Shopping is the check on the sales AI assistants already steer: which of your products they recommend when a shopper types a need, who they put first instead, and whether the store they send the shopper to can sell it.
Shoppers ask for a need — “waterproof trail shoe, wide fit, under £120” — and the assistant answers with a short list. Shopping asks those needs for you and matches every product the answers name back to your catalogue. You find which of your products come up, and which never do, without writing a prompt per product. The example catalogue: 58 of 214 products named at least once. The 156 that never were are the list to work on.
Every product with how often answers name it and how often they recommend it. The two are counted apart, because a mention is not a sale.
For one need, the list each assistant builds, in order, with your product’s place and the store it linked. The answer itself sits underneath.
Discovery, comparison and decision, each as a rate. The example catalogue reads 44.1% at discovery and 12.2% at decision: the name is heard early and lost right before the purchase.
An assistant can recommend your product and link a shop that can’t sell it. Shopping shows, for every answer that named you, which store got the click and whether the shopper could buy there. In the example workspace, 42 of 61 answers that linked a store linked one that cannot sell the product it named: each one a sale your product earned and nobody closed.
| What the screen shows | Why it matters | What it won’t claim |
|---|---|---|
| The store each answer linked | Who gets the click when an assistant recommends you, your own shop included | Anything about stores no answer linked |
| Answers sent to each store | Where the recommendations you earned actually go | A share of sales: the revenue view lives on AI traffic |
| Which of your products each store lists | Whether a retailer can sell what the answer promised | A listing with no feed: that reads couldn’t check, never zero |
| Can’t buy | Answers that sent the shopper somewhere with no stock of the product | Stock guessed from an old feed: the last read date is on the row |
A broken feed never reads as zero. Anything read from a feed turns into couldn’t check, with the date it was last read.
Shopping reads the product feed you already maintain for Google Shopping, or a Shopify shop’s public product list. There is no app to install, no permission to grant and no prompt to write per product.
Seven views, one question each. Every figure opens the answers it was counted from.
| The question | What you get | View |
|---|---|---|
| How much of our range is visible at all? | Products ever named, recommended this time, can’t buy, stores named | Overview |
| For this need, what did each assistant recommend? | The basket in order per assistant, your place, the store linked, the answer itself | Recommendations |
| Where in the buying journey do we disappear? | Discovery, comparison and decision rates, and the competitors at each stage | Journey |
| What do the answers say about our products? | Size, fit, price and material as stated, against your feed | Attributes |
| Who got the click, and could the shopper buy? | Each store, the answers sent there, the products it lists, can’t buy | Merchants |
| Which products are named, and which never? | One row per product, named and recommended kept apart | Items |
| Which shopper needs are we losing? | The needs asked, with the competitors named in each | Prompts |
Shopping measures what assistants say about your catalogue. Here is what it doesn’t do.
| It does not | What it does instead |
|---|---|
| Connect to Shopify through OAuth or an app | Reads the public product list a shop already publishes, or any feed URL, with no permission scope |
| Chart a store’s share of your sales over time | Shows which store each answer linked; the revenue view lives on AI traffic |
| Treat one shopping answer as the ranking | One measurement account is one shopper; the question is asked again, and a personalised answer is read as a sample |
| Report a broken feed as zero | Every feed-read figure turns into couldn’t check, with the date of the last read |
| Promise placement in a shopping answer | Shows where your range stands, who is recommended instead, and whether a change moved it beyond noise |
Check one product page free, or connect your catalogue and measure the whole range.