A new study from researchers at Carnegie Mellon University has found that online shopping platforms, like Amazon or Expedia, can unknowingly help pricing algorithms push prices higher, simply by how they rank products for you.
The key issue lies in how these websites decide the order of products you see. Instead of showing the same product order to everyone, many websites use your past behavior to tailor the ranking, putting what they think you’ll like best at the top.
That sounds helpful. But it turns out, this personalization can make it easier for sellers using AI-powered pricing tools to charge more.
The researchers created a simulated online marketplace to test how two types of product rankings affect pricing:
-
Personalized ranking, which customizes the product order for each shopper.
-
Unpersonalized ranking, which shows the same order to all shoppers, based on general product data like quality and price.
They also used smart pricing algorithms, specifically reinforcement learning models, that learn over time to set prices for products in a competitive environment. These algorithms don’t need any coordination between sellers, they learn from trial and error what prices make the most money.
The result? Personalized rankings gave those algorithms more room to push prices up.
Here’s why: when rankings are tailored to each shopper, people tend to look at fewer products. If the one they want is already on top, they don’t scroll further. This behavior gives sellers less reason to cut prices to compete for attention.
Under unpersonalized rankings, on the other hand, shoppers are more likely to browse around. This pushes sellers to compete harder on price, which keeps things more affordable.
In their tests, the algorithms charged 29% higher prices on average when using personalized rankings compared to unpersonalized ones. Profits jumped by 74%, but consumer satisfaction, measured as the balance of value and effort in searching, fell by 13%.
This shows that personalization, even without charging different people different prices, can still hurt consumers,
Their findings stayed consistent across various tweaks to the model, whether search costs went up or down, or whether there were more sellers.
This study doesn’t accuse any specific platform of wrongdoing. Instead, it reveals a subtle design issue: features meant to help consumers can backfire when combined with smart algorithms built to maximize profit.
What does this mean for everyday shoppers?
It means the convenience of personalized rankings might come with hidden costs. And it means policymakers and tech companies may need to rethink how personalization works, especially when paired with AI pricing tools.
The researchers say one simple fix could be rethinking how products are ordered in search results. By favoring more neutral or transparent ranking methods, platforms might keep prices fairer without hurting product discovery.
For now, if you’re shopping online, it may pay to scroll a little further down the page.
Image: DIW-Aigen
Read next: New Study Reveals Silent Data Theft from Offline PCs via Smartwatches
