Shopify maintains a global product catalog that AI agents query through its agentic commerce tools. When an assistant looks for products to recommend, this catalog is what it searches, and its copy of your product is what it quotes. Octolift asks that catalog the way an agent does, so what you see is what an assistant has to work with.
This panel on the analysis page shows the summary the catalog has built for your product from your content:
The catalog infers all of this from your listing, so a Not specified is a gap worth filling in the description or the category fields. If the product is not in the catalog yet, the panel says so (the score breakdown lists the usual causes); if the catalog could not be reached, it asks you to scan again.
Rather than guessing keywords, Octolift writes a handful of search phrases a shopper might type for your product, from its title, product type and category, in the main language of the country being checked. Each phrase is then run as a real search against the catalog, for a shopper in that country, and the top ten results are read. For every question the table shows:
The suggested questions are refreshed weekly; the rankings are live each time you open the product, and cached only for the rest of your visit. Ranking mixes many signals: listing completeness, popularity, buyer engagement and brand recognition. Completeness is the part you can improve directly, which is what the score breakdown and Improve with AI are for. The rest builds over time.
The catalog answers per buyer country and shipping destination, and shoppers phrase things differently per market, so a rank is always for one country. Octolift defaults to your home market. If your store ships to more than one country, a country dropdown appears at the top of the panel; pick a market and press Check to write new questions in that market's language and rank them there. Questions checked in one country are kept separate from another.
Type any phrase a shopper might ask and press Check rank. It runs the same live catalog search for the selected country and adds a row with your rank, the winning product and its brand. Use it to test the exact wording your customers use.
Press Track on any row to keep that question, with its rank at the time, on the Tracked tab for that product and country. You can track up to 25 questions per country per product, and untrack any of them by pressing Tracking. The Tracked tab is a saved list, so a tracked row keeps the rank it had when you tracked it. To refresh one, untrack it, run it again under Try your own search, and press Track on the fresh row.
Octolift does not ask ChatGPT, Claude, Gemini or Perplexity directly and does not report a position in them. A single API key answers differently from the consumer product, is not deterministic between runs and varies by country, so a number produced that way would look precise while meaning nothing. Measuring the catalog those assistants read is the part that is both true and yours to improve.
Octolift also does not crawl your product pages or your theme, read structured data from them, or measure page speed. The only storefront pages it ever fetches are the policy pages an assistant would follow, when it tests how an assistant answers a shopper. Every other check reads your product through Shopify and the catalog.
Open Support in the app menu to chat with us, or email support@craftshift.com. Tell us your store URL and the product you were looking at and we will pick it up from there.
The two halves of the AI visibility score, every check and its points, the rating bands, and when a product is scored again.
The sales AI assistants brought you: which channels count, how they are attributed, and when the card is hidden.
Short answers about scoring, suggestions, AI visibility, agentic sales, plans, privacy and troubleshooting.