Third Circuit Revives Algorithmic Hotel Pricing Antitrust Case

The Third Circuit revived an algorithmic price-fixing case on July 29, 2026, reversing the dismissal despite no alleged backroom meeting among hotel rivals. The decision behind the Google News headline gives plaintiffs another chance to prove that shared pricing software coordinated Atlantic City casino-hotel rates. It also gives every company using pooled competitor data a new reason to examine how its prices are produced.

The court did not find that Caesars Entertainment, MGM Resorts International, Hard Rock International, Cendyn Group, or the other defendants violated antitrust law. It decided that the guests suing them had alleged enough facts to proceed beyond the pleading stage. That distinction matters because the allegations have not yet been tested through discovery or trial.

The conflict is larger than one hotel market. Businesses increasingly use software to analyze demand, inventory, and customer behavior before recommending prices. The ruling draws a consequential line between software that helps companies compete independently and a shared system that allegedly becomes a single pricing center.

The Third Circuit Restored a Case the District Court Had Dismissed

The immediate change is procedural, but the court’s reasoning gives the ruling significance far beyond one procedural appeal.

The case, Cornish-Adebiyi v. Caesars Entertainment, began with claims by casino-hotel guests who said they paid artificially high room rates in Atlantic City. They sued several hotel operators and Cendyn, the provider of the Rainmaker revenue-management system.

Revenue-management software forecasts demand and recommends prices intended to maximize revenue from limited inventory. Hotels have used such systems for years. The antitrust concern arises when competing hotels allegedly supply current, confidential data to a shared provider that uses the combined information in its recommendations.

The plaintiffs allege that participating hotels sent non-public room-pricing and occupancy data to Rainmaker. They say the software processed each hotel’s information alongside comparable data from competitors. The resulting recommendations allegedly entered the hotels’ room-selling systems automatically.

According to the complaint, the defendant hotels accepted Rainmaker’s recommendations about 90 percent of the time. The plaintiffs also allege that the participating properties represented more than 70 percent of the relevant casino-hotel market. Those figures remain allegations, not established findings.

A federal district court dismissed the complaint in September 2024. It concluded that the plaintiffs had not plausibly alleged the horizontal agreement needed for a hub-and-spoke conspiracy. In that structure, a central intermediary acts as the hub while competing businesses form the spokes.

The district court focused heavily on whether the hotel operators had formed a connecting “rim.” It found the allegations insufficient because the hotels adopted the software at different times, could override its recommendations, and had not allegedly communicated directly about prices.

The appellate court disagreed. In its precedential opinion, the Third Circuit said the complaint must be considered as a whole and in the context of modern pricing technology. It reversed the dismissal and returned the case to the District of New Jersey.

The panel consisted of Judges Theodore McKee, Patty Shwartz Restrepo, and D. Brooks Smith, with Judge McKee writing the opinion. The court heard arguments on September 17, 2025, before filing its decision more than ten months later.

The ruling does not impose liability, award damages, or certify a class. It permits the plaintiffs to pursue their claims and seek evidence. Defendants will still be able to challenge the allegations, the proposed market, causation, and any claimed consumer harm.

That procedural posture is central to understanding the result. Courts reviewing a motion to dismiss generally accept well-pleaded factual allegations as true. They ask whether those facts plausibly support a legal claim, not whether the plaintiff has already proved it.

The Google News framing can therefore make the decision sound more final than it is. The court opened the courthouse door. It did not decide what discovery will reveal once the parties walk through it.

Why Google News Readers Should Notice the Data Flow

The court’s strongest warning concerns the movement of confidential competitor data, not the mere presence of AI.

The word “algorithm” attracts attention, but the alleged data flow does most of the legal work. The plaintiffs describe a common system that collects current pricing and occupancy information from multiple competitors. That system then allegedly converts the pooled information into recommendations for those same competitors.

This arrangement differs from each hotel running an independent forecasting tool with its own records and public market information. Independent software can help a company respond more accurately to demand. It can lower prices when occupancy weakens, raise them during busy periods, or identify customer segments without involving rivals.

