Guides

Web Scraping Amazon: A Guide to Products, Offers & Reviews

Export Amazon product data, offers, and reviews to CSV with Python, no-code tools, or APIs. Verify review pagination and troubleshoot 503 errors.

Kevin LinKevin Lin13 min read

TL;DR: Export Amazon product and offer data to CSV to track prices and stock. For reviews, verify that your source returns individual written text and that page two adds unique records. Validate your Python parser with a local HTML sample, then build a recurring workflow around an API or your own collection pipeline.

If you’re web scraping Amazon to track prices or stock, start by exporting product and offer data to CSV. For customer feedback, collect individual written reviews and check whether page two returns any new ones. Both workflows can start with the same Amazon product ID (ASIN), but they may need different data sources.

If you need a spreadsheet today, start with a no-code export. To build your own pipeline, use the Python example below to validate your CSV parser before connecting a live data source. For recurring jobs, test an API with one ASIN or configure your own collection schedule and network settings. Already getting 503 errors, empty rows, or repeated reviews? Start with troubleshooting or the review pagination section.

What Data Can You Collect from Amazon?

Start with the row you want in your export. A search result is an appearance in a list. A product record describes a specific item or variation. An offer describes a seller’s terms, and a review contains one customer’s feedback. Keeping those records separate makes it easier to update a price without duplicating the product or overwriting a review.

Data type Start with Keep in each row
Amazon SERP (search results page) Search term or category Keyword, position, ASIN, marketplace, sponsored flag when available
Product detail Product URL or ASIN Title, variant, displayed price, currency, availability, average rating
Seller offer ASIN and an offer source Seller, condition, price, shipping and observation time
Individual review ASIN and a review source Review ID when available, stars, date, title and text

For example, search for insulated travel mugs, save the ASINs from the results, then look up details for the ones you want to track. If you need to compare sellers, keep offers in a separate file instead of treating the price you first saw as the only price. Decide the marketplace and delivery context before comparing prices across records.

Find ASINs in Search Results

Search for a product or open a category or Best Sellers list, then copy the ASIN from each product link. If you also need rankings, record the query or list URL, marketplace, position, and whether the result is sponsored. Keep keyword positions separate from category ranks, and look up product details once per marketplace and ASIN.

Amazon search results for insulated travel mugs showing product cards, prices, and ratings

How to Scrape Amazon: Choose the Right Method

Test one representative ASIN or search query before collecting in bulk. Confirm that the source returns the fields you need for the right marketplace, then check what it costs to get a usable row. Use the table to choose how much of the workflow you want to run yourself. If you only need product fields, our comparison of four methods for collecting Amazon product data can help you choose a source.

Method Use it when Check before committing
No-code export You need a small, one-time dataset The export includes your required columns and enough records
Python You need custom fields, cleaning or joins You have access to a suitable source and can maintain the parser
API You need scheduled jobs or queries for new ASINs Sample responses match your fields, marketplace and coverage needs

Build access and data-use requirements into your workflow. Amazon’s published terms for sellers restrict automated extraction within their scope. Check the relevant marketplace terms and your source’s license, use an authorized collection route, and collect only the fields your project needs.

A self-hosted open-source scraper is another option if you want a ready-made parser without writing one from scratch. You manage the runtime, source integration, and network settings. Whether you self-host or pay for a managed API, test one known ASIN or query before committing: check the fields you need, whether page two contains new records, and what failed or repeated requests cost under the provider’s plan.

Export Products Without Code

For a no-code product export, paste a narrow Amazon search or category URL into the Web Scraper product listings template. Run a small sample, download the CSV, and check one row against its product page: match the ASIN and variant, then compare the price and currency. Use the separate review workflow below for written reviews.

How to Scrape Amazon Product Data with Python

When building an Amazon web scraper in Python, separate fetching from parsing. A local HTML fixture lets you check extraction logic and CSV formatting before connecting a live source. The example below uses a synthetic product record with a simplified JSON-LD structure.

