โ Back to Blogs
Amazon Strategy
12 min read
Amazon Review Analysis: How to Mine Competitor Reviews With AI (Bright Data and Apify)

Onieque Edwards
Content Strategist /Blog Writer

A rival's one-star reviews are some of the cheapest product research on Amazon. Someone else paid for the inventory and the returns, and their buyers wrote down what went wrong in their own words. Most sellers read a dozen, nod, and move on.
Amazon review analysis means reading reviews in bulk for the reasons behind the stars, and competitor review mining is the same job pointed at the listings you are trying to beat. With AI review analysis, you pull those reviews into an assistant such as Claude and get back a short list of complaint themes, each with a count and real quotes.
This guide shows how to do that with the official MCP servers from Bright Data and Apify. An MCP server is a connector that lets an assistant run an outside tool for you: you ask in plain language, and the tool fetches the data.
It also covers what those tools cannot reach, where Amazon's terms stand, and how to tell a complaint worth spending on from one that is only loud.
In this article
Why Mine Competitor Reviews on Amazon?
Should You Start With Amazon's Own Review Tools?
How Many Competitor Reviews Can a Scraper Actually Get?
How to Pull Competitor Reviews With Bright Data or Apify
How to Run AI Review Analysis on Amazon Reviews
Is It Legal to Scrape Amazon Reviews?
Competitor Review Mining: A Worked Example
Which Complaint Themes Are Worth Spending On?
FAQ
Why Mine Competitor Reviews on Amazon?
Your own reviews tell you what your buyers think of your product. A competitor's reviews tell you what shoppers in your category wanted and did not get.
That second question is closer to the one that decides a launch or a relaunch. What would make a shopper pick you instead?
The value sits in the pattern, never in the single review. One buyer saying a lid leaked is an anecdote. A third of a rival's reviews saying it is a gap your listing or your product can answer.
Turning that gap into a real point of difference is its own job, covered in how to turn customer research into Amazon product differentiation. This article is about the pulling and the sorting.
Should You Start With Amazon's Own Review Tools?
Yes, for your own brand and your category. They carry less risk than any scraper, and they see data a scraper never will.
Customer Review Insights is a tab inside Opportunity Explorer in Seller Central. Amazon's seller forum announcement lists Review Trends (for brands, showing six-month sentiment shifts), Parent ASIN insights, ASIN Benchmarking against the category average, and Topic Drilldown. The announcement page shows only "2 years ago", so its launch date is not confirmed here. Availability also varies by account.
Voice of the Customer sits under the Performance menu. It shows your NCX rate (the share of orders with negative customer experience feedback), your NCX orders, the top return reason per listing and a CX Health rating.
Return reasons matter most for this job. They come from buyers who never wrote a review, so they catch complaints a review count misses. They cover your own listings only.
"Customers say" summaries on product pages can be read on any listing, a competitor's included. Seller Labs notes that sellers cannot edit them. Forum posts complain they can surface old negatives that were already fixed, so treat them as a hint.
What none of these give you is a competitor's actual review text, in volume, in a file you can sort. That is the gap a scraper fills.
How Many Competitor Reviews Can a Scraper Actually Get?
Fewer than most tool pages suggest.
Amazon puts the full reviews page behind a sign-in. Two scraping vendors report the wall is still in force in late 2026.
Scrape.do, checking in October 2026, found that a logged-out request for the reviews page returns the sign-in page instead. What remains is 13 featured reviews on the product page itself, 8 local and 5 from other countries. Thunderbit puts the logged-out figure at about 8.
When the wall went up is not settled. Thunderbit's guide says 14 November 2024 in one place and late 2025 in another, and no Amazon source was found. The fair wording is late 2024 or 2025, according to scraper vendors.
Tools built to reach further come with caps of their own. Apify's junglee Amazon Reviews Scraper documents a maximum of 100 reviews per star rating or variant, and mentions 500 per product as a best case. Another field on the same page says 100 per product is the most it can get, and it returns only reviews with text.
A second actor, khadinakbar's Amazon Reviews Scraper, lets you ask for 1 to 10 reviews. Its page says Amazon shows about 5 to 10 inline.
Two things follow. A logged-out pull gives you the reviews Amazon chose to feature, which is a sample Amazon picked, not a random one. And any tool promising "all reviews" needs testing on one product before you trust it, because the logged-in routes that reach further also carry more risk under Amazon's terms.
How to Pull Competitor Reviews With Bright Data or Apify
Both companies publish official MCP servers. Both return public page data only, with no sales figures, ad data or search volume.

