Automated Price-Fixing Algorithms and Antitrust Law: A Schematic Analysis

Automated pricing tools do not sit outside antitrust law. Regulators and courts are increasingly treating them as a modern mechanism through which familiar antitrust harms may occur: horizontal price-fixing, hub-and-spoke coordination, information-sharing conspiracies, monopolization through control of pricing infrastructure, and algorithm-enabled restraints in two-sided or data-intensive markets. DOJ has expressly stated that “using software as the sharing mechanism does not immunize this scheme from Sherman Act liability,” and DOJ/FTC have described “algorithmic collusion” as a live enforcement priority (U.S. Dep’t of Justice, Office of Public Affairs, Justice Department Sues RealPage for Algorithmic Pricing Scheme that Harms Millions of American Renters (Aug. 23, 2024)).

I. Core Antitrust Frameworks Into Which Algorithmic Pricing Fits

1. Section 1 of the Sherman Act: Agreement-Based Collusion

This is the primary framework for automated price-fixing claims. The central legal question is still whether there is a contract, combination, or conspiracy that restrains trade. The algorithm does not replace the agreement requirement; it changes how the agreement may be formed, implemented, or inferred.

In practice, algorithmic-pricing theories tend to fall into four categories:

A. Explicit horizontal agreement plus software execution

Competing firms agree to coordinate prices and use software to carry it out. This is the cleanest case and remains classic per se unlawful price-fixing.

B. Hub-and-spoke coordination

A software vendor acts as the hub, and rival firms are the spokes. Liability depends on whether plaintiffs can plausibly show a horizontal agreement or conscious commitment among the rivals, not merely parallel use of the same vendor. The recent Gibson litigation illustrates the point: the district court dismissed with prejudice, and the Ninth Circuit later affirmed dismissal where plaintiffs failed to plausibly allege the necessary restraint or agreement in the hotel-room market.

C. Information-sharing cartel theory

The software aggregates nonpublic, competitively sensitive data from rivals and feeds recommendations back into the market. DOJ’s RealPage case is the clearest current example: DOJ alleges that RealPage contracts with competing landlords to share nonpublic pricing and lease-term information, which the algorithm then uses to generate pricing recommendations based on rivals’ sensitive information.

D. Tacit or autonomous algorithmic collusion

This is the hardest category. If rival firms independently deploy pricing algorithms that learn to coordinate without proof of agreement, current U.S. law is not well-settled on whether parallel algorithmic behavior alone is enough. The doctrinal obstacle is that Section 1 still requires agreement, not merely conscious parallelism.

2. Rule of Reason vs. Per Se Treatment

If plaintiffs can show classic horizontal price-fixing, the conduct may be treated as per se unlawful. But where the theory depends on software licensing agreements, data-sharing architecture, or mixed vertical-horizontal arrangements, courts often move into rule-of-reason territory or dismiss before reaching that full analysis if no cognizable restraint is pleaded. Gibson is important here: the Ninth Circuit held that if the contested software licensing agreements do not restrain trade in the relevant market at all, the case can fail before rule-of-reason balancing even begins.

3. Section 2 of the Sherman Act

Section 2 becomes relevant where the algorithmic-pricing provider itself is alleged to possess or maintain monopoly power in a market for pricing infrastructure or revenue-management software. DOJ’s RealPage announcement expressly paired its Section 1 coordination theory with an allegation that RealPage used the scheme and its data trove to maintain a monopoly in commercial revenue-management software.

II. Schematic: How Courts and Regulators Analyze Algorithmic Price-Fixing

Step 1: Identify the theory of coordination

The first question is not “Was an algorithm used?” but “What legal mechanism is alleged?”

• direct competitor agreement;

• hub-and-spoke conspiracy;

• unlawful information exchange;

• vertical agreements that collectively restrain a downstream market;

• monopolization through control of pricing infrastructure.

  
  

If plaintiffs pick the wrong theory, the case often fails early.

Step 2: Define the relevant market, unless direct effects suffice

Courts often require a relevant market, especially under the rule of reason. The Supreme Court has said courts usually cannot properly apply the rule of reason without an accurate definition of the relevant market.

That said, direct proof of actual anticompetitive effects can sometimes obviate elaborate market analysis. The classic formulation is that proof of actual detrimental effects, such as reduced output, increased price, or deterioration in quality, can substitute for a full market-power inquiry in some settings. That principle remains important in digital cases.

Step 3: Prove a restraint or agreement in that market

This is the key burden in digital-collusion cases.

In Gibson, the Ninth Circuit emphasized that Section 1 requires a causal link between the contested agreement and a restraint of trade in the relevant market. Plaintiffs could not survive merely by alleging that several competitors licensed the same software and prices later rose. The court held that neither the terms nor the operation of the licensing agreements, as alleged, restrained competition in the hotel-room market. Gibson v. Cendyn Grp., LLC, 148 F.4th 1069.

By contrast, in Duffy v. Yardi, the district court denied the motion to dismiss where plaintiffs alleged more than common software usage: they alleged a conspiracy involving sharing detailed nonpublic information, a continuing horizontal agreement among lessors, and a shared understanding that Yardi would use that information to recommend supracompetitive rents. Duffy v. Yardi Sys., 758 F. Supp. 3d 1283.

