10 shopping questions to track for your Shopify store

Why use a fixed set?

AI shopping answers can vary. If you change the wording every time you check, you also change the test. Ten questions are enough to cover your main products without creating a log nobody maintains.

In Cuebase, each question becomes a tracked query in Share of Voice. Keep the prompt wording, engine, date, and result separate. A mention is one observation. It does not prove placement, traffic, or sales.

Build questions, then add them in Cuebase

Use one of these shapes for each core product or collection. Write in the customer’s language, then keep the wording still.

• Use case: “commuter backpack that fits a 16-inch laptop”
• Budget: “everyday backpack under $100”
• Attribute: “water-resistant backpack for cycling to work”
• Constraint: “lightweight backpack for a personal item”
• Comparison: “backpack similar to a commuter bag but under $100”

Build ten questions across those shapes. Cover different products and collections, but do not add ten ways of asking the same thing. Use the phrase a buyer would use, not an internal collection name.

Open Share of Voice in Cuebase and add the questions as tracked queries. If you start with no data, Cuebase can offer Seed queries with AI. Treat those as drafts: check each one against the worksheet, then remove and re-add a query if the wording does not describe a real buying decision for your store.

An illustrative backpack set

The filled download includes a fictional backpack-store example. It does not contain search results. The “EXAMPLE ONLY” dates and empty result fields are deliberate: a real store must capture its own results.

Run a first probe and read the report

After you add your verified set, select Run first probe. Once data exists, Run probe now checks your tracked queries again. Cuebase’s report separates mention rate, average position, citations, and share of voice. Do not combine those measures into one count.

Use the blank worksheet to preserve the question set and conditions you want to compare. Export CSV downloads the Share of Voice summary in long format: one brand per probe, with the query, engine, mention, citation, and position fields. It is not a raw-answer archive.

For a successful answer that does not mention your brand, record a zero mention for that result. Treat a failed or empty probe as missing data, not zero. Save raw answers or screenshots separately when your study needs that evidence.

OpenAI says ChatGPT shopping results depend on the question and its context, and can consider structured metadata such as price and product descriptions. It also says not every available product will necessarily appear. That is why a single result should be read as a captured answer, not a market-wide ranking.

A Cuebase baseline workflow

1. Set the questions before changing product data

   Add the ten tracked queries in Share of Voice, check their wording, and run the first probe.

2. Review results by metric and engine

   Read the Cuebase report per engine. A brand mention, a product link, and a citation answer different questions.

3. Use the result to choose catalog work

   When a question points to an unclear product fact, open Catalog Readiness and review the related product before making a change.

4. Repeat without rewriting history

   Run the same tracked queries again on your planned cadence. Use Export CSV when you need the report outside Cuebase.

For definitions and a reporting approach, read how to measure AI share of voice. If a tracked question exposes a missing product fact, use the first-10 product worksheet to decide where to work next.

Source

1. OpenAI Help Center, Shopping with ChatGPT Search. Reviewed October 2026.