy, it may need to compete for the AI agent’s recommendation.
The Battle Over the “Customer Interface”
Technology companies understand that the organisation controlling the interface often controls the customer relationship.
Google controlled search.
Amazon controlled online-shopping discovery.
Apple and Google controlled mobile applications.
Social platforms controlled content distribution.
AI assistants may become a new layer sitting above all of them.
A user may no longer open five shopping websites.
They may ask one assistant to decide where to shop.
If that future arrives, marketplaces risk becoming invisible suppliers behind an AI-controlled interface.
The retailer stores the products.
The AI owns the conversation.
Consumer Benefits
Agentic shopping could help people:
Compare many products more quickly
Avoid manipulative interface design
Find lower prices across retailers
Understand complex specifications
Identify hidden subscription costs
Make accessible purchases through conversation
Avoid impulse buying
Apply personal ethical preferences
An agent might be instructed to exclude products with poor repairability, suspicious reviews or expensive replacement parts.
Used carefully, this could shift some power back towards consumers.
Consumer Risks
The agent may also:
Misunderstand the user’s needs
Trust fake reviews
Select sponsored products
Make unauthorised purchases
Expose payment information
Buy from unreliable sellers
Conceal commercial relationships
Become manipulated by marketplace content
A helpful agent can become a financial risk the moment its authority exceeds its reliability.
Human approval should therefore remain essential before payment, subscription or delivery commitments are made.
Mary Chuks’ Perspective
This case is the beginning of the AI gatekeeper war.
Amazon wants to protect the environment it built.
Perplexity wants AI to act as the customer’s independent representative.
Both positions contain legitimate concerns.
Platforms need security and predictable access.
Consumers need tools that work for them rather than merely directing them towards whatever generates the most advertising revenue.
The correct answer is not unrestricted agent access or complete platform control.
It is a trusted-agent standard containing:
Verified identity
Clear user authority
Restricted permissions
Transparent commercial relationships
Confirm-before-purchase controls
Complete transaction logs
Easy cancellation
Strong security requirements
An AI agent should enter the shop with the customer’s permission—not behave like a digital burglar carrying their credit card.
Practical Takeaways for Shoppers
Before allowing an AI agent to purchase anything:
Require approval before final payment.
Set a maximum spending limit.
Review the exact seller, delivery date and returns policy.
Use a payment method with buyer protection.
Check whether the agent receives affiliate commission.
Avoid giving permanent access to unnecessary financial information.
Keep transaction records and screenshots.
Start with low-risk purchases before allowing wider autonomy.
Conclusion
Amazon’s failed request does not settle the future of AI shopping.
It opens the door to the real argument.
When a consumer sends an AI agent into a marketplace, whose rules matter most—the platform’s, the agent developer’s or the customer’s?
The answer may determine who controls online commerce in the agentic era.
Original Source and Further Reading
Primary reporting: Reuters, “Amazon Loses US Court Bid to Ban Perplexity’s AI Shopping Tools,” published 4 August 2026. �