How to Scrape Amazon Reviews and Put Them to Work

To scrape Amazon reviews at scale, you give a tool a list of products — ASINs or product URLs — and get back one row per review: the review text, the star rating, the date, the reviewer's display name and the product it belongs to. Amazon publishes no free reviews API, so a scraper is the practical way to get past the first page of feedback and into the full history. The Amazon Reviews Scraper does exactly this across fifteen Amazon marketplaces and exports the result as CSV or Excel.

Collecting the reviews is the easy half. The value is in what you do with them: reading a competitor's one-star reviews as a corpus, lifting the words customers use into your listing copy, running sentiment analysis over thousands of reviews instead of the twenty you can read by hand, and watching how fast a product is gathering feedback. This guide covers why manual approaches break down, exactly what an Amazon review row contains, how to run a job, what it costs, and the analyses that make the data worth paying for.

Why you need to scrape Amazon reviews instead of reading them

Reading reviews on the product page works for one product and a casual question. It stops working the moment the question gets serious. A popular product can carry thousands of reviews, the page shows them a handful at a time, and the order Amazon picks is not the order your analysis needs. Copy-pasting is slow, error-prone and loses structure: by the time you have pasted 300 reviews into a document, the rating, date and variant for each one are gone.

Amazon's own tools do not fill the gap. There is no free public reviews API — review content lives in the product page rather than in a feed you can call. Seller-side reporting is about your own catalogue, not your competitors', and an average star rating tells you almost nothing on its own. A product whose rating moved from 4.3 to 4.2 has a story behind it; the average will never tell you what it is.

That leaves two options: maintain your own parser, or use an Amazon reviews scraper that someone else keeps working. Building your own is a real project — pagination, fifteen storefronts with different domains, layout changes, retries — and it never stays finished. For most teams the question is simply how quickly they can get a clean file of reviews into a spreadsheet or a model.

What Amazon review data comes back: every column

Each row is one review, as Amazon displays it, tagged with the product it was left on — so an export stands alone without a join back to the listing, and a run across many ASINs still reconciles against your input list. These are the nineteen columns that fill, with the header names exactly as the export writes them:

Quirks to handle before you analyse

• The spellings are real. The export spells author as autor across four columns and badge as bage. Write your code against the actual header, not the correct English, or your column lookups will fail silently.
• Dates need parsing. date is a long-form string like "January 14, 2026", not ISO 8601. Convert it once, on import, before any time-series work.
• Blank is not unknown. An empty helpful means nobody marked the review helpful, not that the value is missing. variation, autor_url, url and img_url are present on a share of rows only.
• Some columns never fill. comments, official_comment_banner and autor_descriptor appear in the header but have come back empty; ignore them.
• Text is untouched. body comes back verbatim, in whatever language the shopper wrote it, with no preprocessing — so your sentiment analysis reads the customer's wording rather than ours.

For the definitive header row on your own products and marketplaces, run a small job first and build your import against it.

How to scrape Amazon reviews: step by step

1. Build the product list. Collect the ASINs or product URLs you care about — your own catalogue, the top competitors in a category, or both. Either input works; use whichever your list already holds. Remember that one input row produces many output rows.
2. Open the Amazon Reviews Scraper and paste in the ASINs or product URLs.
3. Decide which marketplaces to cover. Fifteen storefronts are supported: amazon.com, .co.uk, .ca, .de, .es, .fr, .it, .in, .nl, .se, .com.mx, .com.br, .co.jp, .com.au and .ae. The same ASIN is reviewed differently in different countries, so pull several if the regional split matters.
4. Select sorting and filtering options. Narrow the pull up front rather than exporting everything and discarding most of it — this is where you control cost.
5. Choose your output format — CSV or XLSX — and read the pre-flight estimate, which shows the review count and credit cost before the job starts.
6. Run it and import the file, applying the date parsing and column-name handling above. If you would rather skip the download, the same job can be called through the API.

What it costs to export Amazon reviews

Amazon reviews are priced differently from the rest of the platform. It is $0.02 per review — $20 per 1,000 reviews — pay-as-you-go with no subscription and no minimum. The service is available once you have bought credits; the free sign-up credits that new accounts get do not apply to Amazon reviews. The estimator shows the cost before anything runs, so there is no compute-unit arithmetic to do.

That price makes the sorting and filtering step worth taking seriously. Say you want to study ten competing products that carry 2,000 reviews each. Pulling everything is 20,000 reviews, or $400. If the question is "what do unhappy buyers complain about", narrowing the run to the reviews that answer it — and pulling a sample from the rest — might get you the same insight for a fraction of that. Volume pricing for category-scale monitoring or large historical pulls is on request; see the pricing page.

Use case: product research before you launch

The cheapest time to fix a product is before it exists. Pull the reviews for the five to ten best-selling products in the niche you are entering and read the low-star text as one corpus. Recurring failure modes show up as repeated language: "stopped charging after a month", "lid doesn't seal", "smaller than the photos". Each of those is a specification requirement your version can meet.

