AI Search Drives Shopify's Revenue Up 36% – Complements, Not Replaces, Google
Shopify's Q2 2026 earnings reveal that AI search is a significant growth driver, with revenue up 36% to $3.6 billion (beating forecasts) and AI-driven traffic and orders tripling year-over-year. Crucially, AI is complementing—not replacing—traditional search, which still holds roughly a third of storefront sessions and is growing 1.3x over two years.
Why AI Search Works Better for Commerce
Shopify President Harley Finkelstein explained that unlike traditional search engines which rank by popularity against a few keywords, AI agents make multiple calls into Shopify's catalog, using richer structured data to match products to specific buyer intent. Example: "When a buyer asks an AI assistant for the best car seat that fits three across a sedan, traditional search focuses on the keyword 'car seat.' An agent understands the actual need, the dimensions, the vehicle type, and that they need three."
Key Metrics
- AI-referred sessions land directly on product description pages 2.5x more often than traditional search.
- 75% of AI-attributed purchases in Q2 happened outside the top 100 categories—Shopify's "sweet spot"
The company is investing in AI integrations, building connectors to Claude, ChatGPT, Perplexity, Manus, Replit, Vercel, and vibe-coding platforms like Lovable.
📖 Read the full source: HN LLM Tools
👀 See Also
VibecodedHub: A Discovery Platform Built Entirely with Claude Code
A developer built VibecodedHub, a discovery and launch platform for AI-built projects, using Claude Code as the primary engineer across multiple focused sessions. The platform includes landing page, authentication, feed, project submissions, Stripe integrations, admin panel, SEO, and OG images.
Postmortem: Governance System for AI Coding Projects with Claude
A developer shared a postmortem on a 2-week Claude Code project that produced 23k lines of code and 2,629 tests for approximately $100, emphasizing that the governance system was more important than the prompts. The framework is open source.
Multi-agent setup triggers $3,400 in charges due to hallucination loop
A developer hardcoded a corporate virtual card into environment variables for testing a multi-agent MCP setup, resulting in $3,400 in charges when a primary agent got stuck in a hallucination loop that spun up new paid proxy instances every 45 seconds for 14 hours.
OpenClaw as a developer: A two-week experiment with isolated AI coding agents
A developer built an autonomous 'AI developer' on OpenClaw that runs in Slack, uses Claude Code for coding, and operates with read-only configs. After two weeks, he questions if his no-self-learning approach is a dead end.