AI Shopping Results Tracker: Products, Shops & Your Brand Rank

When a shopper asks an AI assistant "what's the best cordless vacuum for pet hair?", which products does it recommend, which shops does it send them to, and is your brand in the answer? Give this Actor a list of shopping prompts and your brand or shop names. For every prompt you get the recommended products in order, their prices, the shops and links, the sources the answer cited, and where your brand, product, shop or website appears.

How the answers are made (please read)

Every prompt is answered by OpenAI's official API with OpenAI's own web search tool, the same kind of live, search-backed answer ChatGPT gives. It is not a copy of the chatgpt.com page or its product cards: nothing is scraped from ChatGPT, and ChatGPT's own answer to the same words can differ (it also knows the user, their history and their location). Treat the results as a close, repeatable stand-in for how ChatGPT-style AI shopping answers see your market, not a screenshot of one person's ChatGPT.

AI answers also change from run to run. Use Answers per prompt (up to 5) to ask the same question several times and see how often your brand comes up, and Monitor mode to watch it over time.

Products, prices and shops are read out of each answer by a second model, then every value is checked against the answer text in code: a product name, price, link or shop that is not written in the answer (or its cited links) is dropped, so nothing invented reaches your data.

What you get for every prompt

• Products: each recommended product in order, with position, name, brand, price, currency, the price as written, the shops it can be bought from (name, link, domain), its product link, and any review sites named for it.
• Your brands: for each name in Brands or shops to look for: found, position (the first product that is that brand, sold by that shop, or linked to that site), whether it appears as a shop, whether it is only mentioned in the text, and whether its site is cited.
• Cited sources: every page the answer cited (link, domain, title) and how many pages the search read.
• The full answer text, the shopper country, and a one-line summary such as 3 product(s) recommended, first: Dyson V15 Detect. Found: Dyson #1.
• Run summary: brand share across all answers (found rate, average position, how often #1 and top 3), the products recommended most, the shops linked most, and the sites cited most.

Your own columns stay next to the results.

Who uses this

• Brands and shops checking whether AI assistants recommend them, or their competitors, for the questions their buyers ask.
• SEO and GEO agencies reporting "AI shopping visibility" to clients, across many prompts and countries.
• Marketplace sellers and affiliates finding which shops and review sites AI answers send buyers to, so they know where to be listed.

Input

Give prompts as a plain list, an Apify dataset (for example a keyword list), a CSV or Google Sheet link (shared as "anyone with the link can view"), or inline rows. The prompt column is detected automatically (prompt, query, question, keyword...), or set Prompt field.

• Brands or shops to look for: brand names (Dyson), shop names (Argos) or websites (argos.co.uk). A Brand field column can add names per row.
• Shopper country: a two-letter code (US, GB, DE, AU...) so search results, shops and currency fit that country. A row's own country column overrides it. Optional city and region.
• Answer like a shopping assistant (on by default): asks for several named products with prices and shops, like ChatGPT's shopping answers. Turn it off to send your prompt with no extra instruction.

Example (the default input):

{
  "prompts": ["What are the best running shoes for flat feet?", "Best cordless vacuum for pet hair", "Which espresso machine under $500 should I buy?"],
  "trackBrands": ["Brooks", "Dyson", "Breville", "Amazon"]
}

Output (one row per prompt answer)

{
  "prompt": "Best cordless vacuum for pet hair",
  "country": "GB",
  "checkStatus": "ok",
  "summary": "3 product(s) recommended, first: Dyson V15 Detect Absolute. Found: Dyson #1.",
  "brandFound": true,
  "bestPosition": 1,
  "productNames": ["Dyson V15 Detect Absolute", "Shark PowerDetect Cordless", "Samsung Bespoke Jet"],
  "products": [
    { "position": 1, "name": "Dyson V15 Detect Absolute", "brand": "Dyson", "price": 479.99, "currency": "GBP", "priceText": "£479.99",
      "shops": [{ "name": "Dyson.co.uk", "url": "https://www.dyson.co.uk/vacuum-cleaners/cordless/v15", "domain": "dyson.co.uk" }] }
  ],
  "brandResults": [
    { "name": "Dyson", "found": true, "position": 1, "asShop": true, "mentionedInAnswer": true, "cited": true },
    { "name": "Amazon", "found": false, "position": null, "asShop": false, "mentionedInAnswer": false, "cited": false }
  ],
  "citedDomains": ["dyson.co.uk"],
  "monitorStatus": null
}

Monitor mode: get told when it changes

Switch on Monitor mode and schedule the run (daily or weekly). The first run records each prompt's starting point. Later runs report what changed: your brand appearing, dropping out or moving position, and products newly recommended or no longer recommended. Set a Webhook URL (Slack, Zapier, Make, n8n) and it is only called when something changed. Only changed answers keeps just those rows. Different watchlists use different Monitor names.

Pricing

Pay per event, no subscription:

• $0.05 per prompt answered (about $50 per 1,000 answers). Lower on higher Apify plans.
• $0.01 per CSV or Excel export file, only when you ask for one.
• $0.01 per webhook delivered (confirmed 2xx only; in monitor mode only when something changed).

You are charged only when an answer comes back. Failed answers, empty prompts, duplicates, and rows skipped because of a time or budget limit are free. The OpenAI and web search costs are included in the price; you need no OpenAI key. Tracking 50 prompts weekly is about $10 a month.

Honest limits

• Answers come from OpenAI's API with web search, not from the chatgpt.com app, so they will not match one person's ChatGPT exactly, and they vary between runs. This is the nature of AI answers, not a fault: ask each prompt several times for a fair picture.
• An answer takes about 10 to 20 seconds. Three prompts run at once by default.
• Prices are the ones written in the answer, which come from the pages the search read. They can be out of date; always check the shop.
• The Actor sends only your prompt text and shopper location to OpenAI, with storage switched off.

FAQ

Is this ChatGPT Shopping data? It is the same kind of answer from OpenAI's own models and search, made through OpenAI's official API, not taken from ChatGPT. It shows what OpenAI's search-backed answers recommend for your prompts.

Can an AI agent call it? Yes. Pay per event works through Apify's MCP server and x402, and the output is plain JSON.

Something looks wrong for one prompt. Open an issue on the Actor page with the prompt and I will look the same day.

If this saved you checking AI shopping answers by hand, a review on the Store helps a lot.