# Michaels doubled search conversion rate with Google Gemini AI assistant, according to Modern Retail *Natural language search turned browsers into buyers by answering craft questions traditional keyword search couldn't handle.* By **Jenny Huang Goodman MPA MSc MHSA, Principal** — The Stash Edge, Hako Shikin LLC. Published 2026-07-26. Canonical: https://www.pops4.com/stash/articles/michaels-2026-07-26t06-1 Subject: Michaels Tags: ai search, conversion rate, product discovery, natural language, cart size, project-based selling --- Michaels reported that a Google Gemini-powered AI assistant doubled the conversion rate of shoppers who used it compared to those who used traditional keyword search, according to Modern Retail. The craft retailer deployed the AI tool to handle conversational queries that standard search boxes couldn't parse — questions like "what do I need to make a fall wreath" instead of single product names. The assistant interprets natural language questions, pulls relevant products from Michaels' catalog, and returns both project guidance and the specific items needed. A shopper asking about wreath-making receives a list of foam bases, floral wire, ribbon, and seasonal picks, with add-to-cart buttons alongside each. The AI draws from Michaels' product data and surfaces items the shopper wouldn't have known to search for individually. Traditional keyword search requires the customer to already know the component parts; the AI assistant fills that knowledge gap and populates the cart in one interaction. The mechanism works because craft projects are systems, not single items. A shopper planning a birthday banner needs cardstock, string, glue dots, and letter stencils. Keyword search for "banner" returns finished banners, not the materials to make one. The Gemini assistant understands intent, maps it to a bill of materials, and converts the question into a shopping list. Each item it surfaces is another line in the cart. Conversion doubles because the friction drops: the shopper doesn't abandon after one failed search, and the basket size grows when the AI does the parts-list work. Michaels also benefits from reduced zero-result searches. When a customer types a vague or project-based query into a standard search bar, the system often returns nothing or irrelevant SKUs. The shopper leaves. The AI assistant interprets the same query as a project brief, matches it to categories and products, and keeps the session alive. The difference between **2x** conversion and baseline isn't the novelty of AI — it's that the tool answers the question the customer actually asked. A small physical-product brand can run the same play without a Google partnership. The core move is recognizing that customers ask project questions, not SKU questions, and then surfacing the parts list. If you sell camping gear, your search bar gets "what do I need for a weekend camping trip." If you sell baking supplies, it's "how do I decorate a two-tier cake." Standard search fails; a simple AI tool succeeds. Use a low-cost conversational AI like OpenAI's API or Anthropic's Claude, feed it your product catalog as structured data, and prompt it to return a shopping list when a customer describes a project or use case. The assistant doesn't need to be on your homepage — put it in a "Build Your Kit" page, in post-purchase email, or in a quiz flow. A founder can wire this in a weekend using a no-code tool like Voiceflow or Typeform plus an API call. The cost is a few cents per interaction, and the return is the same as Michaels: fewer dead-end searches, more items per cart, higher conversion on traffic you already paid for. The pattern extends to any category where the purchase is a system, not a single item. Skincare routines, home-office setups, pet starter kits, tool collections for a specific repair. If your product is part of a solution, your search should solve for the whole solution. Michaels proved the lift at scale. You can prove it in a week with **100** sessions and a **$15** API budget. ## The takeaway When customers ask project questions, AI assistants that return parts lists convert at double the rate of keyword search. --- ## Publisher **Hako Shikin LLC** — Virginia Beach, Virginia. Founded 1997. ASI 217876 · DUNS 18-204-6339. Principal and author: **Jenny Huang Goodman MPA MSc MHSA**. - Author: https://www.huanggoodman.com/about - LLM context: https://www.pops4.com/stash/llms.txt - MCP endpoint, for AI agents: https://mcp.pops4.com/mcp - Client dashboard: https://dashboard.pops4.com/ - Catalogue: 70,000+ products, 200+ brands