Excited to share our latest paper on Pinterest Generative Engine Optimization, a fundamental shift in how we think about search and content visibility with the power of AI Agent.
We’re tackling a problem that’s quickly becoming mission-critical for platforms (e-commerce, social media, SaaS, and enterprise services): How to get your business to directly show up in ChatGPT, Gemini, Claude, and other LLM-powered engines?
We introduce a production-scale system:
🌐 Vision-Language Model (VLM) fine-tuning
We adapt VLMs to generate intent-aligned topics that predicts user search behavior ahead of time and drives better indexing in AI systems.
🤖 Agent-based trend mining system
Instead of waiting for user search, our autonomous agents proactively mine real-time internet trends to surface emerging demand and create content that stays ahead of user interest curves.
🔗 Two-Tower ML + Link Equity for authority signals
To empower billion-scale deployment, we introduce a two-tower MLP model combined with a principled link equity framework that builds authority-signaling structures across large content collections with ANN and multi-modal embeddings.
The result? This pioneer approach that goes beyond metadata tricks used by most startups in the space today, enabling true visibility in generative search environments with 20% search session boost.
Happy to chat with anyone working at the intersection of search, content, and agentic discovery!
date: 2026-02-09 23:25:00
Keep reading