2026-09-15 — E-Commerce · ChatGPT

Track the products ChatGPT recommends

Ask ChatGPT “best wireless earbuds under $150” and you don’t just get prose — you get product cards with names, prices, ratings and links. For e-commerce brands, that’s a new shelf. And unlike a store shelf, you can query it programmatically.

Where products show up

Two places in a ChatGPT answer:

• shoppingCards[] — the card carousel: product name, brand, price, currency, rating, review count, image and product URL.
• inlineProducts[] — products named inside the answer text itself.

Plus entities[] catches the brands mentioned even when no card renders.

Reading them

curl -X POST https://api.answerline.dev/v1/monitor/chatgpt \
  -H "Authorization: Bearer $API_KEY" \
  -d '{
    "prompt": "best wireless earbuds under $150",
    "country": "US",
    "include": { "shoppingCards": true }
  }'

Each card comes back as a record — product.name, product.price, product.rating, product.reviewCount, product.productUrl, position — so a weekly run answers the questions merchandising teams actually ask:

• Which products get recommended for our category prompts?
• Whose price is shown, and is it right?
• Are we recommended at all — and if not, who is?

A monitoring setup that works

1. Pick 20–50 buying-intent prompts per product line (“best X under $Y”, “X vs Y”, “X for ”).
2. Run them weekly per market — product cards vary by country, and US state targeting matters for price-sensitive queries.
3. Diff run over run: new recommendations, price drift, rating changes.
4. Feed the cited sources[] to content: cards are built from pages the engine retrieved — being the cited review is how you get on the card.

The same shape works for Copilot’s shopping cards and AI Mode’s product clusters — one schema across engines, so the pipeline doesn’t change.

See all fields on the ChatGPT engine page and shopping feature page.

Keep reading