AI robots subtly manipulating prices in a digital marketplace.

AI Price Wars: Are Algorithms Secretly Colluding to Raise Prices?

"New research uncovers how artificial intelligence in two-sided markets might be leading to higher prices, and what we can do about it."


Have you ever noticed how prices for certain products online seem to creep up in unison, even when different retailers are selling them? It might not be a coincidence. Algorithmic pricing, powered by artificial intelligence (AI), is becoming increasingly common, and with it comes the potential for tacit collusion – a situation where companies indirectly coordinate to raise prices without explicitly communicating.

A new study dives deep into how AI algorithms, specifically those using a technique called Q-learning, behave in two-sided markets. These are markets where two distinct groups, like buyers and sellers, interact through a platform (think of online marketplaces like Amazon or ride-sharing apps like Uber). The study reveals some unsettling truths about AI's ability to learn and exploit market dynamics to drive up prices.

While AI promises efficiency and innovation, its application in pricing raises significant questions about fairness and transparency. This article will break down the key findings of this research, explore the real-world implications, and discuss potential solutions to keep these AI price wars in check.

AI Search Multiple angles on this topic

A Quiet Expansion of Algorithmic Pricing

Because pricing algorithms are proprietary and rarely publicized, there are no reliable public statistics on how many businesses actually set prices through software. Reported cases of suspected algorithmic coordination are growing in number but remain largely anecdotal and contested. What can be said with confidence is that dynamic pricing software now operates across many consumer markets, from airline tickets and hotel rooms to ride-hailing and fuel. The real impact on consumers, who may be paying higher prices without knowing an algorithm set them, has not yet been rigorously quantified.

Comparable Pricing: The Baseline Algorithms Learn From

Establishing a fair baseline price for almost any good is a comparative exercise, and the used-car market illustrates the standard approach. Listing aggregators such as TrueCar advertise transparent pricing so shoppers can compare deals, presenting more than 68 used 2019 Toyota Camrys in Seattle alone and over 380 listings nationally. Valuation guides such as Kelley Blue Book add reference points, publishing a fair market price alongside depreciation, trade-in, and resale values. When pricing software is trained on such comparables, the resulting price depends heavily on how that market baseline is defined, and algorithms can shift those baselines in ways consumers cannot easily see.

From Handshake Agreements to Automated Markup

Classic price-fixing has a long, well-documented history, with competitors meeting privately to set prices or divide markets. Algorithmic pricing is a comparatively recent development, rising with e-commerce and platforms that reprice goods automatically, sometimes many times a day. It is tempting to read that history as a direct line to modern software, but the analogy is not exact: algorithms respond to data and incentives rather than to explicit agreements. The historical record thus offers cautionary precedents but no settled verdict on whether today's pricing software crosses the same legal line.

Decoding Algorithmic Collusion: How AI Learns to Hike Prices

AI robots subtly manipulating prices in a digital marketplace.

The research focuses on AI agents using Q-learning, a type of reinforcement learning where algorithms learn optimal strategies through trial and error. The study found that AI-driven platforms consistently achieve higher levels of collusion compared to traditional Bertrand competition, a model where companies independently set prices to maximize profits.

But what factors amplify this collusive behavior? The study identifies network externalities as a key driver. Network externalities occur when the value of a product or service increases as more people use it. Think of social media platforms – the more friends you have on a platform, the more valuable it becomes. The research indicates that AI algorithms are adept at exploiting these network effects to maximize profits, leading to even greater collusion.

  • Network Externalities: Increased network externalities significantly enhance collusion, indicating AI algorithms exploit them to maximize profits.
  • User Heterogeneity: Higher user heterogeneity and greater utility from outside options generally reduce collusion.
  • Discount Rates: Higher discount rates increase collusion, especially with significant network externalities.
  • Feasibility: Tacit collusion remains feasible even at low discount rates.
AI Search Multiple angles on this topic

Early Findings, Open Questions

Research on whether pricing algorithms can effectively collude is still in its early stages, and much of the published work is based on simulations rather than field data. Some studies suggest that simple pricing bots can learn to coordinate on higher prices without being programmed to communicate directly. Critics of that line of research note that real markets retain enough competition to undercut high prices, even when some participants use software. Until more real-world evidence accumulates, algorithmic collusion should be treated as a hypothesis under active investigation rather than an established finding.

The Skeptics' Case: Competition and Legal Hurdles

A robust set of counterarguments holds that algorithmic price-fixing is far from proven. Most pricing software optimizes against its own data and incentives rather than coordinating with rivals, and a high price set by one algorithm can be undercut by any competitor. Proving collusion in court also requires showing an agreement or its functional equivalent, which is difficult when the behavior emerges from code rather than conversation. Even among researchers who take algorithmic collusion seriously, some simulations fail to reproduce sustained coordination once markets include enough independent actors. These failures caution against treating automated pricing itself as evidence of wrongdoing.

Uneven Automation Across Markets

The use of algorithmic pricing varies sharply across industries, which makes meaningful comparison difficult. Markets with perishable inventory and real-time demand, such as air travel and hotels, have long relied on dynamic pricing, while many retail categories change prices far less often. Ride-hailing and delivery apps adjust prices continuously based on local supply and demand, attracting the most public scrutiny. Because these systems differ so much in their design and disclosure, drawing a single conclusion about the impact of algorithmic pricing across the economy is not yet possible.

The study also explored the impact of user behavior and market dynamics. They found that higher user heterogeneity (meaning a greater variety in user preferences) and greater availability of outside options (alternatives to the platform) tend to decrease collusion. Conversely, higher discount rates (when companies prioritize short-term gains over long-term consequences) tend to increase collusion, especially when network externalities are strong. Interestingly, the research found that tacit collusion can remain feasible even when companies aren't heavily focused on immediate profits, suggesting this behavior could be more pervasive than previously thought.

