Amazon Reviews Scraper Pro

Amazon product reviews at scale across 20+ marketplaces — verified-purchase filtering, star/keyword filters, media extraction, and MCP-ready providerHealth. Multi-API fallback chain. No paid API keys.

---

Why This Actor Is Better

The most reliable Amazon reviews scraper on the Apify Store. Built to outperform every incumbent on the three things that matter: coverage (20+ marketplaces), reliability (Playwright desktop + mobile HTML fallback), and signal quality (verified-purchase filter, star/keyword filters, media extraction, sentiment bands). Every review carries a confidenceScore and the actor emits providerHealth{} so MCP agents can route around failed providers.

Key Features

• 20+ marketplaces — amazon.com, .co.uk, .de, .fr, .it, .es, .ca, .com.au, .com.mx, .co.jp, .in, .com.br, .nl, .se, .pl, .sg, .com.tr, .ae, .sa
• Multi-API fallback chain — Playwright desktop (JS pagination) primary, mobile HTML fallback when blocked
• Verified-purchase filter — isolate the highest-trust reviews with one checkbox
• Star + keyword filters — slice by rating range and body/title keyword
• Media extraction — image and video URLs attached to reviews
• Sentiment bands — automatic very_positive/positive/neutral/negative/very_negative classification
• Confidence scoring — every review gets a 0.0-1.0 score weighted by verified status, body length, and media presence
• Source attribution + providerHealth — MCP agents can route around failed providers
• Cache-first mode — fast_lookup hits the Apify KVS cache for sub-second repeat queries
• Transparent PPE pricing — $3/1K reviews, $1/1K verified, $0.50/1K with media

---

Architecture

Input (queries + mode)
  |
  v
[Cache check] --hit--> return base data (<800ms)
  |miss
  v
[Primary provider: scrape_amazon_reviews_desktop]
  |-- on failure --> [Fallback provider]
  v
[Normalization + field mapping]
  |
  v
[Enrichment layer]
  |
  v
[Confidence scoring + source attribution + providerHealth]
  |
  v
[Progressive dataset push] (one push per result)
  |
  v
[Webhook + MCP-ready metadata]

Every result includes providerHealth{} tracking per-provider status, latency, and errors — making this actor safe to call from MCP agents that need to route around failures.

---

Modes

Mode	Description	Target latency	Use case
fast_lookup	Cache-first, base data only	<800ms cached	Quick lookups, deduplication
enrich	Full enrichment with contact extraction + scoring	~2-4s/result	Detailed analysis
batch	Queue-based, full enrichment, per-item isolation	varies	Large query lists (100+)

---

Input

Parameter	Type	Required	Default	Description
mode	string	—	enrich	fast_lookup | enrich | batch
queries	array	yes	["B08N5WRWNW"]	Search queries
maxResults	integer	—	25	Max results per query (1-1000)
webhookUrl	string	—	—	Webhook for completion notification

Example input

{
  "mode": "enrich",
  "queries": ["B08N5WRWNW"],
  "maxResults": 50
}

---

Output

Results are stored in the Apify dataset as structured JSON. See .actor/output_schema.json for the canonical schema.

Every result includes:

Field	Type	Description
confidenceScore	number	0.0-1.0 reliability score
sources	array	Provider attribution
providerHealth	object	Per-provider status/latency/error for MCP routing
cacheStatus	string	hit | miss | degraded
mode	string	Execution mode used
extractedAt	string	ISO timestamp

---

Pricing

Transparent pay-per-event (PPE) pricing. You only pay for successful results.

Event	Price	When charged
Actor Start	$0.05/1K runs	One-time per run
Result	$0.003/result	Per result pushed
Verified Review	$0.001/result	Charge per review marked 'Verified Purchase' (higher-value signal).
Media Extracted	$0.0005/result	Charge per review with image/video media URLs extracted.

No monthly subscription required.