Create the product sample

Use Python 3.11 or later in a terminal. The example uses only the standard library. The local CSV checks were verified on November 19, 2026, with Python 3.14.4 on Windows. In an empty directory, save this sample as product.html.

htmlhtml
<script type="application/ld+json">
{
  "@type": "Product",
  "sku": "DEMOASIN01",
  "name": "Insulated Travel Mug",
  "offers": {
    "price": "24.99",
    "priceCurrency": "USD",
    "availability": "https://schema.org/InStock"
  },
  "aggregateRating": {"ratingValue": "4.3", "ratingCount": 87}
}
</script>

Export the product to CSV

Save the script as export_product.py. It reads exactly one Product object from a JSON-LD block, checks the identifier and title, and writes one CSV row. Missing optional fields remain blank.

pythonpython
import csv
import json
from html.parser import HTMLParser
from pathlib import Path


class ProductParser(HTMLParser):
    def __init__(self):
        super().__init__()
        self.parts = None
        self.products = []

    def handle_starttag(self, tag, attrs):
        if tag == "script" and dict(attrs).get("type") == "application/ld+json":
            self.parts = []

    def handle_data(self, data):
        if self.parts is not None:
            self.parts.append(data)

    def handle_endtag(self, tag):
        if tag == "script" and self.parts is not None:
            raw = "".join(self.parts)
            self.parts = None
            try:
                item = json.loads(raw)
            except json.JSONDecodeError:
                return
            if isinstance(item, dict) and item.get("@type") == "Product":
                self.products.append(item)


parser = ProductParser()
parser.feed(Path("product.html").read_text(encoding="utf-8"))
parser.close()
if len(parser.products) != 1:
    raise SystemExit("Expected exactly one Product object.")

product = parser.products[0]
offer = product.get("offers") or {}
rating = product.get("aggregateRating") or {}
if not isinstance(offer, dict) or not isinstance(rating, dict):
    raise SystemExit("Expected one offer and one rating summary.")

row = {
    "asin": product.get("sku"),
    "title": product.get("name"),
    "price": offer.get("price"),
    "currency": offer.get("priceCurrency"),
    "availability": (offer.get("availability") or "").rsplit("/", 1)[-1],
    "rating": rating.get("ratingValue"),
    "rating_count": rating.get("ratingCount"),
}
if not row["asin"] or not row["title"]:
    raise SystemExit("Missing product identifier or title.")

with open("products.csv", "w", encoding="utf-8", newline="") as output:
    writer = csv.DictWriter(output, fieldnames=list(row))
    writer.writeheader()
    writer.writerow(row)

Run python3 export_product.py from the same directory, or py -3 export_product.py on Windows. The generated products.csv should contain:

csvcsv
asin,title,price,currency,availability,rating,rating_count
DEMOASIN01,Insulated Travel Mug,24.99,USD,InStock,4.3,87

Inspect the CSV to verify the ASIN, price, and currency. Remove the sample’s price field and rerun the script: the price cell should stay blank rather than defaulting to zero. The sample uses sku for its demo ASIN; map the ASIN from the appropriate field when integrating another source.

To connect a live source, pass the HTML returned by your HTTP client to parser.feed() and adapt the extraction logic to that response’s structure. Live pages may use DOM elements, nested JSON, or multiple products rather than this simplified JSON-LD object. For variants or multiple seller offers, decide which records get their own rows before updating the CSV schema.

PowerShell showing Python exporting a synthetic Amazon product record to CSV

Export Live Offers via Amazon SP-API

With an authorized SP-API application that has Product Pricing access, you can export offer data directly from Amazon’s Selling Partner API. Apply the same field and CSV checks to one ASIN before scheduling a larger batch.

  1. Request one ASIN. Copy the ASIN from the product’s /dp/ URL. In your app, call getItemOffers. For Amazon US, use the North America endpoint and set MarketplaceId to ATVPDKIKX0DER. Set ItemCondition to New and CustomerType to Consumer. The request path looks like this:
httphttp
GET /products/pricing/v0/items/{ASIN}/offers?MarketplaceId=ATVPDKIKX0DER&ItemCondition=New&CustomerType=Consumer

Replace {ASIN} with your product’s identifier and send the request through the North America SP-API endpoint, https://sellingpartnerapi-na.amazon.com. Use your app’s access token in the x-amz-access-token header and keep it out of CSV exports and logs. For a 401 or 403, check token validity and app permissions before retrying.

