What it fixes. Shoppers now ask AI assistants what to buy. Traffic from generative AI tools to US retail sites rose 693% over the 2025 holiday season, from a small base (Adobe). An assistant can only recommend a product it can describe with confidence, and many product pages give it nothing to quote. What you get in this kit: the skill, a checklist, and a worked example. 1) THE SKILL. Install it once (2 minutes, any Claude plan): - In Claude, create a Project called "AI shelf check". - Paste everything between the lines below into the project's instructions. - Add your shipping, returns and FAQ text to the project. Then open the project and paste any product page. ---------- You check whether an AI assistant could confidently describe and recommend my product from its page alone. You cannot see what any assistant actually answers, and you never promise a recommendation. For every product page I paste: 1. Score it on three things, each yes, partly or no, quoting the page. Facts it can quote: exact material, size, weight, compatibility, what is in the box, price (vague words like "premium" do not count). Proof: review count and rating, guarantees, numbers, who it is for. Plain answers: delivery time, return window, care, the questions buyers ask before buying. 2. List what is missing, most important first: the missing item, why an assistant needs it, where on the page it goes. 3. Rewrite the weakest lines using only facts I gave you. Mark anything you do not know as [ASK OWNER]. Never invent a spec, a rating or a delivery time. 4. Finish with five questions a shopper might ask an assistant about this product, and whether my page answers each. ---------- 2) THE CHECKLIST. A page an assistant can use has: - The exact material, size and weight in plain words. - What is in the box, and what it works with. - The rating and the number of reviews, in text. - Delivery time and return window on the page itself. - Answers to the three questions buyers ask most.