Most people know that Amazon collects data about your shopping history to serve up personalized product recommendations. But what some of us didn’t realize is that you can check exactly what Amazon knows — or thinks it knows — about you in your settings menu. One user’s findings went viral on Threads after she discovered that Amazon assumed she “has flat buttocks” based on her purchases, resulting in a flood of other posts with users sharing the oddly specific information Amazon uses to describe them.
I thought this was worth sharing after I looked at mine, and saw that one of its inferences was:
Gradually built comprehensive gaming setup
This is more or less correct. "Gradually" is load bearing in that description and likely captures a lot about my behavior as a consumer. I was taken aback by how specific it was able to get and the fact that this is possible at scale.
Hahaha, Jesus this is so incredibly inaccurate for me. To be fair, I stopped regularly buying things from Amazon quite a while ago. The only thing I bought from Amazon in 2026 was a soymilk maker which I bought out of points from my old Amazon card, which I am keeping open almost entirely to avoid damaging my credit score. Most of the inferences it had for me were some variation of "probably has a 3D printer" or "probably owns a GM vehicle" - given how many windshield wipers and filaments I've bought, I wonder how they could have ever figured that out?
The irony is that the greatest number of inferences it has for me is stuff that I have bought for other people. It has a huge number of them about how I enjoy mixology and drinks. I don't, I'm not a drinker, and I've never made a mixed drink in my life outside of a mixology class on a vacation. All of the time I spent looking up those products and buying them was because it was a very elaborate gift for my husband. There's a lot of other inferences for other kinds of gifts I've bought for other people. I guess I'm just uniquely altruistic?
That's just the reality of data science at this scale and with so few data points. People like to think that data science and machine learning is literally wizardry, but it's mostly propped up by inaccurate and misleading reporting.
The famous "Target knew daughter was pregnant before parents did" is just fake, for instance.
The reality is that it's INCREDIBLY hard to make ANY inferences about someone from their Amazon purchases. That data is inherently sparse and noisy. If you DID manage to make a lot of inferences from it, it would look very suspicious to any data scientist - that looks more like over-fitting than anything.
It's always funny when people claim that Tinder or Hinge's goal is to selectively match people such that they go on dates but do not become a couple so they spend more time on the app. To anyone that's worked with data, this is the equivalent of telling someone to throw a baseball into a trashcan that's 10 miles away from you. As if, from the shit people put on dating profiles, and swipe information you're going to thread that needle.
Yeah, but here's the thing: I'm an outlier. It's weird that I'm not buying things on Amazon. It's weird that I don't have a Meta account for instagram or whatsapp.
But people who are invested in Amazon are probably not only buying things on amazon.com, they probably have Amazon Prime and are buying compulsively. They are probably buying groceries at Whole Foods Market. They're probably watching shows on Prime Video. They might be reading books on their Kindle or listening to their audiobooks on Audible.
There are many places they can bring datapoints in so I think that a lot of people are going to get weirdly specific predictions.
It's true that a lot of these inferences are either obvious or slightly off-target. Sure I've bought a lot of bike-related gear over the years, so obviously I'm a frequent cyclist. Isn't that obvious? However, I kind of wonder if the general public underestimates the usefulness of these seemingly shallow insights.
The classic one I've seen on the internet is someone buying a toilet seat and getting a bunch of ads for toilet seats afterwards, as if amazon is stupid for doing that. The way I see it, any average person is highly unlikely to purchase a toilet seat at any given time. Someone who's recently purchased a toilet seat has a vastly higher chance of purchasing another toilet seat, since they might return the first one or have another bathroom to work on later.
I don't have any evidence to support it, but my guess is that amazon makes a ton of sales due to their profile-based recommendations. I guess in my case the data is not that sparse. I've purchased about a thousand things on there with purchase history going back to 2002. That's a lot of data.
I don't have any evidence to support it, but my guess is that amazon makes a ton of sales due to their profile-based recommendations.