The high-star reviews matter as much. They tell you what buyers consider table stakes and what genuinely delights them. Use variation to check whether complaints concentrate in one size or colour — a problem with only the large variant is a sourcing issue, not a design flaw — and use helpful to weight reviews other shoppers found persuasive. Our guide to web scraping for market research covers how to frame questions like these before you collect anything.

Use case: mining competitor weaknesses

Feed in a rival's ASINs and read what their buyers dislike, in the buyers' own words. This is the sharpest positioning input you will get, because it is unprompted: nobody asked these customers to compare products, they just described what went wrong.

A practical method:

• Filter to rating of 1 and 2 and group the body text by theme — durability, sizing, instructions, customer service, value.
• Count themes per competitor. A complaint that appears across every rival is a category problem; one that appears for only one rival is their weakness, not the market's.
• Check date once parsed. A complaint that clusters in the last three months may point to a manufacturing change; one spread across years is structural.
• Pull several marketplaces for the same ASIN. A product loved in one country and returned in another is telling you something about expectations, sizing or translation.

The same approach works for local businesses using Google reviews — our post on using Google Maps reviews for competitor analysis applies it to storefronts rather than products.

Use case: voice-of-customer for your listing copy

The best listing copy uses the words buyers already use. Reviews are where those words live. If dozens of five-star reviewers say a backpack "fits under an airline seat", that phrase belongs in your bullet points, because it is what the next buyer is searching for and worrying about.

Pull the reviews for your own ASINs and your closest competitors, filter to the positive ones, and extract recurring phrases from title and body. Review headlines are especially useful: they are short, written by customers, and usually state the single thing the buyer cared about most. Then do the reverse with negative reviews and answer the objections directly in your copy — if competitors' buyers complain about unclear sizing, a precise size chart is a selling point.

Use case: sentiment and AI analysis of Amazon review data

With verbatim text and ratings in one file, you can run your own model or an LLM over a real sample instead of a handful of reviews read manually. Export the top ASINs in a category and ask questions an average rating cannot answer: which product attributes drive five-star reviews, which complaints predict a one-star, how sentiment differs between verified and unverified buyers using the bage column.

The methods are the same ones we use for Google reviews. How AI can analyze thousands of Google reviews walks through prompting, batching and theme extraction, and customer sentiment analysis using Google reviews covers scoring and how to read the results. Two adjustments for Amazon data: translate downstream if you pull several marketplaces and want one language, and keep product_asin on every record you send to a model so results stay attributable to a product.

Use case: tracking review velocity

Review velocity — how many reviews a product gathers per week or month — is one of the clearer public signals of how well it is selling and how that is changing. Once date is parsed, count reviews per product per month and you have a trend line for every ASIN in your set.

Run the same product list again later and compare. total_reviews and overall_rating are repeated on every row, so each export is also a snapshot of the product's headline numbers at that moment. A competitor whose monthly review count doubled after a price change, or a new entrant whose velocity is overtaking an established product, is the kind of shift you want to see in a sheet rather than discover in your sales numbers. De-duplicate across runs on the review id so a repeat pull adds only what is new to your history.

Amazon's terms and responsible use

This is not legal advice; if the stakes are material, take it to a lawyer.

Reviews are published to be read. Amazon shows the reviewer's display name, the text, the rating and the date to anyone visiting the page, signed in or not, and collecting publicly visible feedback for product and market research is long-established practice. Livescraper collects public review content only, touches nothing behind a login, and auto-deletes exports after 30 days.

Two caveats are worth stating plainly. First, Amazon's terms of service restrict automated access, so this is a terms question as well as a legal one — if you have a contractual relationship with Amazon, such as a seller account, check what it says. Second, reviews carry a reviewer's display name, which makes the output partly personal data under GDPR and similar laws. Analysing sentiment across a product or category is a defensible purpose; building a picture of an individual shopper, or contacting reviewers, is not. Drop autor_name and autor_url from your working dataset if your analysis does not need them.

Conclusion

When you scrape Amazon reviews, you turn a page of feedback that is hard to read at scale into a dataset you can sort, count and model. The workflow is short: list the ASINs or product URLs, pick the marketplaces, narrow the run with sorting and filtering, export to CSV or Excel, and handle the export's real column names and long-form dates on import. From there the same file supports pre-launch product research, competitor weakness mining, customer-language listing copy, AI sentiment analysis and review-velocity tracking. The Amazon Reviews Scraper costs $20 per 1,000 reviews once you have bought credits, with the exact cost shown before every run.

Related reading: How to Export 10,000 Google Reviews for AI Analysis, How to Use Google Reviews for Local Market Research, How to Build an AI-Ready Business Database Using Public Business Data.