Pinterest uses AI to identify what phase of a project or interest a user is currently in — for example, whether they are just starting to research home renovation ideas or are close to making a purchase decision. This understanding is then used to send more timely notifications and recommendations. The system, described in an October 2025 engineering post, resulted in an 88% higher email click rate and 32% higher push notification open rate compared to previous approaches.

The system works in two steps: first, extracting keywords from a user's recent searches, saves, and boards; then using a large language model to cluster those keywords into a coherent "journey" — a named theme like "planning a nursery" or "exploring Japanese cuisine." Pinterest initially used GPT (a proprietary model) to establish ground-truth journey labels, then trained a smaller open-source model called Qwen to run the same classification at scale with lower cost and latency. The journey labels are then matched to relevant Pin content and used to personalize email and push notification timing and content.