Fighting Back Against the AI Price Hike

So, what can be done to mitigate the risks of AI-driven price collusion? The researchers propose incorporating a penalty term into the Q-learning algorithm. This penalty would disincentivize algorithms from setting prices that are significantly higher than the average market price. While this is just one potential solution, it highlights the need for proactive measures to ensure that AI benefits consumers rather than exploits them. As AI becomes increasingly integrated into our economic systems, it's crucial to develop regulatory frameworks and ethical guidelines to prevent algorithmic collusion and promote fair competition. Staying informed, demanding transparency, and supporting policies that protect consumers are essential steps in navigating this new landscape.

AI Search Multiple angles on this topic

No Consensus, But a Shared Concern

Expert opinion on algorithmic price-fixing is split, with some arguing that current software poses a genuine antitrust risk while others insist that ordinary competition is sufficient to contain it. What commentators broadly agree on is that the question matters: if algorithms can hold prices above competitive levels without explicit communication, conventional antitrust tools may be poorly suited to detect it. Regulators around the world have begun asking how much visibility they actually have into the pricing logic of major platforms. Until clearer evidence and doctrine emerge, the honest synthesis is that algorithmic collusion remains a serious open risk rather than a documented practice.

The AI Mainstream and What It Portends

Whatever happens next will unfold against the backdrop of an AI boom moving at extraordinary speed. ChatGPT reached 100 million monthly active users within two months of its release and reportedly passed 900 million weekly active users by February 2026, shifting from a research demonstration into an everyday tool. OpenAI describes the underlying model as one that interacts conversationally, answers follow-up questions, and refines its responses, which illustrates how quickly AI has become a normal part of commerce and decision-making. As pricing software rides the same wave of adoption, the practical question is not whether AI will be used in pricing decisions but how visible and accountable those decisions will be.

Transparency as the Real Battleground

At a systemic level, the deeper challenge around algorithmic pricing may not be the technology itself but the visibility of the data behind prices. Many industries already route vast amounts of pricing information through centralized providers, yet consumers and regulators rarely have equivalent access to the logic that produces a final price. Even where the underlying data is nominally available, it can be fragmented and unstructured, making independent scrutiny difficult. The systemic risk is that pricing becomes simultaneously ubiquitous and opaque, a combination that undermines trust far more than any single algorithm could.

What Consumers Stand to Lose: Trust

For ordinary consumers, the stakes of algorithmic pricing are both financial and emotional, since repeatedly paying what feels like an arbitrary, AI-determined price can erode trust in entire marketplaces. The effects may also be uneven, because automated pricing can respond to a shopper's device, location, or history in ways that many customers never notice. Because these systems act invisibly and in real time, consumers have limited ability to shop their way around a price a machine has already set for them. The human-scale consequence, if algorithmic coordination ever becomes widespread, is a sense that the market no longer belongs to buyers.

About this Article -

Written with AI assistance from published research, and reviewed by the Mystum team. See our About page for more information.

This article is based on research published under:

DOI-LINK: https://doi.org/10.48550/arXiv.2407.04088,

Title: Artificial Intelligence And Algorithmic Price Collusion In Two-Sided Markets

Subject: econ.gn cs.ai cs.gt q-fin.ec

Authors: Cristian Chica, Yinglong Guo, Gilad Lerman

Published: 04-07-2024

Everything You Need To Know

1

What is algorithmic pricing and how does it relate to AI price wars?

Algorithmic pricing utilizes artificial intelligence (AI) to automatically adjust prices. The research explores how AI, specifically through Q-learning, is used in two-sided markets, such as online marketplaces, where buyers and sellers interact. This can lead to tacit collusion, where companies indirectly coordinate to raise prices without explicit communication, sparking AI price wars.

2

How does Q-learning contribute to AI-driven price collusion?

Q-learning, a type of reinforcement learning, enables AI algorithms to learn optimal pricing strategies through trial and error. The study found that AI agents using Q-learning in two-sided markets are prone to collusive behavior. They learn to exploit market dynamics, especially network externalities, to maximize profits, often leading to higher prices for consumers.

3

What are network externalities and how do they affect the AI's collusive behavior?

Network externalities refer to the increased value of a product or service as more people use it, like social media platforms. The study indicates that AI algorithms are adept at exploiting network effects. Increased network externalities significantly enhance collusion, indicating AI algorithms exploit them to maximize profits. This means AI algorithms are more likely to collude and raise prices when network effects are strong, as the platform becomes more valuable to users.

4

Besides network externalities, what other factors influence the likelihood of AI-driven price collusion?

User heterogeneity and the availability of outside options can reduce collusion. Higher user heterogeneity (meaning a greater variety in user preferences) and greater availability of outside options (alternatives to the platform) tend to decrease collusion. Conversely, higher discount rates (when companies prioritize short-term gains over long-term consequences) tend to increase collusion, especially when network externalities are strong.

5

What solutions are proposed to mitigate the risks of AI price collusion?

The researchers propose incorporating a penalty term into the Q-learning algorithm. This would disincentivize algorithms from setting prices that are significantly higher than the average market price. This penalty aims to curb the AI's tendency towards collusive pricing. Furthermore, the article highlights the need for regulatory frameworks and ethical guidelines to prevent algorithmic collusion and promote fair competition, and also suggests being informed, demanding transparency, and supporting consumer-protective policies.

Newsletter Subscribe

Subscribe to get the latest articles and insights directly in your inbox.