Amazon product pages can sometimes appear to change from one day to the next. A product may show 4.6 stars and 12,453 ratings on Monday, then 4.5 stars and 12,421 ratings later in the week—even when the product itself has not changed.
This can be confusing for shoppers, sellers, brands, ecommerce teams, and anyone collecting Amazon product data.
The important thing to understand is that Amazon’s displayed rating and review information is not necessarily a permanent historical record. Amazon continuously evaluates customer feedback, applies policies and automated detection systems, and can change how feedback is associated with products.
This article explains why Amazon ratings and review counts change, what the different numbers actually mean, and what businesses should consider when tracking Amazon review data.
1. Ratings and Reviews Are Not the Same Thing
One of the biggest sources of confusion is the difference between a rating and a written review.
A customer can provide a star rating without necessarily writing a substantial review. Amazon therefore distinguishes between the number of ratings associated with a product and the written customer reviews that appear in the Reviews section.
For example, a product might display:
4.6 out of 5 stars — 12,453 ratings
The 12,453 figure represents the number of customer ratings used in the displayed rating system. It should not automatically be interpreted as meaning that Amazon has exactly 12,453 publicly visible written reviews.
This distinction is particularly important when collecting product data programmatically.
If a scraper records:
4.6 | 12,453
it should generally interpret those values as the displayed star rating and rating count, rather than assuming the second number represents the number of written review articles.
2. Amazon Can Remove Customer Reviews and Ratings
One of the most obvious reasons a product’s count can decrease is that Amazon removes customer feedback.
Amazon says customer reviews can disappear when they violate its Community Guidelines, when the customer removes the review, when products were incorrectly associated with one another, or when Amazon detects unusual review behavior.
Amazon’s automated systems can also limit which ratings or reviews are displayed for a product when unusual activity is detected. In some circumstances, Amazon says it may display only reviews associated with Verified Purchases.
This means that a decrease in the displayed count does not necessarily mean customers went back and deleted their reviews.
For example:
| Date | Displayed ratings |
| Monday | 12,453 |
| Tuesday | 12,448 |
| Wednesday | 12,441 |
It would be incorrect to automatically conclude that 12 customers deleted their ratings.
The decrease could result from Amazon removing feedback, filtering activity, or changing the association between products and their reviews.
3. Amazon Uses Automated Systems to Detect Unusual Review Activity
Amazon has extensive systems designed to protect the integrity of its customer-review ecosystem.
Amazon representatives have stated that its review system uses automated filters to help identify potentially problematic review activity. Amazon does not publicly disclose all of the criteria used by these systems.
This is important because a review does not necessarily have to look obviously fraudulent to a seller or shopper for Amazon’s system to restrict or remove it.
Amazon may detect patterns or relationships that are not visible on the product page.
For example, Amazon could identify unusual activity associated with:
- A group of reviews arriving in an unusual pattern
- Promotional or incentivized review activity
- Review manipulation
- Multiple accounts exhibiting suspicious behavior
- Other activity that conflicts with Amazon’s review policies
The result can be a change in the number of ratings or reviews displayed on the product page.
An important distinction
A review being removed does not necessarily mean Amazon has determined that the reviewer is a fraudulent customer.
It means Amazon’s systems have determined that the feedback should not remain displayed under its policies or detection criteria.
Amazon generally does not disclose the precise reasons behind individual automated decisions.
4. A Verified Purchase Does Not Make a Review Permanently Safe
Another common misconception is that a review marked Verified Purchase cannot be removed.
Verified Purchase is useful because it indicates that Amazon has identified a qualifying purchase relationship. However, Amazon’s review policies still apply to those reviews.
Amazon states that reviews violating its Community Guidelines can be removed, and its review systems can also respond to unusual review behavior.
Therefore:
Verified Purchase does not mean “permanent review.”
A verified review can still disappear if it subsequently falls under Amazon’s removal or filtering rules.
5. Customers Can Delete Their Own Reviews
The simplest explanation for a disappearing review is sometimes the correct one: the customer removed it.
Amazon explicitly identifies customer deletion as one of the reasons a review can disappear.
Consider a product with 10,000 ratings.
