How Uber uses AI to charge different riders different prices

Open Uber, type in a destination, and you see a single number: the price. It looks simple and transparent. But behind that one number sits a complex AI system deciding how much you’ll pay, how much your driver will earn, and how much Uber will keep.

From cheap rides to AI-powered pricing

In its early days, Uber grew by being cheaper than taxis. Riders got low fares, drivers kept around 75–80% of each trip, and Uber took a small cut. Prices were based on a straightforward formula: a base fee plus a fixed rate per minute and per mile, like a digital taxi meter.

When demand spiked, Uber added surge pricing, multiplying fares when there weren’t enough drivers. Riders hated it, drivers loved it, and even with surge, Uber still lost billions of dollars as it expanded globally and subsidized rides with investor money.

By the time Uber went public in 2019, it had racked up nearly $9 billion in losses. The company needed a way to turn growth into profit without simply slapping a visible surcharge on every ride.

What upfront pricing actually changed

Uber’s answer was “upfront pricing.” Instead of showing a meter-style breakdown, Uber now gives riders a single price before they confirm the trip, and it shows drivers a fixed payout before they accept.

On the surface, that sounds great: no surprises, no guessing, no watching the meter climb. But behind the scenes, this change broke the tight link between what riders pay, what drivers earn, and how the fare is calculated.

Uber says its algorithm now uses many data points to set prices: estimated trip time, distance, time of day, route, demand patterns, tolls, taxes, and fees. That’s the official story. But researchers argue that upfront pricing also enables something else: algorithmic price discrimination.

Algorithmic price discrimination: charging what you’re likely to accept

Economist Len Sherman from Columbia Business School describes Uber’s system as “algorithmic price discrimination.” In plain language, that means using data and AI to:

• Estimate the maximum price each rider is likely to accept for a given trip
• Estimate the minimum payout nearby drivers are likely to accept
• Set a price and a payout that maximize Uber’s profit in the middle

Instead of one universal formula, the algorithm can nudge prices up or down based on patterns in rider and driver behavior. It doesn’t have to know you personally; it can infer how people like you tend to respond to certain prices and conditions.

Uber denies that it personalizes prices based on personal data and rejects the label “surveillance pricing.” It says prices are based on marketplace conditions, not individual profiles. But the way the system behaves raises questions.

Why the same ride can cost different amounts

One of the clearest signs something has changed is that different people can see different prices for the same trip at the same time.

In one test, multiple people in the same office requested an UberX from the same pickup point to the same destination. The highest quote was about 21% more than the lowest. In another route test, the price gap was smaller—around 2.5%—but still there.

Uber’s explanation is that small GPS differences in pickup points can change the route and therefore the price. That’s plausible in some cases, but larger studies suggest there’s more going on.

Consumer Reports, using over 170 volunteers comparing Uber and Lyft quotes on the same routes at nearly the same time, found some routes where the median gap between the highest and lowest prices was around 50%. Uber called the methodology flawed and said discounts skewed the numbers, but the pattern of inconsistent pricing remains hard to ignore.

What Uber’s own disclosures and patents reveal

If you’re in New York and scroll to the bottom of a quote in the app, you may see a line that says: “This price was set by an algorithm using your personal data.” That disclosure exists because New York law requires companies to tell you when personal data helps set your price.

Uber says that in this context, the only personal data it uses is your location, which legally counts as personal data. But Uber’s patents show how far its technology could go.

One patent describes using signals like tapping accuracy, typing speed, phone angle, and walking speed to infer your state and behavior. Another explains how ride history can be used to infer that someone is a single working parent, their likely age, and gender, based on patterns like daycare and office drop-offs.

Uber says these patents don’t prove that such features are used today and that they don’t relate to pricing. Still, they show the level of behavioral insight Uber’s AI systems can potentially achieve—and how easily that kind of data could feed into pricing decisions in the future.

Riders pay more, but where does the money go?

Since upfront pricing rolled out widely, average fares have climbed sharply. Between 2018 and 2022, average Uber fares rose by about 83%, nearly four times the rate of inflation for those years. Uber doesn’t dispute that fares have gone up significantly, but points to inflation, higher fees, and a post-pandemic driver shortage as key reasons.

