How to choose the first 10 products to improve for AI shopping

Start with ten products

Improving product facts is detail work. Cuebase gives you a practical starting point: scan the catalog, open the lowest-readiness products, and review the gaps it found. Choose ten from that list, finish them, then use what you learn to set the next batch.

Start in Cuebase: scan, sort, review

Open Catalog Readiness and select Scan catalog. Cuebase lists products with the lowest readiness score first and shows a gap count beside each product. Treat that list as the raw material for your first ten; a low score tells you where to inspect, not what to publish.

Open a product with meaningful sales or an important collection role. On its product page, read the Rule breakdown. Then choose Generate improvements only after you have checked the available source facts. Cuebase shows the proposed description, SEO fields, category, and structured attributes for review. Nothing reaches Shopify until you select Apply all to Shopify.

Choose ten from the review queue

Use three plain questions to choose among the products Cuebase has surfaced. Give each a high, medium, or low mark. You do not need a pretend-precise score.

1. Business signal

   Would a clearer page help a product you already sell, promote, or want to make easier to buy? Use recent unit sales, gross margin, inventory you need to move, or a planned campaign. Pick one signal you trust; do not mix figures that mean different things.

2. Question signal

   Can you write a shopper question that this product should answer? “Waterproof hiking boot for wide feet under $200” is useful. “Best boots” is too broad. Start from support tickets, site search, product reviews, sales calls, or your own category knowledge.

3. Fact-gap signal

   Can a buyer confirm the key facts on the product record? Look for a vague title, an incomplete description, no category, missing material or dimensions, unclear variants, old price, or unavailable stock. Shopify recommends assigning a product category and explains that category metafields provide structured attributes for discovery and filtering.

A product that scores well on all three belongs near the top. A popular product with no fact gap can wait. So can a badly documented product with no buyer demand.

A hypothetical example

Example only: imagine a Shopify store selling outdoor gear. Its “Northline 28L daypack” has steady sales, but its page says only “built for every day.” The record does not state laptop sleeve size, pack dimensions, water resistance, weight, or the capacity in the visible description.

That product goes in the first ten. The first fix is not a larger claim. It is finding the verified sleeve dimensions, capacity, materials, water-resistance specification, and variant details, then putting those facts where customers can read them. If a fact cannot be verified, leave it out and ask the supplier or product team for it.

Make the queue in Cuebase

1. Scan, then start with the lowest scores

   In Catalog Readiness, work from the products Cuebase places first. Cross-check that list with products that sell, are in stock, or support an important collection.

2. Write one question for each candidate

   Use an attribute, a use case, and a constraint where it fits: “linen sheet set for hot sleepers,” “ceramic mugs for an office,” or “carry-on backpack under $100.” If you cannot write a useful question, the product may not belong in this batch.

3. Review the flagged gaps

   In each product’s Rule breakdown, identify the fact that keeps the question from being answered. Record that fact in the worksheet, not just “improve copy.”

4. Review suggestions and apply only verified changes

   Use Generate improvements, check each proposed value, then select Apply all to Shopify only when it matches your evidence. Keep each first task small: category, dimensions, a variant, a specific title, or a verified material.

Fix structured facts before polish

Start with product category, brand, price, availability, variants, and the attributes shoppers actually filter on. Shopify says categories should use the predefined Standard Product Taxonomy and that category metafields map to product attributes. Those fields can also help organize your own catalog.

Then make the title and description say the same verified facts in natural language. OpenAI says ChatGPT shopping results consider the shopper’s query and context, as well as structured metadata such as price and product description. That makes clear, current product information useful, but it does not make visibility predictable: ChatGPT selects product results independently, and not every available product is shown.

For a deeper field-by-field review, use our Shopify product data checklist for AI shopping assistants. Once your ten are ready, use the shopping-question worksheet to record the questions you want to observe over time.

Sources

1. Shopify Help Center, Product category. Reviewed October 2026.
2. OpenAI Help Center, Shopping with ChatGPT Search. Reviewed October 2026.