They absolutely do. It's just that it's a numbers game, rather than what people think.
Amazon knows a very, very shallow amount about any particular individual. People think that Amazon knows a lot, but it does not (nor does it really care to). But knowing a very shallow amount about you, is a lot more than TV advertisers, which know pretty much nothing about you other than where you live.
So that very little amount is a definite edge for Amazon. On an individual level, it's on the magnitude of cents. Multiplied by all Amazon customers, it's a large sum.
You don't need to have as much info on a guy as the CIA does on a target they think is a spy to make money. You don't need to make a lot of money from every customer. You can know a bit, and make a bit of money, over a hundred million customers and it adds up.
I guess its not smart enough to figure out all the things I'm buying are literally from other peoples lists that I am already using on their platform, with the "this is a gift" checkbox checked?
Very surprised they can't mesh their data well considering its Amazon.
We haven't saved anything yet. As you shop on Amazon or chat with Alexa, we'll save helpful information to personalize your experience.
Once we have saved some details, you can manage them here.
I guess that's good then - I also haven't bought anything on Amazon in almost a year (not a brag, I just don't really buy things much these days and it likely directly affects these results). Though you'd think at the least it would say something like "buys last minute Christmas gifts here"
Their information on me is fairly accurate. Favors professional-grade tools over consumer versions
Likely manages landscaping and yard improvement projects
Likely maintains vehicles personally
Builds and upgrades computers
Owns a custom-built gaming PC
Likely owns a Toyota vehicle
The one they missed is Gradually building woodworking workshop
My woodshop is damn near complete, and I'm just filling in a few holes with newer types of tools, especially battery operated ones.
I typically only go to Amazon when I can't find an item anywhere else, and it's interesting to see how that gives them a distorted view of who I am. Only one line item really surprised me, and about half of them were incredibly mundane like "owns a car" and "lives in [town where I have things delivered]," etc.
[OP] zoroa | 5 hours ago
Direct link to Amazon's profile of you: https://www.amazon.com/slc/hub?useAiMode=1
I thought this was worth sharing after I looked at mine, and saw that one of its inferences was:
This is more or less correct. "Gradually" is load bearing in that description and likely captures a lot about my behavior as a consumer. I was taken aback by how specific it was able to get and the fact that this is possible at scale.
Which reminded me of this post from a couple months back, "LLMs can unmask pseudonymous users at scale with surprising accuracy", since Amazon is effectively doing the same thing to build a profile.
Akir | 2 hours ago
Hahaha, Jesus this is so incredibly inaccurate for me. To be fair, I stopped regularly buying things from Amazon quite a while ago. The only thing I bought from Amazon in 2026 was a soymilk maker which I bought out of points from my old Amazon card, which I am keeping open almost entirely to avoid damaging my credit score. Most of the inferences it had for me were some variation of "probably has a 3D printer" or "probably owns a GM vehicle" - given how many windshield wipers and filaments I've bought, I wonder how they could have ever figured that out?
The irony is that the greatest number of inferences it has for me is stuff that I have bought for other people. It has a huge number of them about how I enjoy mixology and drinks. I don't, I'm not a drinker, and I've never made a mixed drink in my life outside of a mixology class on a vacation. All of the time I spent looking up those products and buying them was because it was a very elaborate gift for my husband. There's a lot of other inferences for other kinds of gifts I've bought for other people. I guess I'm just uniquely altruistic?
stu2b50 | 2 hours ago
That's just the reality of data science at this scale and with so few data points. People like to think that data science and machine learning is literally wizardry, but it's mostly propped up by inaccurate and misleading reporting.
The famous "Target knew daughter was pregnant before parents did" is just fake, for instance.
The reality is that it's INCREDIBLY hard to make ANY inferences about someone from their Amazon purchases. That data is inherently sparse and noisy. If you DID manage to make a lot of inferences from it, it would look very suspicious to any data scientist - that looks more like over-fitting than anything.
