A pseudonymous Amazon customer, @fangirlinmegan, says the company's internal profiling system inferred she has 'flat buttocks' from her purchase history. She shared her discovery on social media, and the post has since gone viral, drawing attention to how e-commerce platforms use machine learning to build detailed profiles of shoppers.
The label wasn't something she volunteered. It appeared among other data points Amazon had compiled about her, according to her account. The exact mechanism behind the inference hasn't been disclosed, but it points to a broader practice: retailers using purchase patterns to predict characteristics customers never explicitly revealed.
What the purchase history revealed
Amazon's recommendation engine and internal analytics rely on machine learning and natural language processing to sift through what people buy, search for, and view. From that raw data, the system can draw conclusions—some of them surprisingly personal. In this case, the inference was about body shape, a trait that has no obvious connection to the items in a shopping cart.
The label surfaced in a way that allowed the customer to see it directly, which is unusual. Most profiling happens behind the scenes, invisible to the people it describes. That invisibility is part of what makes the episode unsettling for privacy advocates: you can't correct or challenge a characterization you don't know exists.
Amazon has not publicly commented on the specific case. The company's privacy policy does mention that it uses data to personalize recommendations and improve services, but it doesn't detail the full range of inferences its systems can generate.
How machine learning infers physical traits
Machine learning models are designed to find patterns. When fed years of transaction records, browsing behavior, and demographic signals, they can cluster customers into groups that correlate with attributes like age, income, or even health status. Physical traits are no exception. A model might notice that people who buy certain clothing sizes or brands also tend to share a particular body type, and then apply that correlation to new customers.
Natural language processing adds another layer. Product reviews, search queries, and customer service chats contain words that hint at physical characteristics. Even without direct statements, the language people use can be parsed for signals. Over time, the system builds a profile that goes well beyond what a customer intentionally shares.
The problem isn't that the technology is inherently malicious. It's that the inferences can be wrong, and they're often invisible. A mislabeled body type might affect which clothes get recommended, but it could also feed into other systems—insurance, credit, or employment—if the data were ever shared or leaked.
The GDPR question lurking in the background
For users in the European Union, the General Data Protection Regulation gives people the right to access, correct, and erase personal data. Inferred characteristics count as personal data under GDPR, even if the company never explicitly asked for them. That means an EU customer in a similar situation could demand to know what labels have been applied to them and request that inaccurate ones be fixed.
Whether Amazon's profiling practices fully comply with those rights is an open question. The regulation also requires companies to explain the logic behind automated decisions that significantly affect individuals. A label like 'flat buttocks' may not seem significant on its own, but it's part of a larger profile that could influence what a customer sees, what they're offered, and what they pay.
Amazon isn't the only retailer using this kind of analysis. The practice is widespread across e-commerce, social media, and advertising. What's different here is that a customer caught a glimpse of the machinery and decided to share it.
What happens next for the customer and the company
@fangirlinmegan's post has sparked conversations about data transparency, but it's unclear whether she plans to file a formal complaint. Amazon could face scrutiny from regulators if the incident prompts an investigation, particularly in Europe. For now, the company hasn't signaled any changes to its profiling methods.
The bigger question is whether customers will start asking for more visibility into the labels attached to them. Most people have no idea what inferences have been drawn from their shopping habits. This case shows how easy it is for those inferences to cross a line—and how hard it is to know they exist until someone stumbles onto them.
