AI assistants are starting to build short product lists for one customer at one moment. A shopper says what they need. The assistant picks from real products and shows a list. This check helps a small business see if it is ready to be included.
Examples of the direction: Zomato's MCP server lets an AI tool find restaurants, read real menus and build a cart. Amazon's Interests feature turns a written prompt into a list that keeps updating. These are early. Treat this page as preparation, not a promise of any result.
The 10-point check
- Every product has a clear name. Use words a customer would type.
- Every product has a plain description. Two or three sentences, what it is, who it fits.
- Price is correct and current. A wrong price breaks trust at the last step.
- Stock is current. An assistant should never offer an item you cannot send.
- Categories make sense. Few, clear, no overlap.
- Key facts are in fields, not only in images. Size, material, serving, delivery time.
- Product data is shared in a standard way. Google Search Central recommends structured data and Merchant Center feeds, with fast updates for stock.
- Policies are written down. Returns, delivery area, delivery time.
- A human can take over. Say who answers when the assistant is unsure.
- You know what is paid. If you pay for placement anywhere, label it.
How to score it
Count the points you can answer yes to without checking. Fix the "no" answers in the order above. Items 1 to 4 matter most.
Skills that help
Use a data-cleanup skill for names and descriptions, a feed-check skill for price and stock mismatches, and a review skill for policy text. Look for them in the skill library on this site.
Keep expectations honest
Clean data does not guarantee any assistant will show your product. It removes one common reason you would be left out.