How are star ratings distributed?

Each rating is drawn from a weighted distribution. Balanced spreads ratings across all five stars, mostly positive concentrates them at four and five stars, and mostly negative concentrates them at one and two stars. The live average reflects the distribution you chose.

Does the review text match the rating?

Yes. High ratings get positive titles and bodies, three-star ratings get neutral mixed text, and low ratings get negative complaints. So a one-star review never reads like a glowing recommendation.

What fields does each review include?

Every review has an id, an author name, a star rating from one to five, a title, body text, a verified-purchase boolean, a helpful-votes count and a date. This matches the fields most product-review components display.

Are the helpful-vote counts realistic?

They are skewed so most reviews have few helpful votes and a few have many, produced by multiplying two random values. This mirrors real review sites where a handful of reviews accumulate most of the engagement.

Does anything get uploaded?

No. Reviews are generated locally in your browser from a seeded random source. Nothing is sent to a server, and identical settings reproduce identical reviews until you regenerate.

E-commerce Product Review Generator

Believable reviews for shop and product pages

The E-commerce Product Review Generator produces fake customer reviews with star ratings, matching text, verified-purchase flags and helpful-vote counts. It is built for e-commerce demos and review components so star bars, sort controls and review cards have realistic data to render before a real catalog exists.

How it works

You set a review count and a rating distribution. Each rating is drawn from a weighted table: balanced spreads weight across all five stars, mostly positive piles weight onto four and five, and mostly negative onto one and two. A small helper performs the weighted pick by walking the weight array against a random threshold. Crucially, the review title and body are chosen by rating tier — four-plus stars draw from positive copy, three stars from neutral copy, and one-or-two stars from negative copy — so the sentiment of the text always matches the score.

The remaining fields are filled per review: an author name, a verified-purchase flag set roughly three-quarters of the time, a helpful-vote count produced by multiplying two random values so most reviews score low and a few score high, and a date within the past year of a fixed base. A seeded random source keeps a configuration stable between renders, and the tool shows the live average rating so you can confirm the distribution.

Tips and example

Choose mostly positive to mock a well-reviewed bestseller, or mostly negative to test how your UI surfaces a poorly rated product.
Generate a large count to stress-test a paginated reviews list and a star-distribution summary bar.
The verifiedPurchase flag lets you build and test a “verified buyers only” filter directly against the sample data.
Because ratings and text are aligned, screenshots taken from the demo data read naturally and won’t show a five-star review with a complaint.