📰 Key Highlights

Pinterest announced on June 17, 2026, the launch of an experimental app called “Ask Pinterest,” featuring a conversational shopping experience that runs parallel to the main app to avoid disrupting existing users. At its core is Pinterest’s internally built “Taste Graph” — a graph system that maps user interests and aesthetic preferences — allowing users to ask questions in natural language and receive more personalized inspiration and product recommendations.

Unlike traditional Pinterest search, which works well for single keyword queries, Ask Pinterest is designed for multi-step or complex requests — like asking how to plan a dinner party or step-by-step room decor. The system can retain the user’s contextual thread across conversations. It can also read the user’s previously saved Pins and Boards for personalized responses. Initially, only limited access is available.

The advertising tools launched simultaneously include: an AI assistant for ad management, which is already live in the US (still in beta); Performance+ creative optimization model, available globally, which automatically selects the best ad creatives for each impression; and Pinterest MCP infrastructure layer, which allows advertisers to manage and monitor campaigns in a standardized way through third-party proxy tools.

This wave of updates debuted ahead of the Cannes Lions advertising tech conference, timing that also aligns with AI chatbots aggressively taking over traditional search engine territory. Google, ChatGPT, Meta, and Shopify have all rolled out conversational shopping features. Pinterest chose to prioritize training models on its own data rather than licensing to third-party AI services.

💬 JudyAI Lab Perspective

At a critical moment when AI chatbots are making a full assault on traditional search, Pinterest chose to launch “Ask Pinterest” — fighting with its own data rather than borrowing from external AI services. Both the timing and strategic direction make it clear: the next battle for image discovery platforms will be fought in natural language.

This case shows two design decisions worth paying attention to. First, Pinterest chose its own “Taste Graph” — a graph database of user tastes and preferences — as the core input for the conversational model, rather than just plugging into a third-party AI service. This shows the platform is increasingly treating its own unique data as a competitive foundation: anyone can connect a conversational interface, but whether you have your own training data is what creates long-term differentiation. Second, running the new app in parallel with the main app and limiting access initially is a classic “isolated testing” approach — letting AI features not interfere with existing user experience, collecting data while running, then gradually rolling it out. The same logic applies on the ad side: the MCP infrastructure layer lets advertisers connect in a standardized way rather than redesigning an entire set of tools.

If you’re planning an AI feature launch, start by asking: what proprietary data can third parties not replicate? That’s the input that’s most worth prioritizing for your model.

📅 Source Information

🔗 Further Reading