If several customers delete their feedback, the displayed count could decline.
However, when monitoring large product catalogs, it is usually impossible to determine from the product page alone whether a decrease was caused by:
- Customer deletion
- Amazon moderation
- Unusual-activity filtering
- Product-listing changes
- Another change in Amazon’s aggregation system
This is one reason historical tracking is important.
6. Amazon Can Separate Reviews When Products Were Incorrectly Combined
Another particularly important reason for sudden changes is product association.
Amazon explains that multiple products can sometimes be incorrectly listed as the same product. When Amazon separates those products, reviews associated with them can also be separated.
This can produce a dramatic change.
Imagine:
Product A
- 8,000 ratings
Product B
- 5,000 ratings
If Amazon previously associated the two products together, a customer might have seen:
13,000 ratings
If Amazon subsequently determines that the products should be separated, the counts could become:
Product A: 8,000
Product B: 5,000
Nothing necessarily happened to the customers’ reviews themselves. Amazon simply changed which product those reviews belong to.
This is particularly relevant to products with many variations.
7. Product Variations Can Complicate Review Counts
Amazon product catalogs frequently contain variations.
Examples include:
- Black vs. white
- 128 GB vs. 256 GB
- Small vs. large
- Different colors
- Different bundles
- Different configurations
Depending on how Amazon associates products and variations, the reviews displayed to a shopper can change.
A product’s apparent review history can therefore be affected by changes to the underlying catalog structure.
For ecommerce analytics, it is important to distinguish between:
ASIN-level data
and
Parent/variation-level data.
A change in the displayed review count does not necessarily indicate that customers suddenly changed their opinion of the product.
It may indicate that Amazon changed how the products are grouped.
8. Amazon Can Change the Displayed Star Rating Without a Simple Average
Another important point is that Amazon’s displayed star rating should not necessarily be calculated using a basic arithmetic average.
Amazon states that it uses machine-learned models rather than a simple average to calculate a product’s star rating.
This means a calculation such as:
(5-star ratings + 4-star ratings + 3-star ratings + 2-star ratings + 1-star ratings) ÷ total ratings
may not reproduce the star rating displayed on Amazon.
This is a critical distinction for anyone building an Amazon data-analysis system.
Example
Suppose a product has:
- 800 five-star ratings
- 100 four-star ratings
- 50 three-star ratings
- 30 two-star ratings
- 20 one-star ratings
A simple mathematical average would produce one number.
Amazon’s displayed rating may not exactly match that number because Amazon’s rating calculation uses additional modeling rather than simply averaging all ratings.
Consequently, you should generally treat the displayed rating as its own data point.
9. Why the Star Rating Can Change Even When the Count Barely Changes
Consider:
Yesterday
4.6 stars — 10,000 ratings
Today
4.5 stars — 10,020 ratings
Only 20 additional ratings were added, but the displayed rating changed.
That does not necessarily mean the new 20 ratings were overwhelmingly negative.
Several things can influence the displayed rating, including Amazon’s rating methodology and the underlying distribution and treatment of ratings.
This is another reason why attempting to reverse-engineer Amazon’s displayed star rating using only the visible count is unreliable.
10. Why the Rating Count Can Go Down
A decrease is particularly interesting because people generally assume review counts can only increase.
In reality, a displayed count can decrease.
Common explanations include:
A. Amazon removed policy-violating feedback
Amazon may remove reviews that violate its Community Guidelines.
B. The customer removed the review
Customers can remove their own reviews.
C. Amazon detected unusual review activity
Amazon can restrict the display of reviews and ratings when it detects unusual activity.
D. Product associations changed
If products were incorrectly combined, Amazon can separate their reviews when it separates the products.
E. Review eligibility or filtering changed
Amazon can change which feedback is accepted or displayed for a particular item, including situations where only Verified Purchase reviews are displayed.
11. Why the Count Can Increase in Large Jumps
The opposite can happen as well.
A product might go from:
5,200 ratings
to:
5,500 ratings
without anyone observing 300 individual review events.
Possible explanations include changes to product associations, aggregation, or previously unavailable feedback becoming reflected in the displayed product data.