---

Use Cases

• Product research — aggregate review sentiment before launching a competing product
• Brand monitoring — track review velocity and sentiment across your ASIN portfolio
• SEO/ASO — extract review keywords to inform listing optimization
• Market research — benchmark competitor review counts, ratings, and verified-purchase ratios
• Reputation management — surface negative reviews with media for fast response
• MCP agent workflows — call from LLM agents; providerHealth lets agents route around marketplace failures
• QA / bug tracking — mine 1-2 star reviews for reproducible product issues
• Influencer / reviewer outreach — build lists of high-helpful-vote reviewers

---

Integration Examples

Python (Apify SDK)

from apify_client import ApifyClient

client = ApifyClient("YOUR_APIFY_TOKEN")
run = client.actor("YOUR_USERNAME/amazon-reviews-scraper-pro").call(run_input={
    "mode": "enrich",
    "queries": ["B08N5WRWNW"],
    "maxResults": 50,
})

for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item)

cURL

curl -X POST "https://api.apify.com/v2/acts/YOUR_USERNAME~amazon-reviews-scraper-pro/runs?token=YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"mode":"enrich","queries":["B08N5WRWNW"],"maxResults":25}'

MCP (Model Context Protocol)

This actor is MCP-ready. Register it via @apify/actors-mcp-server:

$npx -y @apify/actors-mcp-server --tools actors,YOUR_USERNAME/amazon-reviews-scraper-pro

Agents can then call call-actor and use providerHealth + confidenceScore to route around failed providers and filter result quality.

---

FAQ

Q: Do I need an Amazon API key?

No. This actor scrapes public review pages via Playwright + httpx. No Amazon Product Advertising API key or seller credentials required.

Q: Which marketplaces are supported?

20+ marketplaces via the marketplace input: com, co.uk, de, fr, it, es, ca, com.au, com.mx, co.jp, in, com.br, nl, se, pl, sg, com.tr, ae, sa. Default is com.

Q: How does the fallback chain work?

The primary Playwright desktop scrape handles JS-rendered pagination. If Amazon blocks Playwright (CAPTCHA, 403, timeout), the actor falls back to a mobile-HTML httpx fetch. The providerHealth field on every result shows which providers succeeded and their latency.

Q: What does verified-purchase filtering do?

When verifiedOnly is true, only reviews marked 'Verified Purchase' by Amazon are returned. These are higher-trust signals for product research.

Q: How are reviews paginated?

The actor follows Amazon's review pagination (up to 50 pages on desktop, 10 on mobile fallback) until maxReviewsPerProduct is reached or the well runs dry.

Q: Can I call this from an LLM agent?

Yes. The actor is MCP-ready. Every result includes providerHealth{}, confidenceScore, ratingBand, and verifiedPurchase so agents can make routing and filtering decisions.

---

Legal & Compliance

This actor scrapes publicly available data. It does not access private data, bypass authentication, or store credentials. Users are responsible for complying with applicable data protection laws (GDPR, CCPA, etc.) and the target platform's Terms of Service.

This actor is intended for legitimate research, analysis, and outreach use cases. It must not be used for spam, harassment, or unlawful activity.

---

AI-DLC / Data Lifecycle

This actor follows AI-DLC principles for ethical data handling:

• Collection — Public data only; respects robots.txt and rate limits.
• Processing — In-memory normalization; no PII logging.
• Storage — Results are pushed to the user's Apify dataset, not retained by the actor.
• Usage — Designed for analysis, enrichment, and legitimate outreach.
• Disposal — No long-term caching of user data between runs (cache TTL 3600s for base results only).

---

Enhancement Roadmap (API / MCP Integrations)

• Multi-step orchestration via LangGraph conditional workflows
• Vector store integration for semantic deduplication across runs
• Additional paid enrichment APIs (user keys, disabled by default)

---

Changelog

• v1.0 — Initial release: 20+ marketplaces, Playwright+mobile fallback, verified-purchase filter, star/keyword filters, media extraction, MCP-ready metadata

---

Links

• Apify Store: https://apify.com/YOUR_USERNAME/amazon-reviews-scraper-pro
• Documentation: See this README
• Support: Open an issue on the actor's Apify Store page