AI agents are ready to shop. Some retailers are wary

Introduction

AI‑driven shopping agents—tools that can browse, compare, and even purchase products on a user’s behalf—are no longer a futuristic concept. Companies like Meta, Google, OpenAI, and Amazon have rolled out prototypes that promise to streamline the buying process. Yet, as early adopters discover, the reality is a patchwork of technical glitches, retailer resistance, and new ethical dilemmas that temper the hype.

Retailer resistance and technical roadblocks

The most visible sign of retailer wariness is Amazon’s aggressive stance. After a lawsuit against Perplexity for ignoring Amazon’s request to stop its agent from shopping on the site, Amazon also blocked Meta’s Muse agent. Similar restrictions appear across other sectors: JPMorgan Chase refuses external agents access to accounts, and airlines such as Delta and Amtrak have been reported to reject agent‑initiated bookings. Even when a retailer does not issue a blanket ban, generic bot‑blocking systems can intermittently stop agents, leading to an erratic experience. For instance, Muse could retrieve a product from Target one minute, only to be halted by a bot check the next, sometimes returning irrelevant alternatives or missing regional shipping constraints.

Inconsistent performance and user effort

When agents do manage to access a site, they often require extensive, precise instructions. Users must specify price ranges, shipping preferences, ethical criteria, and more, turning a simple search into a multi‑step dialogue. The agents may complete the task, but the results can be off‑target or incomplete—especially when a major marketplace like Amazon is unavailable. In practice, the time spent refining prompts and handling errors can outweigh the convenience the agent promises.

Ethical and competitive concerns

Beyond technical friction, agentic shopping raises questions about fairness and market impact. An experiment with Muse on The RealReal—a luxury resale platform—showed how an agent could repeatedly reserve items, effectively crowding out human shoppers. While the platform allows third‑party agents, it also implements safeguards to prevent automated abuse. Retailers worry that unchecked agents could manipulate inventory, scrape product data without consent, or exploit price‑monitoring features to undercut smaller competitors.

The path forward for AI shopping agents

For agents to become genuinely useful, several conditions must be met:

1. Retailer collaboration – Open APIs and clear usage policies would reduce ad‑hoc blocking and give agents reliable access to inventory, pricing, and shipping data.
2. Robust user controls – Users need transparent ways to pause, intervene, or override agent actions, ensuring they retain ultimate decision‑making power.
3. Ethical safeguards – Platforms should enforce limits on bulk‑adding or reservation‑hijacking to protect the shopping experience for all participants.
4. Focused use cases – Agents excel at narrow, data‑heavy tasks such as finding the lowest price for a known product or monitoring price drops, rather than broad discovery journeys.

Conclusion

AI shopping agents showcase the promise of automating tedious purchasing tasks, but their current rollout is hampered by retailer pushback, technical instability, and ethical gray areas. Until retailers and AI developers align on standards and safeguards, the technology will remain a niche tool that often requires more human oversight than it saves. For shoppers, the best approach today is to treat agents as assistants for specific, well‑defined queries while staying prepared to step in when the AI hits a wall.