Step 1: Check the first-party tools. Read Customer Review Insights and your own return reasons first, so you know what your buyers already say before you read anyone else's.
Step 2: Pick the competitor listings. Choose three to five ASINs a shopper would actually compare with yours, meaning the same use at a similar price. A rival at twice your price draws different complaints.
Step 3: Connect one MCP server.
Bright Data. The official server is brightdata/brightdata-mcp on GitHub, run hosted or on your own machine with a Bright Data API token. Its README does not name Amazon tools. Amazon data comes through its general search_scrapers and run_scraper flow, which reaches Bright Data's Amazon scrapers.
One of those is a Reviews scraper. It takes an Amazon product URL and returns the review text, rating, reviewer details, a verified-purchase flag and the date. Bright Data does not publish a review-count cap for it.
On cost, Bright Data's MCP pricing page lists a free tier of 5,000 requests a month. Published per-record figures for its scrapers run from $0.75 to $1.50 per 1,000 records depending on the source, so confirm with Bright Data before you budget.
Apify. The official server is apify/apify-mcp-server, hosted at mcp.apify.com and connected by OAuth or a token. Any actor in the Apify Store can become a tool, which is how you reach a reviews scraper such as the two named above.
Connecting is free. Each actor run bills to your Apify account at that actor's own price, and the free plan includes $5 of platform usage a month. Many Amazon actors are built by community developers rather than Apify, so quality and upkeep vary.
Step 4: Cap the spend before the first run. Apify supports maxTotalChargeUsd and maxItems limits on a run, so set both. Listed prices for reviews actors ran from $0.06 to $7.50 per 1,000 reviews when this was researched, and they are not like-for-like, because some add per-event charges, platform usage and start fees. Run one ASIN, check the bill, then scale.
Step 5: Check what came back. Count the reviews per ASIN and read the dates. If you asked for 100 and got 12, you hit the sign-in wall, not a quiet product. The junglee actor's changelog notes a January 2026 fix for errors caused by Amazon's anti-scraping changes, so expect a run to break now and then.
How to Run AI Review Analysis on Amazon Reviews
The sorting is where the hours are saved, and also where most of the mistakes get made.
Ask for themes in a fixed shape, with counts and quotes. A prompt along these lines works:
Group these reviews into complaint themes. For each theme, give the number of reviews that mention it, which ASINs they came from, the star ratings of those reviews, and two short quotes. Put praise in a separate list. Only merge themes when the reviews describe the same failure.
What you get back is a list of complaint themes you can check line by line against the file. Four rules keep that list honest.
Count reviews, not mentions. One angry review that says "leaks" four times is one review.
Split every theme by ASIN. A theme that comes from one listing is that product's flaw. A theme spread across all of them is a category gap, and that is the one worth building around.
Spot-check the quotes. Open five quoted reviews in the raw file and confirm the words are there. A theme you cannot trace to real text is not a theme.
Set a floor. On samples this small, a theme in 3 reviews is noise. Write it down and wait for it to grow.
Is It Legal to Scrape Amazon Reviews?
This is not legal advice, and nobody can give you a blanket yes.
Amazon's Conditions of Use, last updated 14 August 2026, exclude from the site licence "any use of data mining, robots, or similar data gathering and extraction tools." That is Amazon's position, in its own terms.
Two court cases get quoted as proof that scraping is fine. Neither is about Amazon.
In hiQ v. LinkedIn, the Ninth Circuit said in April 2022 that scraping public profiles raised serious questions about whether it broke the federal anti-hacking law. A district court later found hiQ had breached LinkedIn's user agreement. The case settled in December 2022 with a $500,000 judgment and a permanent bar on scraping. The contract claim won.
In Meta v. Bright Data, a federal judge in California ruled on 23 January 2024 that Meta's terms did not bar Bright Data from scraping public data while logged out. That ruling turned on Meta's terms, not Amazon's.
Amazon v. Perplexity is sometimes added to the list. The Ninth Circuit vacated Amazon's preliminary injunction in August 2026, but that case is about an AI browsing agent working inside password-protected areas, and it is not resolved.
As a seller you have also signed Amazon's Business Solutions Agreement. SellerSprite, a seller-tool vendor, reports that Amazon added a section on AI agents in 2026, with new limits on data mining. Amazon's own text was not checked for this article, so read your current agreement yourself.
The fair summary: Amazon's terms prohibit it, courts have treated public-data scraping differently from case to case, and the risk sits with the account holder. That is the reason the first-party tools come first. If you do run a scraper, logged-out reads of public pages are the lower-risk end, and logged-in scraping is the higher-risk end.
Competitor Review Mining: A Worked Example
The numbers in this example are invented to show the method. They do not come from real listings.
Say you sell an insulated water bottle. You pull reviews for three rival bottles at a similar price, logged out, and get 36 reviews back, 12 per listing, which fits the featured-review ceiling. The themes come back like this:
Lid leaks: 14 reviews, spread across all three listings
Hard to clean: 9 reviews, 7 of them on one listing
Too heavy: 4 reviews
Paint chips: 3 reviews
Now read each theme against the rest of the picture.
Lid leaks clears the floor and crosses all three rivals. Next, check what their listings promise. If two of the three say leakproof in the title or on the main image, this is a broken promise, which is exactly what produces one-star reviews.
Then check your own side. If your Voice of the Customer dashboard shows leaking among your top return reasons, the gap is shared across the category, and claiming it would be a promise you also break. If your return reasons are clean on it, you have something your listing can claim, with proof shown in an image.
Price and stars finish the read. If the leak complaints sit in the 1-star and 2-star reviews of the cheapest rival only, price may explain them. If they sit across every price point, the gap belongs to the category.
Hard to clean looks large at 9, but 7 of those come from one listing. That is one rival's design flaw. It is useful to know, and it is not a category gap.
Too heavy and Paint chips sit at or near the floor. Park them, pull again next quarter, and see whether they grow.

