Amazon Seller Data Extractor: AI Sentiment & Feedback Analytics

Amazon Seller Scraper — Ratings, Feedback and Reputation Data

Amazon seller scraper that extracts a storefront's name, star rating, 30/90/12-month rating-window counts, and individual buyer feedback comments as structured JSON — no HTML parsing required. It also computes a derived reputation score from that data and can optionally run AI sentiment/theme analysis on feedback text using your own LLM key. Feed it ASINs, seller IDs, storefront URLs, or search keywords and start pulling seller data in minutes.

What is Amazon Seller Data Extractor: AI Sentiment & Feedback Analytics?

It's an Amazon seller scraper that opens a product or seller storefront page and extracts the seller's profile, ratings, feedback comments, and policy text, then derives non-AI reputation metrics (numeric rating, positive-feedback score, feedback velocity, recency, a blended reputation score) from the same data. An optional module sends each feedback comment's text to an AI provider (Claude, GPT, Gemini, Grok, DeepSeek, Perplexity, or Mistral — your own API key) for sentiment and theme classification. No Amazon account or login is required — it scrapes publicly visible storefront pages over HTTP with residential proxies. Built for sourcing/procurement analysts, marketplace researchers, and developers building reputation-scoring or lead-gen pipelines.

What Amazon seller feedback data is publicly available to scrape?

Amazon publishes a seller's storefront page — name, rating, feedback comments, and policies — to any visitor, no account needed. What's gated sits behind the seller's own Seller Central login.

Data Category	Publicly Available	Restricted behind Seller Central login
Seller name, ID, storefront URL	Yes	—
Star rating & % positive feedback	Yes	—
Buyer feedback comments & dates	Yes	—
30d/90d/12mo/lifetime rating counts	Yes	—
Shipping, return, privacy policy text	Yes	—
Business name/address/phone	If seller discloses	—
Order history, inventory, sales data	No	Seller's own account only
Direct buyer-seller messaging	No	Amazon's messaging system

This Actor only returns publicly visible data — what any visitor sees. Nothing behind a login wall.

What data can I extract with Amazon Seller Data Extractor: AI Sentiment & Feedback Analytics?

Every run returns seller identity fields, ratings and reputation metrics, and full feedback records — with AI sentiment/theme fields added when enabled.

Field Name	Description
seller_name	Seller's storefront display name
seller_id	Amazon seller ID (e.g. A33W53J5GVPZ8K)
url	Seller storefront URL that was scraped
products_link	Link to the seller's full product listing on Amazon
country	Marketplace country, derived from marketplaceDomain
seller_phone_number	Phone number, only when the seller discloses one on the page
description	"About seller" free text
detailed_info[]	Every business-info label row Amazon renders (title/value pairs — business name, address, etc., when disclosed)
return_policy[]	Return/refund policy text, split into sentences
shipping_policies	Shipping policy text
privacy_security	Privacy & security disclosure text
privacy_policy	Privacy policy text
help_content	Return/refund/contact help text
stars	Amazon's own rating text (e.g. "4.5 out of 5 stars")
starRatingNumeric	stars parsed to a float (e.g. 4.5)
rating_positive	Amazon's own positive-feedback text (e.g. "88% positive")
positiveFeedbackScore	rating_positive parsed to a number (e.g. 88.0)
feedbacks_percentages	Star-by-star percentage breakdown (star_5…star_1)
rating_count_m1	Real 30-day rating count, parsed from the page's embedded ratings-window data
rating_count_m3	Real 90-day rating count
rating_count_m12	Real 12-month rating count
rating_count_lifetime	Lifetime rating count
feedbackVelocity	rating_count_m1 ÷ rating_count_lifetime — recent activity as a share of all-time volume
feedbackRecencyDays	Days since the most recent feedback entry
reputationScore	0–100 blended score: 50% positive-feedback rate, 30% log-scaled lifetime rating volume, 20% recency
feedbacks[]	Buyer feedback comments, each with stars, text, date (plus sentiment, sentimentScore, themes when AI is enabled)
sentimentSummary	Per-seller tally of positive/neutral/negative/failed AI classifications (null unless AI ran)
aiSentimentEnabled	Whether AI sentiment/theme analysis actually ran for this row
section_index, section	Which input (ASIN, seller ID, URL, or keyword) produced this row
timestamp	UTC time the row was scraped