So the dividing line is not “algorithm yes/no.” It is whether plaintiffs can plausibly connect the software architecture and data-sharing arrangement to an actual restraint of trade among market participants.

Step 4: Show anticompetitive effects

If the case reaches effects analysis, plaintiffs generally must show actual adverse effects on competition or facts supporting market power plus likely harm.

The Supreme Court in Amex restated that plaintiffs bear the burden at the first step to prove substantial anticompetitive effects in the relevant market. In the two-sided platform setting, the Court required proof that the restraint increased the overall transaction cost above a competitive level, reduced output, or otherwise stifled competition in the two-sided market.

That framework matters for digital-collusion cases involving marketplaces, booking platforms, ad tech, or other transaction platforms. If the alleged algorithmic restraint affects only one side of a two-sided platform, plaintiffs may have to prove harm to the market as a whole, depending on how the platform is characterized.

Step 5: Rebut procompetitive justifications

If defendants offer efficiency explanations, such as better forecasting, reduced vacancy, faster response to demand, or lower search costs, the burden typically shifts to them to substantiate those justifications. Plaintiffs then may need to show that the justifications are pretextual, less restrictive alternatives exist, or the harms outweigh the benefits. The Microsoft burden-shifting formulation remains influential in antitrust analysis generally.

III. The Burden of Proof in Digital Collusion Cases

The burden varies by the theory, but the practical requirements are these:

A. Pleading stage

A complaint must do more than allege:

• use of the same algorithm,

• higher prices after adoption,

• parallel conduct, or

• a vendor-client relationship.

That was not enough in Gibson. The Ninth Circuit rejected the proposition that independent decisions by competitors to buy the same pricing software, followed by higher prices, automatically trigger Section 1 scrutiny.

At the pleading stage, plaintiffs usually need nonconclusory facts supporting one or more of the following:

• a horizontal agreement or conscious commitment to a common scheme;

• exchange of confidential or competitively sensitive information;

• coordinated adherence to algorithmic recommendations;

• software rules that meaningfully constrain competitive independence;

• actual adverse effects in the relevant market.

  
  

B. Proof of agreement

This remains the hardest element.

For hub-and-spoke cases, plaintiffs typically must show:

1. vertical agreements between the hub and each spoke;

2. a horizontal agreement, tacit understanding, or common commitment among the spokes; and

3. a restraint in the relevant market caused by that arrangement.

Without the “rim,” the wheel collapses.

C. Proof of effects

Plaintiffs may proceed by:

• direct evidence of higher prices, reduced output, worse quality, or restrained competitive incentives; or

• indirect evidence through market definition, market power, and likely anticompetitive effect.

As a matter of doctrine, actual detrimental effects can sometimes substitute for a full market-power showing. But many courts remain reluctant to dispense with market definition unless the direct-effects evidence is strong.

D. Two-sided platform complications

If the platform is legally characterized as two-sided, Amex can raise the plaintiff’s burden significantly. The plaintiff may have to prove harm across the platform, not just on one side. That can make algorithmic-collusion claims harder in booking, ad-tech, marketplace, or payment ecosystems.

IV. What Regulators Appear to Be Doing

Current agency activity suggests three trends.

1. Treating algorithmic coordination as ordinary antitrust, not a loophole

DOJ’s RealPage case and the joint FTC/DOJ hotel-room statement make clear that agencies view algorithmic coordination as fully reachable under existing antitrust law.

2. Focusing on data-sharing architecture

Agencies appear especially interested where rivals feed nonpublic, competitively sensitive information into a common system that then generates pricing outputs for them. That architecture looks less like independent optimization and more like a centralized coordination mechanism.

3. Studying surveillance-pricing and individualized pricing ecosystems

FTC’s 2024 orders and 2025 issue spotlight show concern not only with cartel-style coordination but also with AI-driven “surveillance pricing,” where firms use extensive consumer data to target prices. That is not identical to collusion, but it reflects the broader regulatory concern with algorithmic pricing systems built on large data ecosystems. (Fed. Trade Comm’n, FTC Issues Orders to Eight Companies Seeking Information on Surveillance Pricing (July 23, 2024)).

V. Practical Schematic for Litigation

A useful schematic for courts and litigants is:

Algorithm used→ not enough by itself

Common vendor + parallel price increases→ still usually not enough by itself under Section 1

Common vendor + nonpublic rival data sharing + algorithmic recommendations + adherence or constrained independence→ plausible coordination theory

Plus direct effects or market-power evidence→ stronger rule-of-reason case

Plus evidence of explicit agreement among rivals→ strongest path toward per se treatment

VI. Bottom Line

Automated price-fixing claims are being absorbed into existing antitrust doctrine, not replacing it. The core questions remain familiar: Is there an agreement? Is there a restraint in the relevant market? Are there actual adverse effects on competition? Does the conduct fit per se condemnation or require rule-of-reason analysis?

The real doctrinal pressure point is the burden of proving agreement in a digital environment. Courts have shown they will not infer unlawful collusion merely because rivals use the same software and prices rise. But where plaintiffs can allege or prove shared nonpublic data flows, algorithmic feedback loops, constrained independent decision-making, and actual market effects, the existing antitrust framework is fully capable of reaching digital collusion. The unsettled frontier is not whether antitrust applies. It is how much algorithmic interdependence, without more, is enough to count as agreement.