To take a theme like the lid leak one step further and score it against what fixing it would cost, use the method in which Amazon customer complaint is worth fixing. It turns a counted theme into a yes or no on the spend.
Which Complaint Themes Are Worth Spending On?
The count is never the finish line. A complaint theme is a claim about what buyers want, and it only becomes a finding when other signals agree with it.
Read every theme against the star ratings of the reviews carrying it, the price gap between you and the rival, what the rival's listing promises, and, for your own product, the return reasons in Voice of the Customer. One flag is noise. Correlated flags are a finding.
Rivals also fix things. A theme that was loud in the spring can vanish by autumn once a competitor changes its lid. Re-pull on a schedule, the same way you would monitor Amazon competitors for price and rank.
The themes that survive route to two very different budgets. A theme your product already solves is a listing and image decision, and it is cheap. A theme your product shares with the category is a product decision, and it costs a supplier conversation and a new production run.
So the decision is where the next dollar goes. Put it into words and images that claim a fix you already have, or into a product change for a gap the whole category shares. Only commit to the second when the theme crosses several listings, clears the floor, and shows up in your own return reasons too.
Holding a list of competitor complaints and unsure which one is costing you sales, or whether your own listing carries the same gap? The Expedition Audit sets your review themes against your return reasons and your listing, and names the one worth paying to fix. Book it here.
FAQ
What is Amazon review analysis?
It is reading reviews in bulk to find the reasons behind the star ratings, usually grouped into complaint and praise themes with a count for each. Run on competitors' listings, it shows what shoppers in your category wanted and did not get.
Can I get all of a competitor's reviews with a scraper?
Usually not. Amazon puts the full reviews page behind a sign-in, and scraper vendors report that a logged-out read returns roughly 8 to 13 featured reviews per product. Some Apify actors document caps of 100 per product, with 500 as a stated best case. Test any "all reviews" claim on one product first.
Should I use Bright Data or Apify for Amazon reviews?
Neither is ranked here. Bright Data runs its own Amazon Reviews scraper, which takes a product URL. Apify connects you to reviews actors in its Store, many of them built by community developers, each with its own caps and price. Published prices vary and conflict, so confirm costs with each company and set a spend cap before the first run.
Is it legal to scrape Amazon reviews?
Amazon's Conditions of Use exclude data mining and robots from the site licence. Court cases often quoted in favour of scraping, hiQ v. LinkedIn and Meta v. Bright Data, were about other companies' terms, and hiQ lost on contract. The risk sits with the account holder, and this is not legal advice.
Where do I find Customer Review Insights on Amazon?
It is a tab inside Opportunity Explorer in Seller Central, with Review Trends, Parent ASIN insights, ASIN Benchmarking and Topic Drilldown. Availability varies by account.
How many reviews make a complaint theme worth acting on?
There is no official number. On the small samples a logged-out pull returns, treat a theme in 3 reviews as noise. Act when a theme crosses several competitor listings and agrees with your own return reasons, star ratings and the promises rival listings make.
Sources
Amazon (official): Conditions of Use, last updated 14 August 2026, an Amazon-hosted page, for the data mining exclusion. Amazon Seller Forums announcement of Customer Review Insights, page dated "2 years ago" only, for the tab and its four views.
Tool makers (each describing its own product): Bright Data's official MCP server on GitHub, its MCP pricing page (published 3 December 2025, modified 1 September 2026) and its Amazon scrapers documentation. Apify's official MCP server on GitHub. Apify Store pages for the junglee Amazon Reviews Scraper and its changelog, and the khadinakbar Amazon Reviews Scraper, read 10 November 2026.
Scraping vendors, for the sign-in wall: Scrape.do, updated 29 September 2026, checked October 2026. Thunderbit, updated 21 August 2026.
Seller-tool vendors: Seller Labs on Amazon's AI review highlights. SellerSprite on the Business Solutions Agreement agent policy, a secondhand account.
Legal coverage: Proskauer on the hiQ and LinkedIn settlement, 8 December 2022. Farella Braun + Martel on Meta v. Bright Data. Engadget on the Amazon v. Perplexity appeal.

Onieque Edwards
Content Strategist /Blog Writer
Onieque is the brain behind bold Amazon growth strategies and structured business execution. He enjoys turning scattered ideas into clear, actionable systems that actually drive results. When heโs not building out growth plans or refining campaigns, youโll likely find him exploring new coffee spots or getting lost in ideas that connect strategy with creativity.