Seller identity & storefront details

seller_name, seller_id, url, products_link, country, seller_phone_number, description, detailed_info[]

Ratings, feedback volume & reputation metrics

stars, starRatingNumeric, rating_positive, positiveFeedbackScore, feedbacks_percentages, rating_count_m1, rating_count_m3, rating_count_m12, rating_count_lifetime, feedbackVelocity, feedbackRecencyDays, reputationScore

Buyer feedback & AI sentiment/theme analysis

feedbacks[], sentimentSummary, aiSentimentEnabled

🤖 Add-on: Need additional Amazon data?

Pair this seller scraper with Amazon Best Sellers Scraper to pull top-100 category rankings, price, and review counts, or Amazon Search Products Scraper: Top Rated & Best Reviewed to find and rank products by star rating before resolving their sellers here.

How does Amazon Seller Data Extractor: AI Sentiment & Feedback Analytics differ from the official Amazon API?

Amazon's Selling Partner API (SP-API) is scoped to the seller account that authorizes it — a third-party developer cannot use it to pull another seller's storefront, feedback, or rating data without that seller's own authorization, per Amazon's SP-API developer documentation.

Feature	Amazon SP-API	Amazon Seller Data Extractor
Scope of accessible sellers	Only the authorizing seller's own account	Any public seller storefront, by ASIN/ID/URL/keyword
Approval process	Seller must authorize the app (public or private app workflow)	None — runs directly on Apify
Feedback/rating-window data	Own-account order/feedback reports only	Public 30d/90d/12mo/lifetime rating counts for any seller
Sentiment/theme analysis	Not provided	Optional, via your own AI provider key
Setup	Developer registration + per-seller authorization	Provide ASINs, IDs, or URLs and run

Use SP-API when you manage your own seller account and need order/inventory operations. Use this Actor when you need public reputation data on sellers you don't control — competitors, prospects, or suppliers.

How to use Amazon Seller Data Extractor: AI Sentiment & Feedback Analytics

This Actor runs on the Apify platform — no separate signup or API integration is required to try it.

1. Open the Actor's page in Apify Console and click Start.
2. Provide at least one input — none are required by the schema, but you'll typically fill productAsins, sellerIds, sellerProfileUrls, or keywordInputs.
3. Optionally set marketplaceDomain (e.g. amazon.co.uk) and maxFeedbackPerSeller.
4. Start the run.
5. Download results as JSON, CSV, or another format from the Apify dataset export options once the run finishes.

How to scale to bulk Amazon seller extraction

productAsins, sellerIds, sellerProfileUrls, and keywordInputs are all arrays — list as many rows as you want in a single run, and each is processed concurrently up to maxParallelSellers. keywordInputs additionally runs a real search → product → seller discovery pipeline per keyword, de-duplicating discovered sellers by seller ID, so one keyword can surface many sellers without listing them individually.

What can you do with Amazon seller feedback data?

• A procurement analyst vetting suppliers uses reputationScore and rating_count_lifetime to shortlist storefronts before requesting quotes.
• A marketplace researcher tracking competitor sellers uses rating_count_m1/rating_count_m3/rating_count_m12 against rating_count_lifetime to see whether a seller's recent activity is accelerating or stalling.
• A lead-gen team building an outreach list uses seller_name, detailed_info[], and seller_phone_number (when disclosed) to compile contactable storefronts.
• A brand-reputation team monitors positiveFeedbackScore and feedbackRecencyDays over repeated runs to catch a seller's rating trending down.
• An AI engineer feeds feedbacks[].text through the built-in AI sentiment/theme module (or into their own RAG pipeline) to surface recurring complaint themes — shipping, packaging, service — across hundreds of comments without reading each one.

