What if the price you see is not simply the price of a product, but a price calculated specifically for you? Imagine two customers purchasing exactly the same product at the same time. One pays $80; the other pays $60. Is that fair? Most of us would probably want more information before answering. Yet AI is making this question increasingly relevant, and increasingly difficult to avoid.
We already accept different prices in many parts of our lives. Two passengers sitting next to each other on the same flight may have paid different fares. Students receive discounts at museums. Universities offer scholarships based on financial need. Hotels charge more during periods of high demand.
As AI becomes better at predicting what individual consumers are willing to pay, questions about fairness, privacy, transparency, and consumer welfare will become increasingly important.
Price differences are not inherently new, nor are they necessarily unfair. What is changing is the ability of companies to use data and AI to estimate, with increasing precision, what different consumers are willing to pay. As I discuss in my recent book, Pricing in the Age of AI, this technological shift creates both new opportunities and risks. It also transforms pricing from primarily a business decision into an increasingly important societal and policy question.
From One Price to Your Price
For much of modern commerce, pricing was relatively simple from the consumer’s perspective. A store posted a price and customers decided whether to buy. Companies could change that price over time or offer discounts, but their ability to tailor prices to individual customers was limited. Digital commerce began to change this as companies gained access to much more information about customers and could adjust prices more frequently.
AI accelerates this transformation by allowing firms to analyze vast amounts of data and make pricing decisions at enormous scale. Pricing algorithms can potentially draw on information such as transaction histories, browsing behavior, location, timing, and other contextual signals.
Instead of dividing customers into a few broad groups, AI can identify much more subtle patterns and tailor offers with high granularity.
Generative AI (GenAI) adds another dimension. Tools such as ChatGPT, Claude, and Gemini can now provide pricing recommendations without requiring a company to build a sophisticated pricing algorithm from scratch.
This democratizes access to AI-based pricing capabilities, making them available to a much wider range of businesses. The important question is no longer simply whether companies can personalize prices, but how far they should go.
Who Wins and Who Loses?
It is tempting to assume that personalized pricing must be bad for consumers. The economics are more complicated. Consider a museum that charges $20 to adults but $10 to students. The lower student price may allow people to visit who would otherwise find the museum too expensive, while the museum still collects a higher price from customers who are willing to pay more. In such cases, price differentiation can benefit both the seller and consumers at the aggregate level.
Pricing algorithms can continuously observe market conditions and competitors’ prices and adjust their own prices in response.
The same principle applies in many other markets. Different prices can help companies serve customers with different willingness to pay, potentially expanding access to products and services. Personalized discounts can also mean that some consumers receive lower prices than they would under a single-price system. But the picture changes as personalization becomes increasingly precise. Economists use the concept of consumer surplus to describe the difference between what someone is willing to pay and what they actually pay. If I am willing to pay $100 for something but purchase it for $70, I receive $30 of surplus.
Now imagine that an AI system could estimate almost perfectly that I am willing to pay $100. Instead of offering me the standard $70 price, the company could theoretically charge me close to $100. Another consumer willing to pay only $60 could receive a much lower price for the same product. In the theoretical extreme, perfect price personalization could allow companies to capture all the consumer surplus.
The product itself has not changed. What has changed is how the economic value created by the transaction is divided between the seller and the buyers. AI-driven pricing is thus not only a question of efficiency or profitability. It is also becoming a question about the distribution of economic value between firms and consumers.
What Is a Fair Price?
This leads to an apparently simple question: How can we ensure that AI sets prices fairly? The problem is that there is no single definition of a fair price. In my research with Adam Elmachtoub and Xiao Lei on price discrimination with fairness constraints, we proposed four definitions of fairness. Price fairness means offering similar prices across groups. Demand fairness focuses on similar access to a product. Consumer-surplus fairness focuses on whether different groups retain similar value after paying. A fourth definition focuses on consumers who do not purchase because the price is too high.
When GenAI is used to recommend prices, evaluating the model alone may not be enough.
Each definition sounds reasonable. But they can be incompatible. Suppose that two groups of consumers value the same product differently. Offering both groups the same price may satisfy price fairness, but it does not necessarily provide similar access.
Conversely, achieving similar purchase rates across the two groups may require charging them different prices. Our research shows that it can be impossible to satisfy even two of these fairness definitions simultaneously.
We also find that imposing fairness does not always have the same effect on economic welfare. In some settings, moderate fairness constraints can improve social welfare, while overly rigid constraints can reduce it. This distinction is important for AI pricing. Saying that an algorithm should be “fair” is not enough. We first need to decide what fairness means in a particular context. That leads to a broader point about AI. AI can optimize a definition of fairness once we give it one. But AI cannot decide which definition society should value. That remains a human decision.