It's always funny when people claim that Tinder or Hinge's goal is to selectively match people such that they go on dates but do not become a couple so they spend more time on the app. To anyone that's worked with data, this is the equivalent of telling someone to throw a baseball into a trashcan that's 10 miles away from you. As if, from the shit people put on dating profiles, and swipe information you're going to thread that needle.
Akir | an hour ago
Yeah, but here's the thing: I'm an outlier. It's weird that I'm not buying things on Amazon. It's weird that I don't have a Meta account for instagram or whatsapp.
But people who are invested in Amazon are probably not only buying things on amazon.com, they probably have Amazon Prime and are buying compulsively. They are probably buying groceries at Whole Foods Market. They're probably watching shows on Prime Video. They might be reading books on their Kindle or listening to their audiobooks on Audible.
There are many places they can bring datapoints in so I think that a lot of people are going to get weirdly specific predictions.
cdb | 22 minutes ago
It's true that a lot of these inferences are either obvious or slightly off-target. Sure I've bought a lot of bike-related gear over the years, so obviously I'm a frequent cyclist. Isn't that obvious? However, I kind of wonder if the general public underestimates the usefulness of these seemingly shallow insights.
The classic one I've seen on the internet is someone buying a toilet seat and getting a bunch of ads for toilet seats afterwards, as if amazon is stupid for doing that. The way I see it, any average person is highly unlikely to purchase a toilet seat at any given time. Someone who's recently purchased a toilet seat has a vastly higher chance of purchasing another toilet seat, since they might return the first one or have another bathroom to work on later.
I don't have any evidence to support it, but my guess is that amazon makes a ton of sales due to their profile-based recommendations. I guess in my case the data is not that sparse. I've purchased about a thousand things on there with purchase history going back to 2002. That's a lot of data.
stu2b50 | 2 minutes ago
They absolutely do. It's just that it's a numbers game, rather than what people think.
Amazon knows a very, very shallow amount about any particular individual. People think that Amazon knows a lot, but it does not (nor does it really care to). But knowing a very shallow amount about you, is a lot more than TV advertisers, which know pretty much nothing about you other than where you live.
So that very little amount is a definite edge for Amazon. On an individual level, it's on the magnitude of cents. Multiplied by all Amazon customers, it's a large sum.
You don't need to have as much info on a guy as the CIA does on a target they think is a spy to make money. You don't need to make a lot of money from every customer. You can know a bit, and make a bit of money, over a hundred million customers and it adds up.
Grenno | 54 minutes ago
I guess its not smart enough to figure out all the things I'm buying are literally from other peoples lists that I am already using on their platform, with the "this is a gift" checkbox checked?
Very surprised they can't mesh their data well considering its Amazon.
sparkle | 3 hours ago
I guess that's good then - I also haven't bought anything on Amazon in almost a year (not a brag, I just don't really buy things much these days and it likely directly affects these results). Though you'd think at the least it would say something like "buys last minute Christmas gifts here"
Kerry56 | an hour ago
Their information on me is fairly accurate.
Favors professional-grade tools over consumer versions
Likely manages landscaping and yard improvement projects
Likely maintains vehicles personally
Builds and upgrades computers
Owns a custom-built gaming PC
Likely owns a Toyota vehicle
The one they missed is Gradually building woodworking workshop
My woodshop is damn near complete, and I'm just filling in a few holes with newer types of tools, especially battery operated ones.
Perryapsis | 38 minutes ago
I typically only go to Amazon when I can't find an item anywhere else, and it's interesting to see how that gives them a distorted view of who I am. Only one line item really surprised me, and about half of them were incredibly mundane like "owns a car" and "lives in [town where I have things delivered]," etc.
kingofsnake | 9 minutes ago
I wonder if this is only for US customers? Amazon.ca doesn't have this listed.