For this reason, a data collection system should not assume:
Today’s count − yesterday’s count = number of new reviews
That assumption can produce inaccurate analytics.
12. Amazon’s Review System Is Dynamic
The most important concept is this:
Amazon’s product-review data is dynamic rather than immutable.
A product page represents Amazon’s current view of the customer feedback associated with that product.
That view can change because of:
- New customer ratings
- New written reviews
- Customer deletions
- Amazon moderation
- Automated review filtering
- Unusual-activity detection
- Product catalog changes
- Variation changes
- Review/product association changes
- Changes in Amazon’s rating methodology
Therefore, the value displayed today should be considered a snapshot.
13. Why This Matters for Amazon Product Scraping
This is particularly important if you are collecting Amazon data automatically.
Suppose a scraper collects:
ASIN: B012345678
Rating: 4.6
Rating Count: 12,453
on August 1.
On August 15 it collects:
ASIN: B012345678
Rating: 4.5
Rating Count: 12,421
A naive analytics system might conclude:
The product received 32 fewer ratings.
That is not necessarily what happened.
A better interpretation is:
Amazon’s currently displayed rating count decreased by 32 between the two observations.
That distinction is extremely important.
Recommended data model
Instead of storing only the current value, store historical snapshots:
| Date | ASIN | Rating | Rating Count |
| Aug. 1 | B012345678 | 4.6 | 12,453 |
| Aug. 2 | B012345678 | 4.6 | 12,460 |
| Aug. 3 | B012345678 | 4.6 | 12,467 |
| Aug. 10 | B012345678 | 4.5 | 12,430 |
| Aug. 15 | B012345678 | 4.5 | 12,421 |
This allows you to identify changes in the displayed data without incorrectly claiming to know exactly what caused each change.
14. Do Not Treat Rating-Count Changes as Exact Review Velocity
Suppose:
Monday: 10,000 ratings
Tuesday: 10,030 ratings
It is tempting to report:
“The product received 30 new reviews.”
A more accurate statement would be:
“The displayed Amazon rating count increased by 30.”
The first statement assumes that every difference represents a newly submitted rating.
The second statement reports exactly what your data actually establishes.
This distinction becomes especially important when creating:
- Competitive intelligence dashboards
- Amazon monitoring systems
- Product analytics
- Executive reports
- Brand performance reports
- SEO/ecommerce reports
- Automated alerts
15. A Better Way to Measure Review Growth
If you’re monitoring products, track several metrics separately.
Rating count
The current number displayed by Amazon.
Written review count
The number of written reviews your data source can identify.
Star rating
The displayed Amazon rating.
Rating-count change
Current rating count − Previous rating count
Rating growth rate
(Current count − Previous count) / Previous count × 100
Observation date
The exact date and time the information was collected.
ASIN
The product identifier associated with the observation.
This gives you a historical record without assuming that every count change represents a new customer review.
16. What a Negative Review-Count Change Means
Suppose your monitoring system produces:
Previous count: 25,431
Current count: 25,397
Change: -34
You should report:
Rating count decreased by 34.
You should not automatically report:
Amazon deleted 34 reviews.
The latter requires information you generally cannot obtain from the public product page.
The decrease could potentially involve multiple factors, including customer deletions, Amazon moderation, unusual-activity filtering, or product association changes.
Amazon itself states that it does not always provide the specific reason for individual automated review decisions.
17. What a Positive Change Means
Likewise:
Previous count: 25,397
Current count: 25,431
Change: +34
means:
The displayed rating count increased by 34.
It does not necessarily prove that exactly 34 new written reviews were published.
The count could represent new ratings, changes in displayed/associated feedback, or other changes in Amazon’s review system.
18. Why Historical Screenshots Can Be Valuable
For important products, retaining screenshots or raw HTML snapshots can be useful.
For example:
August 1
4.6 out of 5 stars — 12,453 ratings
August 10
4.5 out of 5 stars — 12,421 ratings
A historical snapshot gives you evidence of what Amazon displayed at a particular point in time.
