On June 17, ahead of the Cannes Creativity Festival, Pinterest announced the launch of an experimental AI shopping application, “Ask Pinterest,” which uses Pinterest’s “Taste Graph” (an internal data graph that links users to their interests and aesthetic preferences in real time) to provide a conversational shopping experience that can preserve context across sessions.
Ask Pinterest’s Taste Graph and cross-session conversational shopping
At the core of Ask Pinterest is using Pinterest’s “Taste Graph” to deliver context-aware product recommendations. Users can ask multi-step, complex queries, such as “help me plan a dinner party” or “I want to slowly decorate this room over the next three months.” The app will provide suggestions based on the user’s past saved Pins and Boards.
The most important differentiating feature is cross-session context retention: Ask Pinterest will remember the Scandinavian-style sofa searched for last week and the dining table setting saved two days ago, and naturally incorporate them into subsequent recommendations. Users don’t need to re-describe their needs from scratch each time. Pinterest CBO Lee Brown said in an official announcement: “In the future, product discovery won’t be driven only by keywords—it will be shaped by context, taste, and trustworthy recommendations.”
AI shopping strategies from Google, Amazon, Meta vs. Pinterest’s differentiated route
The launch of Ask Pinterest comes at a key moment of competition between AI chatbots and traditional search engines for consumer attention: Google has rolled out AI shopping features (covering product discovery, price tracking, and AI agent checkout); ChatGPT launched an agentic shopping system in 2025; Meta and Shopify are also actively laying out their strategies in AI commerce.
Pinterest’s approach is markedly different from the competitors above: it chooses to train its own AI models using its data and power its own AI products, rather than providing product recommendation data to third-party AI services via licensing agreements.
Three ad-owner AI tools released in parallel: Performance+ creative goes global, Pinterest MCP infrastructure
Pinterest is simultaneously rolling out three tools for advertisers:
U.S. Ads Manager AI Assistant (Beta): helps advertisers optimize ad delivery; currently in the testing phase.
Performance+ creative: a new AI model, already live globally; it can automatically select, from multiple ad creatives at each ad impression, the one predicted to perform best for delivery.
Pinterest Model Context Protocol (MCP) infrastructure layer: enables advertisers to manage and monitor ad campaigns in standardized ways through third-party agent tools. It is similar to the concept of an MCP open standard proposed by Anthropic, but designed specifically for the ad ecosystem.
FAQ
Will Ask Pinterest replace the existing main Pinterest app?
According to Pinterest’s official explanation, Ask Pinterest is a standalone experimental app, and in the short term it will not disrupt the user experience of the main app. Pinterest positions it as a testbed to trial AI shopping features. Pinterest says that in the future, the results from Ask Pinterest will be gradually integrated into the main app, becoming a foundation for AI features in its flagship product.
What is Taste Graph, and how does it integrate with Ask Pinterest?
According to the article, Taste Graph is Pinterest’s internal data graph that links users to their interests and aesthetic preferences in real time. Ask Pinterest uses Taste Graph to analyze users’ previously saved Pins and Boards, and continuously retains these preferences during cross-session conversations to provide context-aware product recommendations, rather than static results based on a single keyword query.
How is Pinterest’s MCP infrastructure layer related to Anthropic’s MCP standard?
According to the article, Pinterest MCP is a standardized ad management agreement that Pinterest designed for the ad ecosystem—“similar to the concept of the MCP open standard proposed by Anthropic”—but tailored specifically for managing and monitoring ad campaigns by advertisers, and not a direct extension or licensed implementation of Anthropic MCP.
