Ecommerce Academy · Chapter 02
Track products, not only the brand
Shopping answers recommend products, and a count of brand mentions can miss what happened to yours. Your product can sit in the product cards while your store's name never appears in the text. Your brand can be named while a rival's product takes the recommendation. Or your product can be shown with another store as the seller. Track your key products by name, and for each answer record whether the product appeared, how it was described, at what price, and which seller got the link.
Why the brand count misses it
OpenAI's documentation shows how many separate decisions sit inside one ChatGPT shopping answer. Choosing the products comes first. Then, when a shopper opens one, ChatGPT ranks the merchants offering it by availability, price, quality and whether the merchant is the maker or main seller. The price in the first response usually comes from the first merchant listed, which may not be the cheapest. ChatGPT may also rewrite product titles into simpler ones, add labels such as "Most popular", and summarize reviews.
Each of those steps is a place where your product can win or lose while your brand count stays exactly where it was.
What to record for each tracked product
| Field | What it tells you |
|---|---|
| Shown, and under what name | Whether it appeared, even with a shortened title |
| Position among the products | How prominent it was |
| Price and currency shown | Whether the answer matches your own price |
| Seller listed first | Who is likely to get the click |
| Label or review summary | How the answer framed it |
| Page cited | Your product page, or someone else's |
Search each answer for your model numbers and short names as well as the full title. A simplified title can drop exactly the part you'd search for.
Which products to track
Not the whole catalog. Pick the products that carry the business: top sellers, high-margin lines, a recent launch, and one representative from each category. Start with ten or so and add more only when a product earns its place.
For each one, add a question that names it, such as "Is the [product] good for [use]?", and note the category questions where it ought to be recommended. Keep the named-product questions in their own list, next to your store prompt set, so a single best seller can't lift the numbers for its whole category.
Where to start
Choose five products. For each, write one question that names it and pick two category questions where it should be recommended. Run them in the assistants you care about, fill in the six fields above, and look for the gap that repeats: the product missing entirely, shown at the wrong price, or shown with another seller first. The third pattern has its own lesson: retailers and marketplaces cited instead of you.