The complaint describes something more integrated. It alleges that non-public data moved from several competing hotels into one shared platform. The platform allegedly returned recommended prices shaped by that collective pool.

The court treated this exchange as a possible “plus factor.” A plus factor is circumstantial evidence that parallel business behavior reflects coordination rather than independent responses to market conditions. Other alleged factors included high recommendation acceptance, parallel pricing, falling occupancy, and participation across much of the market.

Cendyn Rainmaker pricing recommendations were not automatically unlawful simply because several hotels received them. The court explicitly acknowledged legitimate business reasons for using dynamic-pricing systems. It also said that common software alone does not establish an antitrust conspiracy.

The concern sharpened because the plaintiffs alleged that competitors benefited from each other’s current, proprietary information. The court said such information exchanges can facilitate coordination when they reduce uncertainty about how rivals will behave.

That reduction in uncertainty is economically important. Competition depends partly on each seller deciding whether to discount without knowing exactly how its rivals will respond. A hotel with empty rooms might cut rates to attract guests from a nearby property.

If a shared platform allegedly coordinates recommendations across major competitors, each participant can become more confident that another hotel will not undercut it. The plaintiffs argue that this confidence allowed higher room rates to persist even as occupancy declined.

The court did not independently inspect Rainmaker’s source code, data pipelines, model architecture, or recommendation logic. It specifically declined to assume how the software actually operates. Its analysis addressed what the complaint plausibly alleged.

That limitation separates a legal pleading from a technical audit. Discovery will need to determine which data entered the system, how current those data were, and whether competitors’ information affected individual recommendations. Investigators will also need to examine overrides, human review, configuration differences, and safeguards.

The distinction is essential for companies assessing the ruling. A vendor cannot evaluate its exposure by asking only whether its product uses AI. It must map where data originate, how information from one customer influences another, and how much independent control each customer retains.

The answer may vary across deployments of the same product. One customer might use only internal data and public signals. Another configuration might combine sensitive information from several market participants.

This is why the decision reaches beyond hospitality. Airlines, landlords, insurers, retailers, healthcare networks, and employers use systems that recommend prices, payments, rents, or wages. The legal risk follows the information and decision structure, not the industry label.

Shared Pricing Software Can Become the Alleged Hub

The central dispute pits independent pricing assistance against the alleged creation of a single decision-making center for competitors.

Section 1 of the Sherman Act requires concerted action. It generally does not prohibit a company from making an independent pricing decision, even when that decision tracks public competitor prices. The law targets an agreement that unreasonably restrains trade.

That requirement once made a familiar kind of evidence especially valuable. Investigators looked for meetings, telephone calls, messages, or explicit commitments among rivals. Shared software complicates that picture because companies can allegedly coordinate through an intermediary without speaking directly.

In a hub-and-spoke theory, each competitor forms a vertical relationship with the central provider. The plaintiff must also plausibly connect the spokes through a horizontal understanding. Without that connection, several ordinary vendor contracts do not become a cartel.

The Third Circuit concluded that the complaint plausibly supplied that connection. It pointed to allegations that the hotels knowingly contributed confidential information, expected competitors to participate, and delegated substantial pricing influence to the same system.

The defendants’ ability to reject recommendations did not end the analysis. The court reasoned that price-fixing law does not require total adherence or the elimination of every degree of discretion. A recommendation can still coordinate an important starting point, even when a company sometimes changes the final price.

The alleged 90 percent acceptance rate reinforced that conclusion at the pleading stage. If verified, it would suggest that recommendations shaped actual pricing rather than serving as one minor input. Yet the figure still requires evidentiary testing.

Timing also did not defeat plausibility. The hotels allegedly joined Rainmaker at different points rather than signing up simultaneously. The court found that a conspiracy can develop over time as companies join an existing coordinated arrangement.

This reasoning creates the principal tension for algorithmic-pricing vendors. Their products become more useful as they process more information, produce more accurate forecasts, and integrate more closely with customer systems. Those same features can increase concern when customers compete in the same concentrated market.