  1. Save one offer to CSV. Under payload.Offers, choose one offer and read ListingPrice.Amount and ListingPrice.CurrencyCode from that same entry. Save the condition, marketplace and UTC time with the price. Listing price excludes the separately reported shipping charge. If the response has no offers, record that outcome separately instead of adding a zero-price row. Example CSV output:
csvcsv
asin,marketplace_id,condition,listing_price,currency,observed_at_utc
DEMOASIN01,ATVPDKIKX0DER,New,24.99,USD,2026-11-17T10:00:00Z
  1. Check the row before scaling. Compare the returned ASIN and marketplace ID with your request. Then check that the CSV amount and currency match the selected offer and that the file has exactly one data row. If you also compare against the storefront, match the variant, seller, condition and delivery location. A different offer or shipping charge can explain a price mismatch.

How to Collect Amazon Reviews Beyond the First Page

Before you run an Amazon review scraper, check a sample response from your target marketplace. Look for one record per written review with actual text. Catalog ratings and review insights are useful for other tasks, but they won’t give you the individual comments you need for a review CSV.

Get Your First Written Review

To find out why buyers say the mug lid leaks, start with one product URL in an Amazon review scraper or review feed. If it accepts only an ASIN, select the marketplace separately. Keep the first run to one page or a small review limit.

Preview the output. If you see only ratings or a summary, choose a source that returns individual review text. Export one review per CSV row and keep the ASIN, marketplace, review ID (if available), stars, date, title, and text.

Example CSV row: DEMOASIN01, amazon.com, DEMO-R1, 2, 2026-11-01, Lid leaks, The lid leaks when tipped.

Test Whether Page Two Adds Reviews

If your source supports pagination, keep the same marketplace, sort order, and filters when requesting page two. Count the review IDs you did not see on page one. If IDs are missing, use the date, title, and text to flag possible duplicates. If there is no next page or the same records repeat, report the unique reviews and date range you actually collected.

Illustrative Amazon review pagination showing one new review and one duplicate on page two

What Does an Amazon Reviews API Actually Return?

If you plan to automate review collection, compare what each type of Amazon reviews API returns before integrating it.

Source What to expect What to verify
Amazon Creators API Catalog information, offers, images and variations in the documented operations Access eligibility and whether the catalog fields meet your needs
Amazon SP-API Customer Feedback Review and return insights, with review insights at the ASIN and browse-node level Marketplace support, access roles and which insight fields you can use
Third-party review feed Depends on the provider: it may return summaries, a limited sample or individual records Actual text fields, pagination, date range, marketplaces, usage rights and billing

Amazon’s Creators API reference describes catalog operations, while its Customer Feedback API documentation covers review and return insights. For individual written reviews, evaluate a source that explicitly supplies review records. Test several ASINs and check whether retries, empty responses, and duplicate pages consume your quota. Endpoint status matters too: Scrapingdog’s Amazon Reviews API documentation lists pagination and filter settings but displayed a temporary-unavailability notice when checked on November 19, 2026.

How to Fix Amazon Scraping Errors and Bad Data

Start with the symptom in the table. Check whether the request returned the expected page or JSON first, then inspect missing fields or inconsistent values.

Symptom Check first Next step
HTTP 503 or 429 Response body, Retry-After header, source status, and errors by exit IP Respect Retry-After and reduce concurrency. If failures cluster on one exit, compare a small residential-proxy sample at the same request rate before choosing a rotation strategy.
CAPTCHA or access denial Error frequency by exit IP, request rate, cookies, and session changes Pause automatic retries and isolate the cause. For IP-related failures, evaluate residential exits; keep related requests in the same session.
Sign-in page or expired session Final URL, authentication state, and session cookies Refresh the supported authentication flow and keep session settings consistent.
Browser shows a field but Python returns nothing Whether the value exists in the actual input HTML or JSON Fix the selector only if the field is present. Otherwise use a source that exposes the field or leave it missing.
Price differs from the one you expected Marketplace, currency, delivery location, seller, variant and timestamp Compare matching records. Keep separate fields for current price, list price and shipping.
Review export repeats or stops early Next-page token or link, unique IDs and date range Stop if the page repeats. Record the verified coverage and check whether the source offers more.

