Sellers play a big part in flagging reviews that violate the guidelines. In fact, I think most of the culling of bad reviews comes from customers and sellers complaining and very little from automation as people here might imagine.

[1] - https://www.amazon.com/gp/help/customer/display.html?nodeId=...

So if you and your spouse like to keep their own opinions and review separately never connect your accounts in any way.

That of course, makes it a pointless statistic, and hopefully, not something, they seriously tell themselves matters.

I'm sure it sounds great for the press though!

IE you really want to know STH like "What percent of reviews are fake for the first two pages of items in the top 100/1000 categories on amazon"

or something that is weighted towards things people see/look at, not "amazon's entire catalog"

Also you have grey fakes with solicited reviews. I have a few vendors who offer significant discounts in exchange for reviews.

Plus these Chinese companies just rebrand and post new product listings free of the bad reviews, leaving the old ones to die.

Amazon obviously has a huge catalog. There's lots of niche products, where quite honestly, there just isn't enough money for fake reviews.

Not to mention some products like books, where most seem to be real.

The question Amazon isn't answering -- and what I think you're getting at -- is something like for top products in certain products like electronics -- what % are fake?

The answer to that I would wager is >50%.

1. Amazon uses good ML techniques, and they don't work.

2. Amazon is doing something else to detect fraud, and it doesn't work.

3. Amazon is lying about the issue.

A good friend of mine worked on the recommendation engine ML side for AMZN until fairly recently. His review was that it's worse, and more expensive, than having hired people do it by hand.

But the sibling comment is more correct -- all three of these things are currently the case.