This topic covers AI-driven shopping and product-discovery assistants — from chat-based research helpers like a ChatGPT Shopping Research Assistant to visual or taste‑graph systems that map preferences to products. Those assistants rely on two technical pillars: Web Data Extraction (search, scraping, structured page extraction) and Chat API Integrations (conversational LLMs orchestrating searches, summarization, and actions). Recent tooling has standardized access patterns via Model Context Protocol (MCP) servers and browser automation layers so LLMs can perform live research without hardcoded API keys or brittle screen scraping. Key components include Web Search MCP (full web search, summaries, page extraction), Perplexity’s Sonar integration (real‑time research and reasoning), and Exa (AI‑focused search and code search). For interaction with rendered pages and complex flows, Playwright, Skyvern, and Browserbase offer MCP‑driven browser control and structured accessibility data, while Firecrawl provides scalable scraping and crawling. These tools enable assistants to fetch live prices, stock status, user reviews, and product specs and to perform form filling or checkout automation when permitted. As of late 2025, relevance comes from consumer demand for faster, more personalized product discovery and from commerce use cases that need reliable, auditable web data. Important evaluation criteria are data freshness, traceability of sources, privacy/compliance with site terms, image and taste‑graph support for visual recommendations, and vendor‑agnostic MCP integrations that let teams mix search, scraping, and automation capabilities. This topic helps readers compare approaches that balance timely web access, conversational UX, and responsible data collection for practical shopping assistants.