Retailers Must Balance Bot Blocking with AI Agent Access

Retailers have spent years improving bot detection, but the rise of AI-driven shopping agents means that some automated traffic is desirable. Blocking too aggressively risks shutting off emerging AI-agent-led customer journeys, while allowing too freely invites familiar fraud and scraping risks.

Who AI agents are The article notes three categories of automated visitors that retailers already want to serve:

• An agent scraping product data for a marketplace, such as Amazon gathering data for its “Buy for Me” catalog.
• An answer engine like ChatGPT or Google Gemini accessing product content to surface items in synthesized answers.
• A consumer-owned AI shopping agent (e.g., Instinct or Muse) acting as a digital delegate in product discovery or purchase.

Survey of bot-and-agent trust software users In Forrester’s Buyer’s Guide: Bot And Agent Trust Management Software, 2026, the research team interviewed 22 reference customers. Of those, 17 used bot-and-agent trust management software for account fraud, 14 for e-commerce fraud, and 14 for web and LLM scraping. Thirteen had started or were about to start applying this software to AI-agent trust.

What retail technology leaders should do now

• Stop treating every bot as the same problem. Define policies for malicious automation, known partners, answer engines, commerce agents, and customer-authorized shopping agents, and assign clear business owners for each category.
• Demand visibility and explainability from vendors. Require contextual information behind classifications and decisions, not just an allow/block outcome, and ensure teams can access sufficient signal to investigate incidents.
• Test for false positives in commerce-focused scenarios. Include legitimate AI-agent use cases in proofs of concept and evaluate detection strength, data quality, support responsiveness, and model-tuning capability.
• Bring commerce and security leaders to the same table. Commerce teams must decide which external agents to serve; security teams must determine how that access can be trusted and governed.
• Evaluate the vendor, not just its detection engine. Assess threat research, tuning ability, support, pricing model, and service options.

Our read: The data shows that a majority of surveyed retailers already use trust-management tools for traditional fraud vectors, yet only a minority have begun to extend those controls to AI-agent traffic. This gap suggests an emerging risk surface that vendors and retailers must address through granular policy frameworks and transparent, explainable tooling.