How does Amazon Seller Data Extractor: AI Sentiment & Feedback Analytics handle rate limits and blocking?

Every request runs through a residential proxy session (Apify's RESIDENTIAL proxy group), with a session warm-up request to the marketplace homepage before each page fetch and randomized delays between requests. Each page fetch retries up to 4 times on a blocked or empty response (robot-check pages, CAPTCHA interstitials, empty bodies), with increasing backoff between attempts; turning on useCaptchaSolver doubles that to 8 attempts with longer backoff — it does not call any third-party CAPTCHA-solving service, it just retries harder. Proxy sessions rotate automatically on a blocked attempt. If all attempts for a given seller are exhausted, that seller is skipped (logged, not pushed to the dataset) and the run continues with the remaining inputs; if nothing is extracted at all, the run logs a suggestion to retry later or set proxyCountryCode to match the marketplace domain.

⬇️ Input

None of the following parameters are required — the Actor accepts any combination of productAsins, sellerIds, sellerProfileUrls, or keywordInputs.

Parameter	Required	Type	Description	Example Value
productAsins	No	array	One ASIN per row; the seller behind each product's offer is resolved and scraped	["B00P8XQPY4"]
sellerIds	No	array	Amazon seller IDs to scrape directly	["A33W53J5GVPZ8K"]
sellerProfileUrls	No	array	Full Amazon seller storefront URLs	["https://www.amazon.com/sp?seller=A33W53J5GVPZ8K"]
keywordInputs	No	array	Search phrases; runs search → product → seller discovery, de-duplicated by seller ID	["handmade ceramics UK"]
marketplaceDomain	No	string	Amazon marketplace domain to scrape	"amazon.com"
language	No	string	Locale for the Accept-Language header; blank lets Amazon decide	"en-US"
aiSentimentEnabled	No	boolean	Turn on AI sentiment/theme analysis of feedback text (default false)	false
aiModel	No	string	AI model/provider to use when AI sentiment is enabled	"claude-haiku-4-5"
aiApiKey	No	string	Your own API key for the selected provider; falls back to a matching env var	"sk-ant-..."
maxFeedbacksForAi	No	integer	Max feedback comments analyzed by AI per seller (1–200, default 20)	20
maxFeedbackPerSeller	No	integer	Max feedback rows collected per seller (0–1000, 0 = all available)	10
maxParallelSellers	No	integer	Max sellers fetched concurrently (1–20)	10
proxyCountryCode	No	string	Proxy country for routing, or "AUTO"	"AUTO"
useCaptchaSolver	No	boolean	Escalate the anti-bot retry ladder from 4 to 8 attempts (default false)	false
sortOrder	No	string	Order of feedbacks[] before truncation: "recent", "helpful", or "rating" (default "recent")	"recent"
proxyConfiguration	No	object	Proxy configuration; this Actor always runs on residential proxies regardless of this setting	{"useApifyProxy": false}

Example input

{
  "productAsins": ["B00P8XQPY4"],
  "sellerIds": ["A33W53J5GVPZ8K"],
  "sellerProfileUrls": [],
  "keywordInputs": ["handmade ceramics UK"],
  "marketplaceDomain": "amazon.com",
  "language": "",
  "aiSentimentEnabled": true,
  "aiModel": "claude-haiku-4-5",
  "aiApiKey": "sk-ant-your-own-key",
  "maxFeedbacksForAi": 20,
  "maxFeedbackPerSeller": 10,
  "maxParallelSellers": 10,
  "proxyCountryCode": "AUTO",
  "useCaptchaSolver": false,
  "sortOrder": "recent",
  "proxyConfiguration": { "useApifyProxy": false }
}

⬆️ Output

Results are pushed to the Apify dataset as one typed JSON row per seller, with a consistent field set across every run. Download as JSON, CSV, Excel, or another format from the dataset's export options.