This can be especially useful when investigating:
- Sudden rating changes
- Review-count drops
- Product merges
- Product splits
- Variation changes
- Competitor activity
- Marketplace changes
- Data-quality issues
19. Why Amazon Review Data Should Be Treated as a Snapshot
A useful mental model is:
Amazon product page
↓
Current Amazon review/rating state
↓
Your scraper
↓
Your historical database
Your scraper is not necessarily retrieving an immutable database of every review ever submitted.
It is observing what Amazon is currently making available on the product page.
That means your database should preserve historical observations rather than continuously overwriting the previous value.
20. Recommended Approach for Ecommerce Teams
If you’re using Amazon review data for competitive or product analysis, I recommend capturing at least:
| Field | Purpose |
| ASIN | Identify the product |
| Product title | Identify the product in reports |
| Brand | Competitive analysis |
| Star rating | Current customer sentiment |
| Rating count | Size of customer-rating base |
| Written review count | Review-content volume |
| Product URL | Source reference |
| Marketplace | Amazon.com, Amazon.co.uk, etc. |
| Collection timestamp | Establish historical state |
| Variation | Identify variation-level changes |
| Seller | Context for marketplace changes |
Then calculate changes separately.
For example:
Rating Count Change
= Current Rating Count
– Previous Rating Count
and:
Rating Growth %
= Rating Count Change
/ Previous Rating Count × 100
This approach is much safer than assuming that every difference represents newly submitted reviews.
Frequently Asked Questions
Can Amazon remove legitimate reviews?
Yes. Amazon’s review system can remove or stop displaying feedback for policy reasons, customer deletion, product-association issues, or unusual review activity. Amazon’s automated systems may also restrict the display of ratings or reviews when unusual activity is detected.
Can Amazon remove Verified Purchase reviews?
Yes. Verified Purchase status does not override Amazon’s Community Guidelines or other review controls.
Can a review disappear without the seller doing anything?
Yes. Amazon identifies several circumstances in which reviews can disappear, including customer deletion, policy violations, product-association corrections, and unusual review activity.
Does a decrease in rating count mean Amazon deleted reviews?
Not necessarily.
It means the displayed rating count decreased. The exact cause may not be publicly identifiable.
Does an increase in rating count mean that many new reviews were posted?
Not necessarily. It is safer to describe it as an increase in the displayed rating count.
Is Amazon’s star rating a simple average?
No. Amazon states that it uses machine-learned models rather than a simple arithmetic average to calculate a product’s star rating.
Why did my competitor’s rating count suddenly drop?
There is no reliable way to determine the exact reason from the public product page alone. Possible explanations include customer deletions, Amazon moderation, unusual-activity filtering, or changes to how products and reviews are associated.
Should Amazon review counts be used for competitive analysis?
Yes—but as observational data, not as a perfect record of review activity.
The best practice is to collect the data regularly and maintain a historical record.
Conclusion
Amazon’s rating and review system is considerably more dynamic than it may appear.
A product’s displayed rating and rating count can change because customers submit new ratings, customers remove feedback, Amazon removes feedback that violates its policies, automated systems detect unusual activity, or Amazon changes the relationship between products and their reviews.
Amazon also states that its star-rating calculation uses machine-learned models rather than a simple average, meaning the displayed star rating cannot reliably be recreated by simply averaging visible star counts.
The key takeaway for shoppers, sellers, brands, and ecommerce analytics teams is:
A change in Amazon’s displayed rating or review count is evidence of a change in what Amazon is currently displaying—not necessarily evidence of exactly how many new reviews were submitted or deleted.
For anyone monitoring Amazon products programmatically, the safest strategy is to capture periodic snapshots, retain historical values, and report changes in the displayed rating/count without assuming the underlying cause unless it can be independently verified.
This approach produces much more reliable Amazon competitive intelligence and avoids one of the most common mistakes in review analytics: treating a dynamic marketplace display as though it were a permanent transaction ledger.
Sources
Amazon’s published customer-review documentation explains its review moderation process, missing-review scenarios, and machine-learned rating calculation.
Amazon Seller Central’s customer-review guidance provides additional explanations for disappearing reviews, including policy violations, customer deletion, incorrect product associations, and unusual review activity.
Amazon’s seller documentation also confirms that sellers cannot directly change or remove customer reviews and that reviews violating Amazon’s policies can be reported for removal.