Those higher prices helped transform Uber’s finances. In 2023, the company reported its first-ever annual profit of $1.1 billion, and its stock price more than quadrupled from a 2022 low. Upfront pricing is widely seen as a core part of that turnaround.

But higher fares don’t automatically mean drivers are earning more. In fact, many drivers say they’re getting a shrinking share of what riders pay.

How AI squeezes driver earnings

Under the old model, drivers typically kept around 75–80% of the fare. With upfront pricing, drivers see a fixed payout and have only a few seconds to decide whether to accept a trip. They no longer see a clear formula tied to time and distance.

Drivers like Levi in Syracuse and Bill, a veteran with nearly 40,000 rides, say that now the real skill isn’t driving well—it’s “playing the game” of the algorithm. They have to quickly estimate whether a trip is profitable, knowing that many offers are barely worth taking once gas, maintenance, and time are factored in.

In one example, a rider paid $70.52 for a trip. The driver earned $27.31—about 39% of the initial fare before tip. Uber’s service fee was $21.29, around 30% of the ride, and the company also kept a portion of the wait-time fee, even though the driver was the one actually waiting.

Studies back up these experiences. Research on hundreds of drivers and tens of thousands of trips found that Uber’s median take rate rose from around 25% to 29% under upfront pricing and sometimes exceeded 50%. Another analysis of roughly 50,000 trips by three experienced drivers found Uber’s average take rate above 50% for each of them.

Uber disputes that its take rate has increased and says the growing gap between rider payments and driver payouts is largely due to external costs, especially commercial auto insurance. But when drivers track their own numbers, they often see insurance and operational fees fluctuating in ways that don’t clearly match distance, time, or risk.

The black box problem

The core issue for both riders and drivers is that the system has become a black box. You see the final price or payout, but not how the algorithm arrived there.

For riders, that means you don’t know if you’re paying more than the person next to you for the same trip, or why. For drivers, it means you can’t easily predict what you’ll earn for a given type of ride, and you have little visibility into how much Uber is taking on each trip.

Uber’s CEO has talked about “offering the right trip at the right price to the right driver” based on driver preferences and behavioral patterns. Uber says this is about reducing driver downtime and rider wait times, but it also shows how deeply behavior is now baked into matching and pricing.

How this model is spreading across the gig economy

Uber may have pioneered upfront pricing, but it’s no longer alone. By late 2022, Lyft had expanded upfront pay for drivers across dozens of markets and reported its first annual profit two years later. Delivery platforms like Instacart and DoorDash use similar variable payout systems, even if they don’t call it “upfront pricing.”

This same pattern—AI-driven, opaque pricing and payouts—is spreading across gig work and digital marketplaces. As AI models become more capable and easier to deploy, we’re seeing similar optimization logic show up in other industries too, from e-commerce to digital services. If you’re interested in how powerful AI systems are being integrated into everyday tools, it’s worth looking at how they’re being orchestrated across platforms in one workspace for every AI model.

Why trust is the real casualty

For many riders, the feeling is simple: prices are higher, and they don’t know why. For many drivers, the feeling is harsher: they’re doing the work, taking on the risk and costs, and watching an algorithm slice up the money in ways they can’t understand or challenge.

That’s the deeper cost of AI-driven, opaque pricing systems—trust erodes. People don’t necessarily object to a company making money; they object to feeling manipulated by an invisible system that has far more information than they do.

As AI becomes more embedded in pricing, work, and everyday apps, understanding how these systems operate—and demanding more transparency—will matter far beyond ride-hailing. The same skills that help people work effectively with advanced models like Claude or ChatGPT, such as interpreting outputs and questioning incentives, are increasingly important for navigating algorithmic platforms in general. If you want to turn those skills into real-world value, it’s worth exploring practical applications like those in Claude AI skills that can earn more than a college degree.

For now, Uber’s AI-driven upfront pricing has delivered what investors wanted: profit. Whether it can rebuild what riders and drivers say they’ve lost—trust—is a harder problem no algorithm can solve on its own.