Example output

{
  "section_index": 1,
  "section": "B00P8XQPY4",
  "url": "https://www.amazon.com/sp?seller=A33W53J5GVPZ8K",
  "seller_id": "A33W53J5GVPZ8K",
  "seller_name": "Coastal Home Goods",
  "products_link": "https://www.amazon.com/s?ie=UTF8&me=A33W53J5GVPZ8K",
  "description": "Family-run storefront specializing in handmade ceramics and home decor.",
  "detailed_info": [
    { "title": "Business Name", "value": "Coastal Home Goods LLC" },
    { "title": "Business Address", "value": "500 Harbor Rd, Portland, ME, 04101, US" }
  ],
  "seller_phone_number": null,
  "stars": "4.5 out of 5 stars",
  "starRatingNumeric": 4.5,
  "rating_positive": "88% positive lifetime",
  "positiveFeedbackScore": 88.0,
  "feedbacks_percentages": { "star_5": "72%", "star_4": "16%", "star_3": "6%", "star_2": "3%", "star_1": "3%" },
  "rating_count_m1": 41,
  "rating_count_m3": 118,
  "rating_count_m12": 512,
  "rating_count_lifetime": 3204,
  "feedbackVelocity": 0.0128,
  "feedbackRecencyDays": 2,
  "reputationScore": 81.4,
  "feedbacks": [
    {
      "stars": "5.0 out of 5 stars",
      "text": "Fast shipping and the mug was packaged very carefully.",
      "date": "on July 21, 2026",
      "sentiment": "positive",
      "sentimentScore": 0.9,
      "themes": ["shipping speed", "packaging"]
    }
  ],
  "sentimentSummary": { "positive": 9, "neutral": 1, "negative": 0, "failed": 0, "analyzed": 10 },
  "aiSentimentEnabled": true,
  "return_policy": ["Items can be returned within 30 days of receipt."],
  "shipping_policies": "Ships within 2 business days via standard carrier.",
  "privacy_security": "",
  "privacy_policy": "",
  "help_content": "",
  "country": "US",
  "timestamp": "2026-07-25T14:02:11Z"
}

How does it work?

Requests go out over Apify's residential proxy network in a sticky session per seller, with a homepage warm-up fetch before the target page to look like normal browser traffic — no headless browser is used, pages are fetched over plain HTTP and parsed with BeautifulSoup. Responses are checked against known Amazon block signals (robot-check pages, CAPTCHA interstitials, short/empty bodies); a blocked response triggers a proxy-session rotation and a retry with backoff. Only what Amazon renders on the public storefront and product pages is extracted — nothing requires a login. The output field set (seller_name, rating_count_m1, reputationScore, and the rest) stays the same from run to run even if Amazon changes its page layout, because the underlying label/section lookups are re-verified against live pages as part of maintaining the Actor.

Integrations

Amazon Seller Data Extractor: AI Sentiment & Feedback Analytics runs on Apify, so it works with anything that can call the Apify API or client libraries.

Calling Amazon Seller Data Extractor: AI Sentiment & Feedback Analytics programmatically

from apify_client import ApifyClient

client = ApifyClient("<YOUR_APIFY_TOKEN>")

run = client.actor("scraper-engine/amazon-seller-data-extractor-ai-sentiment-feedback-analytics").call(
    run_input={
        "productAsins": ["B00P8XQPY4"],
        "marketplaceDomain": "amazon.com",
    }
)

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

Works in Go, Ruby, Node.js, cURL — any language that can make an HTTP request.

No-code tools (n8n, Make, LangChain)

In n8n or Make, use the HTTP Request node (or the dedicated Apify node) pointed at this Actor's run endpoint with your Apify API token, then parse the returned dataset items downstream. In LangChain or a similar agent framework, wrap the Apify client call above as a tool so an agent can request seller reputation data as part of a larger chain.

Is it legal to scrape Amazon seller feedback data?

