Assistant sessions are in your analytics with nothing to credit them to; find which SKUs the answers name, and whose store gets the sale.
“Waterproof trail shoe, wide fit, under £120” is a sentence now, and the answer is a short list with links. AI traffic shows you those visits in your own analytics, and Shopping reads the catalogue you already keep.
No prompt per SKU. Praised asks the way a shopper asks and matches every product the answers name back to your catalogue. For each one you get its share of answers and how often it is described right.
A stale price or a wrong stock claim is checked against the feed you already publish for Google, and counts against described right.
Every shopping question, per assistant, as the answer came back, with your products marked in the list. Under it, the sites the assistant leaned on, and the fixes worth trying first.
Every shopping prompt is classified discovery, comparison or decision, and you find where each one loses the shopper. A brand can look healthy while shoppers browse and lose every purchase.
When an answer names your product, Praised records which merchant it sent the shopper to, whether that merchant lists your product, and whether it was in stock. The split between your store and everyone else’s is a margin problem.
Shopping runs beside your brand runs and its findings land in the same places: one queue, content built from your product facts, an experiment that says whether the change moved anything.
It closes only when a re-check confirms the change landed.
Graded 0–100 by the same linter that decides whether it publishes. A draft can fail.
With “no measurable change” as a possible, and honest, result.
A Shopify shop domain is read without OAuth or an app, and any feed URL works. Compare the plans →
If it isn’t here, the methodology is public, limitations included →
Paste one product page and get a verdict before any ask: what an assistant can read from it, and what it has to guess.