# Michaels doubles search conversion with Google Gemini AI assistant, early tests show *Guided product discovery beats keyword guessing for arts and crafts buyers who don't know what they need yet.* By **Jenny Huang Goodman MPA MSc MHSA, Principal** — The Stash Edge, Hako Shikin LLC. Published 2026-07-27. Canonical: https://www.pops4.com/stash/articles/michaels-2026-07-27t15-2 Subject: Michaels Tags: conversion, ai, search, guided selling, product discovery, michaels --- Michaels reported that a Google Gemini-powered AI assistant doubled the conversion rate of traditional search on its website, according to early deployment results shared with Digiday. The craft retailer integrated the conversational tool to help shoppers navigate a catalog with thousands of SKUs across overlapping categories — beads, paints, adhesives, fabrics — where shoppers often search for a project outcome rather than a specific product name. The assistant works by asking clarifying questions before showing results. A customer typing "birthday party decorations" might be asked the age of the child, the theme, and whether they want ready-made or DIY. That guided flow replaces the traditional search bar's single keyword input and page of ranked results. According to Michaels, the conversion lift came from shoppers who previously abandoned after seeing too many options or the wrong category of product. The mechanism is scoped discovery. Traditional search assumes the buyer knows what they want and can describe it in a few words. AI-assisted search assumes the opposite — that the buyer has a goal but needs help translating it into the right product or combination of products. The assistant narrows the field through dialogue, filtering not just by keyword match but by use case, skill level, and budget. The system can also suggest complementary items in a single response, which increases basket size without requiring the shopper to run multiple searches. This works especially well for physical products with high SKU density and overlapping use cases. A shopper looking for "fabric paint" might need textile medium, brushes, stencils, and sealant, but won't search for all five. The assistant surfaces the bundle. A buyer searching "wood glue" might be building furniture or repairing a toy, and the best adhesive differs. The assistant asks. The result is fewer dead-end searches and more carts that contain what the project actually requires. A small physical-product brand can steal this play without Google's backend. Use a simple quiz on your product page or a conversational form on your homepage. Ask three to five questions that segment by use case, experience level, or project type. Route each answer path to a specific product or pre-built bundle. If you sell coffee, ask: hot or iced, strong or smooth, morning or afternoon. If you sell skincare, ask: oily or dry, sensitive or not, morning or night routine. Each path shows one product or a three-item kit with a single add-to-cart button. No search bar needed. The cost is a **$30-per-month** Typeform or Fillout account and two hours to script the decision tree. Write the questions in plain language. Map five to seven end states — the most common buyer profiles or project types you see in support emails or returns. Link each end state to a product page with the answer pre-filled or a Shopify cart permalink with the bundle already loaded. Embed the quiz above the fold on your homepage or as the first step in your email welcome series. Track conversion from quiz completion to checkout in Google Analytics as a goal. For larger operations with engineering resource, integrate a lightweight conversational API like OpenAI's Assistant or Anthropic's Claude using function calling to pull live inventory and pricing. You can build a working prototype in a week for under **$500** in developer time plus API costs that run **$0.02 to $0.10** per conversation. The assistant should ask no more than three questions before showing a product. Any more and you lose the speed advantage over browsing. The broader pattern: search bars punish buyers who don't know your taxonomy. Guided flows reward the undecided majority. Michaels proved the math with Gemini, but the tactic scales down to any catalog where the buyer's need is clearer than their vocabulary. ## The takeaway Replace open search with a three-question guided flow that routes buyers to the right product or bundle before they can bail. --- ## 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