When Data Determines What You Pay
AI-driven pricing raises another difficult question: What information should influence the price someone receives? AI systems can identify relationships that would be difficult for a human decision-maker to detect. That is part of their value. But it also means that removing sensitive information from a dataset does not necessarily eliminate the possibility of unequal outcomes.
Pricing algorithms can continuously observe market conditions and competitors’ prices and adjust their own prices in response.
Consider geographic location. Using a customer’s location may seem very different from using a sensitive characteristic such as race or gender. Yet location can be correlated with demographic characteristics. An algorithm could thus generate systematic differences across groups even without directly observing group membership.
GenAI introduces an additional complication: outcomes can depend not only on the model and data, but also on the instructions, or prompts, given to the model. My research with Eddy Hage-Youssef and Warut Khern-am-nuai illustrates this issue in the labor market.
As we explain in What Happens When AI Sets Wages, we asked leading AI models to recommend hourly rates for thousands of freelancers. The recommendations varied with characteristics such as geography and age, and the way prompts were formulated could also influence the results. The implication extends beyond wages. When GenAI is used to recommend prices, evaluating the model alone may not be enough. We also need to understand the data, prompts, objectives, and other choices surrounding the system.
From Better Pricing to Responsible Pricing
None of this means that AI-based pricing is inherently harmful. AI can help businesses make better decisions, respond quickly to changes in supply and demand, offer targeted discounts, improve access, and use scarce resources more efficiently. In many cases, pricing algorithms can benefit consumers as well as firms. The challenge is to capture these benefits without ignoring the risks that accompany increasingly automated and personalized decisions. These technologies can also complicate familiar concerns surrounding pricing.
In a recent article on how AI is reshaping pricing litigation, I discuss four areas in particular: algorithmic collusion, price gouging, deceptive practices, and discriminatory pricing. Algorithmic collusion provides an interesting example. Traditionally, collusion involves competitors coordinating their behavior, such as agreeing to raise prices. With algorithmic pricing, the situation can become more complex and harder to detect.
Pricing algorithms can continuously observe market conditions and competitors’ prices and adjust their own prices in response. Under some conditions, algorithms may learn pricing strategies that reduce price competition even without explicit communication among firms. At the same time, similar prices can simply reflect firms independently responding to the same market conditions. Distinguishing between these possibilities creates new challenges for competition policy.
AI does not create these issues, but its ability to process large amounts of data, adapt rapidly, and automate decisions can change how they arise and how difficult they are to detect. Pricing algorithms have already become the subject of litigation and growing regulatory scrutiny, including in rental housing and other markets. Courts and competition authorities are increasingly examining the role of algorithms and data in pricing.
Personalized pricing is also receiving growing attention. In August 2026, the U.S. Federal Trade Commission issued a proposed enforcement policy statement addressing the use of personal data to determine the price offered to an individual consumer. The FTC did not propose banning personalized pricing. Instead, the statement focuses in part on consumer expectations and transparency, including disclosure of whether a price is personalized, the basis for the personalization, and the types of data being used. AI makes this issue even more complicated.
Pricing has always reflected more than supply and demand.
Traditional discussions about data-driven pricing often center on the question: What information does a company have about me? In the age of AI, an equally important question may be: What can the company’s algorithm infer about me from the information it has? An AI system may infer economically relevant characteristics from combinations of location, browsing behavior, purchase history, device information, and other signals, even when those characteristics were never directly provided by the consumer.
These developments make transparency, auditing, and human oversight increasingly relevant. Businesses need to understand how their pricing systems behave, test them for unexpected outcomes, and determine when humans should intervene. Consumers may increasingly want to know when prices are personalized and what kinds of information influence the offers they receive.
For policymakers, AI-driven pricing intersects with issues that have traditionally been considered separately, including privacy, discrimination, consumer protection, competition, and price transparency. The goal is not simply to prevent companies from using AI for pricing, nor to assume that more personalization is always desirable. The challenge is to preserve the economic benefits of AI while addressing the new risks created by increasingly automated and individualized prices.
The New Politics of Pricing
Pricing has always reflected more than supply and demand. It also affects access, affordability, and how the gains from economic exchange are divided. AI does not change these fundamental questions. What it changes is our ability to make pricing decisions with unprecedented speed, scale, and precision. Companies can increasingly estimate not only what a product is worth on average, but potentially what it is worth to you.
In Pricing in the Age of AI, I argue that this evolution toward increasingly personalized pricing will bring both economic opportunities and new challenges. As AI becomes better at predicting what individual consumers are willing to pay, questions about fairness, privacy, transparency, and consumer welfare will become increasingly important.
Ultimately, technology cannot answer these questions for us. Businesses can decide how aggressively to personalize. Algorithms can optimize the objectives they are given. Policymakers can establish safeguards. But humans still have to determine what outcomes they consider acceptable. When AI can increasingly estimate what each of us is willing to pay, the critical question is no longer only what the right price is, but what principles should guide how AI sets prices.
