Amazon Public Review Snapshots

Extract the limited subset of customer reviews Amazon exposes on public product pages by ASIN or product URL. Export review text, ratings, dates and product metadata to JSON, CSV or Excel.

Full-review scraping is no longer supported in any marketplace. This Actor does not enumerate all reviews, paginate dedicated review listings, or reliably collect the newest reviews. Amazon usually exposes 8–14 top/featured reviews per product, but this is a typical source subset, not a guaranteed yield: some pages expose fewer reviews, metadata only, or no usable response. A successful run is not proof of complete coverage.

What does it do?

The Actor requests public desktop product pages, with a mobile product-page fallback when no review records are exposed. It extracts the records present on those pages in Amazon's source order. It does not sign in, use Amazon account credentials, or request dedicated review-list pagination.

The available subset can differ between runs. Raising the limit, changing filters or running daily cannot recover reviews Amazon does not expose in that subset. Product rating counts describe the product, not the number of reviews this Actor can retrieve.

October 2026 functionality change

This Actor cannot reliably deliver full review history or newest-first listings without sign-in, which it does not use. It is now a best-effort public product-page snapshot tool across all marketplaces, not a full-reviews scraper or complete new-review monitoring service.

• Full-review pagination has been removed globally.
• sort is no longer an input option. Saved sort=recent or sort=helpful inputs remain accepted for compatibility but are ignored with an explicit warning; output retains product-page order.
• Star/date filters operate only on available snapshot records.
• Japan remains a best-effort snapshot marketplace, just like the other countries. This is not a Japan-only discontinuation or a claim that Japan reviews have been recovered.
• The Actor slug and output fields remain unchanged. Existing API integrations still use automation-lab/amazon-reviews-scraper.

Who is it for?

Analysts and product teams who can use a small, source-limited sample for exploratory sentiment analysis or qualitative feedback. It is not suitable for exhaustive review corpora, daily newest-review alerts, guaranteed star-specific coverage or Seller Central parity.

Why use this Actor?

• Public product-page snapshots from 11 Amazon marketplaces.
• ASIN or product URL input, up to 50 unique ASINs per run.
• Typed review records and separate product metadata.
• Local star and date filters over the available subset.
• Bounded retries and proxy recovery with explicit access-limit messages.
• Familiar export formats and unchanged review output fields.

No minimum review count, completeness, newest-first coverage or repeatability is promised.

Marketplaces and access limits

Best-effort domains: US (amazon.com), UK (amazon.co.uk), DE (amazon.de), FR (amazon.fr), IT (amazon.it), ES (amazon.es), CA (amazon.ca), JP (amazon.co.jp), IN (amazon.in), AU (amazon.com.au), SG (amazon.sg).

Select the domain using marketplace. Product URLs supply an ASIN; their domain does not override that field. Proxy country does not automatically match each marketplace: US is pinned to US; other markets are unpinned. The Actor tries the account's US datacenter pool and bounded residential recovery. Neither guarantees access.

Amazon can return CAPTCHA, sign-in pages or synthetic 404s for live listings. A 404 alone does not prove a product is unavailable. Seller Central visibility does not establish public anonymous review availability.

How much does it cost to extract Amazon public review snapshots?

Pricing remains Pay-Per-Event:

Event	Price
Run start	$0.01 once per run
Review emitted	$0.002 FREE, $0.0018 BRONZE, $0.0016 SILVER, $0.0013 GOLD, $0.001 PLATINUM, $0.0008 DIAMOND

Worked example: at FREE pricing, 5 emitted reviews cost $0.02 including start; 10 cost $0.03. These examples describe actual emitted records, not guaranteed yields. The run-start charge applies even if access fails or filters leave no reviews. Metadata is not charged as a review. Review records filtered out before emission are not charged. Consult the Store pricing tab for the active tariff.

How to use it

1. Enter ASINs or product URLs.
2. Select the matching marketplace.
3. Choose a small snapshot limit; the default is 20, with a maximum of 100 per ASIN. A larger limit does not unlock more reviews.
4. Optionally filter the available subset by star rating or date.
5. Run and inspect both review output and the terminal access/status message.
6. Export results from the dataset. Do not interpret empty output as proof that no new or matching reviews exist.