The Third Circuit algorithmic collusion analysis therefore turns on functional control. Courts will look beyond labels such as “advisory,” “automated,” or “AI-powered.” They will ask whether the system preserves independent decisions or allegedly substitutes collective intelligence for them.

The opinion uses a simple analogy associated with former Federal Trade Commission acting chair Maureen Ohlhausen. If a person collected confidential pricing strategies from competitors and told each one how to price, calling that person an algorithm would not change the legal concern.

The Department of Justice and Federal Trade Commission made a similar point in their 2024 hotel pricing statement. They argued that competitors cannot use software to undertake conduct that would be unlawful if performed by people.

The agencies also said direct communications among competitors are not always necessary to allege an agreement. They emphasized that retaining some pricing discretion does not necessarily insulate coordinated recommendations.

The appellate ruling broadly aligns with that technology-neutral position. It does not create a separate antitrust statute for AI. It applies longstanding principles about agreements, information sharing, and independent decision-making to a newer technical mechanism.

That approach pressures vendors and customers alike. A software provider must consider how its platform combines client information. A customer must understand whether adoption connects its decisions to confidential data supplied by rivals.

Contract language alone will not resolve those questions. A clause saying that recommendations are optional carries less weight if operational evidence shows near-universal acceptance. A promise to aggregate data also needs technical support concerning recency, granularity, and re-identification risk.

The legal issue is not whether machines are allowed to set prices. It is whether companies remain genuinely independent when the machine sits between direct competitors.

A Different Hotel Appeal Shows the Boundary Is Fact-Specific

The Third Circuit did not declare that every shared pricing platform creates a conspiracy, and another appellate case illustrates that limit.

In August 2025, the Ninth Circuit addressed related claims involving Cendyn software and Las Vegas hotels. Gibson v. Cendyn Group arose from similar concerns about algorithmic hotel pricing, but its procedural and factual posture differed.

The Ninth Circuit did not revive the plaintiffs’ abandoned hub-and-spoke claim. It affirmed dismissal of a separate theory challenging the aggregate effect of vertical licensing agreements. The court found insufficient allegations that those agreements restrained each hotel’s independent pricing authority.

The Ninth Circuit decision underscores why companies should resist broad summaries of either case. Different complaints can describe different data practices, participation patterns, agreements, and theories of liability.

The Third Circuit emphasized allegations that Atlantic City hotels contributed current, non-public pricing and occupancy data. It also considered allegations that the system pooled those inputs, returned collective recommendations, and automatically uploaded rates.

The Las Vegas litigation did not reach the Third Circuit’s present conclusion on the same record. Differences in the claims pursued on appeal also limited the comparison. One ruling therefore does not simply overrule or contradict the other.

This fact sensitivity is a feature of antitrust analysis. Parallel prices can result from an unlawful agreement, but they can also reflect common costs, public demand, seasonal events, or rational independent behavior. Concentrated markets often produce similar decisions without collusion.

Dynamic pricing can benefit consumers as well. Faster responses to weak demand can create discounts, while better forecasts can help firms allocate inventory. The Justice Department has previously recognized that algorithmic pricing can support competition when businesses act independently.

The Third Circuit acknowledged that concern directly. An amicus brief from the International Center for Law and Economics asked whether companies become conspirators merely by using identical spreadsheets or the same market-research firm. The court said those examples oversimplified the complaint.

Its answer rested on the combination of alleged facts. The confidential data exchange, high acceptance rate, automatic implementation, market coverage, and expectation of mutual participation collectively supported an inference. No single feature established liability by itself.

That makes compliance more complicated than banning a software category. A company cannot assume safety merely because humans can override a recommendation. It also should not conclude that using a common vendor is inherently unlawful.

A serious review must distinguish public information from non-public competitor data. It should determine whether inputs are historical or current, aggregated or property-specific, delayed or real-time. Each difference affects how much strategic knowledge the system can transmit.

Reviewers should also test actual behavior. How often do customers reject recommendations? Do different customers receive materially different suggestions? Can one participant’s discount influence another participant’s recommended response?

