Google Tests Five AI Shopping Agents but None Complete Payments Autonomously
Google’s agentic shopping protocol enables AI agents to work across discovery and post-purchase support, but tests on ChatGPT, Gemini, Claude, Alexa and Sparky reveal none can complete payments without user input. Foundational AI tools like Gemini lack direct catalogue access, while retail-native agents are precise but limited to their assortments.
1. Google’s Agentic Shopping Protocol
Google’s protocol allows AI agents to handle discovery and post-purchase support across multiple shopping stages on behalf of customers. It integrates features like product identification, comparison, and customer assistance but currently stops short of completing purchase transactions without manual input.
2. Test of Five Agentic Shopping Tools and Findings
A comparative test of ChatGPT, Gemini, Claude, Alexa and Sparky showed none could execute an end-to-end purchase autonomously, with all tools requiring human steps for payment and fulfillment. Foundational models like Gemini excel in cross-platform recommendations but lack real-time catalogue data, while retail-native assistants offer precise inventory info yet remain limited to their own assortments.
3. Challenges and Path to Autonomous E-Commerce
The main obstacle to full agentic shopping is embedding payment and fulfillment capabilities directly into AI workflows, bridging the gap between discovery and transaction. Partnerships between foundational AI providers and retailers are emerging to integrate structured catalogue access, real-time pricing and delivery options for a seamless end-to-end experience.