Input parameters

Parameter	Default	Semantics
asins	required	ASINs or product URLs; up to 50 unique ASINs
marketplace	US	Selects domain; best-effort snapshot in all 11 markets
maxReviewsPerProduct	20	Upper bound, not yield; product-page subset only, no pagination
filterByStars	all	Exact stars, positive (4–5) or critical (1–2), within subset only
newerThan	none	Strict YYYY-MM-DD cutoff over subset, not complete monitoring
maxRequestRetries	2	Bounded request retries, maximum 2

With newerThan, the review limit bounds deduplicated star-matching date candidates before cutoff filtering. Localized dates that cannot be safely parsed are retained and logged. IDs are deduplicated across the run; sibling variations can therefore reduce counts. Reviews are attributed to the variation actually served by Amazon.

Legacy sort is ignored with a warning. It does not request newest-first or most-helpful enumeration.

{
  "asins": ["0321965515"],
  "marketplace": "SG",
  "maxReviewsPerProduct": 10,
  "filterByStars": "all",
  "maxRequestRetries": 2
}

Output

The default dataset and named reviews dataset contain the same emitted review records. The named product_metadata dataset contains metadata separately. Access aliases through the run's Storage tab or the Apify API.

Review field	Meaning
asin, marketplace	Served product variation and marketplace
reviewId	Stable source review ID used for deduplication
title, body	Source review headline and text
rating	Numeric star rating, where exposed
date	Original localized source date text
author, authorUrl	Public reviewer display name/profile link
isVerifiedPurchase	Boolean purchase badge
helpfulVotes	Numeric helpful vote count
reviewUrl	Source review link; it may require sign-in when opened
scrapedAt	ISO timestamp

Product metadata includes asin, productName, productUrl, overallRating, totalReviewCount, marketplace, scrapedAt. Retry attempts can produce more than one metadata record for a requested ASIN. Rating totals are not snapshot coverage guarantees.

Status and empty results

• Useful reviews: succeeds with source-limited output.
• Available reviews excluded by filters: succeeds with no matches in this subset, not a claim about all reviews.
• Useful metadata but no review records: can succeed with an explicit access-limit status and metadata only.
• No useful source response after bounded recovery: fails; it does not fabricate product unavailability.

The run-start fee still applies to empty or failed runs. No review events are charged without emitted reviews.

Tips for best results

Use small batches, inspect the available sample and keep source limitations in downstream reports. Date filters can reduce repeated snapshot output, but cannot establish a complete daily checkpoint. Never infer that an unseen review was deleted or that no recent complaints exist.

The runtime parses each page once and reclaims page trees between requests. The 512 MB allocation has a bounded JavaScript heap; memory pressure does not silently skip requested products. Access limitations can still prevent requested output.

Integrations

Export snapshots to Google Sheets, Pandas, BigQuery, Snowflake or BI tools for qualitative analysis. Apify webhooks, Zapier and Make can forward completed snapshots. Such integrations do not expand source coverage and should not advertise exhaustive new-review alerts.

API usage

import { ApifyClient } from 'apify-client';
const client = new ApifyClient({ token: 'YOUR_API_TOKEN' });
const run = await client.actor('automation-lab/amazon-reviews-scraper').call({
    asins: ['0321965515'], marketplace: 'SG', maxReviewsPerProduct: 10,
    filterByStars: 'all',
});
const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items); // Limited public product-page subset, not all/newest reviews.

Python:

from apify_client import ApifyClient
client = ApifyClient("YOUR_API_TOKEN")
run = client.actor("automation-lab/amazon-reviews-scraper").call(run_input={
    "asins": ["0321965515"], "marketplace": "SG",
    "maxReviewsPerProduct": 10, "filterByStars": "all",
})
print(client.dataset(run["defaultDatasetId"]).list_items().items)

cURL (set your own APIFY_TOKEN locally; never share it):

curl -X POST \
  'https://api.apify.com/v2/acts/automation-lab~amazon-reviews-scraper/runs' \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H 'Content-Type: application/json' \
  -d '{"asins":["0321965515"],"marketplace":"SG","maxReviewsPerProduct":10,"filterByStars":"all"}'

This asynchronous request returns run metadata, not review rows. Retain data.id, check /v2/actor-runs/{runId} until terminal, and read /v2/datasets/{defaultDatasetId}/items?limit=20&offset=0 only after success. Do not restart a run to check its status.