Keep fetch failures separate from records with missing fields. If a listing is out of stock and no price appears, it can still be a usable availability record. An error page accidentally saved as a product is a different problem.

When errors cluster on a shared or datacenter exit, IP reputation and traffic concentrated on that address are worth investigating. Test a small ASIN batch through residential exits while keeping the marketplace, request rate, and other settings unchanged. Compare valid rows, 503/429 responses, and CAPTCHA frequency. This gives you a practical basis for choosing an exit network and rotation schedule. For connection timeouts, separately check the proxy host, port, and credentials.

Check the Response Before Retrying

For HTTP 503, inspect the response body and Retry-After header to distinguish a temporary service error from an access challenge. A status code alone does not identify the cause.

When a value appears in Chrome but is missing from Python output, open DevTools → Network, reload the page, select the main document request, and search the Response tab. If the value is present, update your selector. Otherwise, compare the browser and script’s marketplace, delivery context, and response content, then check whether a separate Fetch/XHR request supplies the value. Use the source’s documented endpoint or browser-rendering workflow to capture the field.

How to Scale Amazon Data Collection

Scale once a small sample has the right fields. Start with several ASINs that include a variation and an unavailable item. For product snapshots, keep marketplace, ASIN and observation time so a later price doesn’t overwrite the earlier one. For reviews, retain the source’s record ID when available. Track three numbers on each batch: how many records you expected, how many distinct records you received, and how many passed your field checks.

Compare sources by cost per 1,000 usable rows, not by request price alone. If a batch costs $5 and 800 records pass your field checks, the cost is $6.25 per 1,000 usable rows. Apply the same checks to each source so duplicates and unusable rows do not inflate the count.

Pilot a Daily Run

For a daily price check, start with five ASINs in one marketplace and schedule one run per day. Save each result as a new row keyed by ASIN, marketplace, offer type and observation time. Log missing offers separately from failed requests so an outage doesn’t look like a price drop or an out-of-stock item. At the end of the first week, compare expected rows with valid rows and inspect a few against the source before adding more items or increasing frequency.

When to Use Proxies for Amazon Scraping

For recurring Amazon collection, proxy selection affects three operational choices: the exit network used by each batch, the country requests come from, and whether related requests share an IP. Residential exits are worth testing when a shared or datacenter address repeatedly encounters IP-related challenges. Rotation lets independent requests use different exits; sticky sessions keep related requests together. Continue to respect source rate limits when scheduling batches.

Socks5.IO’s rotating residential proxies combine residential exits, country targeting, and rotating or sticky sessions for teams running their own scrapers. Use rotating sessions for independent product snapshots and sticky sessions for offer comparisons or review pagination. For localized price and stock checks, align the exit country with the Amazon marketplace and delivery ZIP code, then compare matching variants and sellers.

For multi-step collection, combine country targeting with a Sticky Session. In Socks5.IO Proxy Setup, select your target country under Country/Region and choose a session window, such as 10 minutes. Confirm the exit region, then check the IP twice with the same session settings. A sticky IP can change when its window expires, after inactivity, or if the node becomes unavailable. Start a small Socks5.IO pilot and compare cost per usable row, error rates, and session consistency before increasing the batch size.

Socks5.IO residential proxy setup with United States selected and a 10-minute sticky session

FAQs

Yes. Import the CSV as a new sheet. Keep ASINs and other identifiers as text, and leave the currency in its own column so prices from different markets aren’t summed together by accident.

Latest articles

Global proxy network illustration

Plug into the global proxy network — every connection faster and more stable

Start for Free