Scraping publicly available web data, like the seller storefront pages this Actor reads, is generally legal. Amazon Seller Data Extractor: AI Sentiment & Feedback Analytics only returns data any visitor can already see on a storefront page — no login-gated content. This is business/storefront data (seller name, ratings, feedback text, policies) rather than personal data in most cases, though detailed_info[] or seller_phone_number can occasionally expose a sole proprietor's personal name, address, or phone number when Amazon discloses it on the page — treat those fields with care if your use case involves storing them at scale. Consult legal counsel if your use case involves bulk storage of personal data.

Frequently asked questions

What Amazon seller fields does this Actor return?

The top fields are seller_name, seller_id, stars/starRatingNumeric, rating_count_m1/m3/m12/lifetime, and reputationScore — see the full data fields table earlier in this README for the complete list.

Does this Actor require an Amazon account or login?

No. It scrapes publicly visible seller storefront and product pages over HTTP; no Amazon account, cookies, or login session are used.

How many sellers can I extract in one run?

There's no fixed cap in the input schema — productAsins, sellerIds, sellerProfileUrls, and keywordInputs all accept arrays, and each entry is processed up to maxParallelSellers concurrent fetches. Keyword-based discovery finds up to 15 unique sellers per keyword.

What happens if a seller page is missing or the storefront can't be found?

That input is skipped — the run logs it and continues with the remaining inputs. No placeholder or fabricated row is pushed to the dataset for a seller that couldn't be resolved.

Can I scrape multiple Amazon sellers at once?

Yes — pass multiple ASINs, seller IDs, URLs, or keywords in their respective array inputs in a single run; they're processed concurrently up to maxParallelSellers.

Does this Actor work with Claude, ChatGPT, and other AI agent tools?

It's callable as an HTTP endpoint by any agent framework via the Apify API/client, and its own AI sentiment module already calls Claude, GPT, Gemini, Grok, DeepSeek, Perplexity, or Mistral directly using your own provider key.

Is the sentiment analysis genuinely AI-generated, and by what model?

Yes — when aiSentimentEnabled is true, each feedback comment's text is sent to a real LLM API call (Anthropic, OpenAI, or an OpenAI-compatible endpoint for Gemini/Grok/DeepSeek/Perplexity/Mistral) using the model you pick in aiModel (default claude-haiku-4-5) and your own API key. It is not a keyword-matching label; a failed or missing key returns null sentiment fields for that run rather than a faked result.

Does this Actor return data in a format LLMs can use directly?

Yes. Every field is typed, normalized JSON with consistent names across runs — pass it directly to an LLM, index it into a vector store, or feed it to an agent tool without any HTML parsing.

What happens when Amazon changes its layout or anti-bot system?

The output schema is designed to stay stable — field names don't change — and the underlying page selectors are maintained against Amazon's current markup. No specific update turnaround time is published.

Can I use this Actor without managing proxies or browser infrastructure?

Yes — residential proxy sessions, request headers, and block-retry logic are all handled internally; you don't configure or run any proxy or browser infrastructure yourself.

Which fields work best for AI training data and RAG indexing?

For RAG, index feedbacks[].text alongside description and detailed_info[] — the highest-information free text on each row. For structured training data, starRatingNumeric, positiveFeedbackScore, rating_count_m1/m3/m12/lifetime, and reputationScore return as consistent typed numbers across every record.

Related scrapers

Scraper Name	What it extracts
Amazon Best Sellers Scraper	Top 100 best-selling products per category — name, price, ASIN, rank, rating, review count
Amazon Search Products Scraper: Top Rated & Best Reviewed	Amazon search results filtered and ranked by star rating and review count
eBay Seller Scraper	Same seller-reputation extraction pattern, for eBay storefronts
eBay Product Scraper	eBay product listings and pricing data

Your feedback

Found a bug or missing a field? Let us know through this Actor's Issues tab on Apify — reports like these directly shape what gets fixed and added next.