Use with AI tools via MCP

Ask an MCP-compatible client to extract the available public product-page review snapshot and summarize its limitations, not to get all reviews or fetch the newest 50 reviews. A tool call has the same source limits as a regular run.

Claude Code:

claude mcp add --transport http apify \
  "https://mcp.apify.com?tools=automation-lab/amazon-reviews-scraper"

Cursor (.cursor/mcp.json), using its HTTP MCP support:

{
  "mcpServers": {
    "apify": {
      "url": "https://mcp.apify.com?tools=automation-lab/amazon-reviews-scraper"
    }
  }
}

Claude Desktop (claude_desktop_config.json), using the mcp-remote bridge when direct remote connections are unavailable (requires Node.js):

{
  "mcpServers": {
    "apify": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.apify.com?tools=automation-lab/amazon-reviews-scraper"]
    }
  }
}

VS Code (.vscode/mcp.json):

{
  "servers": {
    "apify": {
      "type": "http",
      "url": "https://mcp.apify.com?tools=automation-lab/amazon-reviews-scraper"
    }
  }
}

Complete the client's Apify authentication prompt with your own account; do not put credentials in a shared configuration. Discover actual tool names and argument schemas with scoped tools/list. Actor selection also exposes run/storage helpers, not exactly one tool. Discovery alone does not authorize a run.

Example prompt: "For ASIN 0321965515 on SG, extract up to 10 available public product-page reviews and summarize the sample with its coverage limitations."

For an authorized invocation, start once and retain the run ID, status and storage IDs. call-actor waits at most 45 seconds (waitSecs, default 30); a returned ID is not a completed result. Follow the same run with get-actor-run, for example at 2, 4, 8 and then 10-second intervals with a 120-second total deadline. Keep the client timeout longer than the requested server wait plus transport margin. On timeout check the known run instead of restarting; if its ID is unknown, report uncertainty. At the deadline stop polling and report pending with the ID; terminal failures remain failures and any rows are partial.

After success, retrieve get-dataset-items in bounded pages with explicit limit, offset and needed fields. For example, use 20 source rows per page and a total budget of 100 rows or 64 KiB of serialized UTF-8 results admitted to model context, whichever comes first. Track source offsets and pagination metadata, not transformed row counts. Prefer untransformed JSON; enforce the byte budget host-side before model injection, disclosing omitted or summarized oversized rows. If your client cannot intercept responses, do not claim a hard byte bound. Stop at exhaustion or budget and report count, next offset and any partial/truncated status. Full exports stay in storage or downstream processing. These are consumer context limits, not changes to Actor collection scope. Discovery/cache behavior depends on the client; no numeric token-savings promise is made.

Private error diagnostics

Failed operations send sanitized diagnostic input, exceptions, and Actor/build/run IDs to our private GlitchTip service for repair. Secret fields and URL queries are removed; reports are retained for 30 days. These diagnostics do not expand access to Amazon or provide signed-in review coverage. No AI is used to generate review output. Product requests use company-funded Apify proxies; no additional service account is required. Public reviewer names/profile links are included only as part of requested review records. Run output and logs follow your Apify account retention settings and can be deleted through Apify; proxy sessions are kept in memory, not persisted. This Actor is independently operated and is not affiliated with or endorsed by Amazon.

Legal considerations

Only anonymously accessible product-page content is requested. Amazon's terms can restrict automated access; users must assess applicable terms, privacy obligations and legal requirements. This Actor does not bypass authentication or accept Amazon credentials. Do not impersonate reviewers or misrepresent the scope or source of these records.

FAQ

Can I get all reviews or the latest reviews? No. Full-review scraping and dedicated pagination are no longer supported across any Amazon country.

Does Japan work? It remains a best-effort public product-page snapshot domain. Specific listings may expose no usable records; the prior customer's full/newest JP monitoring need has not been recovered.

Will a higher limit or retries recover missing reviews? No guarantee. Limits cap the subset; retries handle transient failures only.

Are prices or images included? No. Use Amazon Scraper for product price and listing workflows.

Related Actors

Amazon Scraper provides product listing and price data, not a promise of full review coverage.

Built by automation-lab.