Amazon Reviews & "Customers Say" Scraper — public pages, no login

Reviews and Amazon's AI "Customers say" summary from public product pages, across up to 12 marketplaces, merged and deduplicated. No login, no cookies, no accounts. Products that fail or get blocked are never charged.

What you get, stated plainly

Amazon shows about 8–13 reviews per product per marketplace without signing in. That's a hard limit, and it applies to every public-page scraper. This actor gets more legitimately:

• Multi-marketplace merge: US + CA + AU etc. Each country's page adds its own top reviews; duplicates are removed. Tested: one product returned 20 unique reviews from US + CA + AU.
• "Customers say" topics: each topic (e.g. comfort, fit, sound quality) with positive / negative / mixed sentiment, mention counts, and Amazon's one-line summary. This summarises all of a product's reviews, not just the ones shown.
• Product details: price, list price, availability, rating, rating count, star histogram (%).
• An honest summary row per product: which marketplaces worked, how many reviews were on the pages, how many you received.

Not possible without sign-in, so not offered: full review lists, "all 1★ reviews", date sorting. Public pages show mostly top (often positive) reviews, so a 1–2★ filter may return only a few; use the "Customers say" negative topics for complaints.

Switching from another Amazon reviews scraper?

Paste your existing input: productUrls (strings or { "url": ... }), asins, maxReviews and includeGdprSensitive are understood as-is.

Input

{
  "products": ["B09B8V1LZ3", "https://www.amazon.com/dp/B0DYHY9B6M"],
  "marketplaces": ["US", "CA", "AU"],
  "stars": [],
  "verifiedOnly": false,
  "keywords": []
}

Marketplaces: US, CA, UK, DE, FR, IT, ES, IN, AU, AE, SA, JP. Each is fetched through a proxy in that country.

Output (one dataset; recordType = product, review or summary)

{"recordType": "review", "asin": "B09B8V1LZ3", "marketplace": "US", "reviewId": "R3N0H8AJQZM3P0",
 "rating": 4, "title": "Sweet little device, but a tad hard of hearing",
 "text": "I bought this little guy for my kitchen and over all, love it...",
 "submittedAt": "2026-09-21", "reviewedIn": "United States", "verifiedPurchase": true,
 "variant": "Color: Deep Sea Blue Size: Pack of 1 Configuration: Device only", "helpfulVotes": 3,
 "url": "https://www.amazon.com/gp/customer-reviews/R3N0H8AJQZM3P0"}

{"recordType": "product", "asin": "B09B8V1LZ3", "marketplace": "US", "price": 39.99, "wasPrice": 79.99,
 "availability": "In Stock", "averageRating": 4.7, "ratingCount": 200201,
 "ratingHistogramPercent": {"5": 82, "4": 11, "3": 3, "2": 1, "1": 3},
 "customersSay": "Customers find the Echo Dot to be fantastic...",
 "topics": [{"topic": "quality", "sentiment": "positive", "mentions": 6392, "summary": "Customers find the Echo Dot to be fantastic..."}]}

Pricing

$0.004 per product page (includes "Customers say" and details) and $0.002 per review. Failed or blocked marketplaces and summary rows are free. Typical cost: 1 product × 3 marketplaces ≈ 3 × $0.004 + ~25 reviews × $0.002 ≈ $0.06.

Use it from Claude, Cursor or other AI assistants (MCP)

Your AI assistant can call this scraper directly and pay per result like any other run. Add Apify's MCP server with the Metqo tools:

https://mcp.apify.com?tools=metqo/walmart-product-reviews,metqo/amazon-public-reviews

• Claude (desktop or claude.ai): Settings → Connectors → Add custom connector → paste the URL → sign in to Apify.
• Cursor / VS Code: add to your MCP config:

{ "mcpServers": { "metqo": { "url": "https://mcp.apify.com?tools=metqo/walmart-product-reviews,metqo/amazon-public-reviews" } } }

Then just ask, e.g. "What do verified buyers complain about most for Walmart item 10450114? Group by topic."

More from Metqo Data

• Walmart Product & Reviews Scraper: every Walmart review (not just the top ones) with verified, star, date, keyword and topic-sentiment filters, plus monitoring mode.
• Need scheduled feeds or other retailers? support@metqo.com