Governance matters alongside model behavior. Companies need records showing who approves pricing rules, which data features are permitted, and how antitrust safeguards work. An undocumented human override button offers little comfort if employees rarely use it.

The Cendyn Rainmaker pricing dispute will now test those distinctions more closely. Discovery might support the plaintiffs’ account, contradict it, or reveal a mixed system that fits neither side’s simplified description.

Until that evidence emerges, the Third Circuit’s ruling should be read as a warning about plausible allegations. It is not a final judgment that Rainmaker coordinated prices or harmed consumers.

The Decision Raises the Cost of Ignoring Algorithm Governance

Companies now face legal risk when they cannot explain how shared data becomes a price, even if a vendor built the system.

For corporate buyers, outsourcing a pricing function does not outsource responsibility. A hotel might license software from a third party, but its executives still need to understand the system’s inputs, recommendations, and market effects.

That responsibility begins during procurement. Buyers should ask whether a platform uses data from direct competitors, how those data are aggregated, and how quickly new information enters the model. They should also ask whether customer-specific data can influence another customer’s output.

The same questions belong in technical due diligence. Engineers can trace data lineage, which records where information originated and how it changed through a system. They can test whether removing competitor data materially changes recommendations.

Independent model instances can reduce some risks, but only if separation exists in practice. Two nominally separate models might still learn from a shared training set containing sensitive customer information. Architecture diagrams should match operational reality.

Human oversight must also be substantive. A pricing manager who automatically approves each suggestion adds little independence. Reviewers need authority, relevant market information, and documented reasons for accepting or rejecting recommendations.

Companies should monitor outcomes without assuming that similar prices prove collusion. Useful checks include recommendation acceptance, override patterns, price dispersion, and responses to demand shocks. Sudden convergence across competitors deserves investigation, not an automatic legal conclusion.

Information retention creates another issue. Pricing platforms generate extensive logs, model outputs, access records, and communications. Those materials can help demonstrate independent decision-making, but they can also reveal that optional recommendations functioned as commands.

The Third Circuit algorithmic collusion ruling increases the importance of preserving that evidence. Deleting operational records under an ordinary short retention policy can make later explanations harder. Legal, engineering, and product teams should coordinate before litigation appears.

Companies also need clear escalation paths. A pricing analyst who notices recommendations rising despite weak demand should know whom to contact. Product teams should be able to suspend questionable data features without disabling an entire commercial system.

The risk extends to generative AI. Businesses are beginning to ask general-purpose models for pricing guidance, negotiation ranges, and revenue forecasts. If several competitors use the same model with similar sensitive inputs, investigators will ask whether the setup preserves independent judgment.

A 2026 pricing litigation analysis identifies data sources, update frequency, granularity, safeguards, and human oversight as central investigative questions. Those are practical engineering questions as much as legal ones.

Knowledge management can support that work when teams must connect model documentation, vendor contracts, data maps, and review decisions. A searchable technical knowledge base can help teams preserve that context without treating compliance as a final checklist.

The Department of Justice has signaled that algorithmic collusion remains an enforcement priority. In a 2025 digital markets speech, officials highlighted housing and healthcare cases involving shared pricing systems.

Private litigation adds another pressure. Customers who believe they paid inflated prices can file claims even before regulators finish an investigation. A precedential appellate decision offers those plaintiffs a roadmap for drafting more detailed complaints.

Still, firms should avoid panic-driven restrictions that prevent legitimate forecasting. Independent pricing systems can improve inventory use and help markets react to changing demand. The governance goal is to preserve those benefits while preventing the transfer of strategic competitor information.

That requires more than adding “AI risk” to a policy. Teams must understand the specific commercial decision being automated. They must identify who supplies the inputs, who receives the output, and whose incentives the system serves.

Three Signals Will Show How Far the Ruling Reaches

Discovery, judicial treatment outside the Third Circuit, and vendor design changes will determine whether this becomes a narrow hotel case or a broader market rule.

The first signal is what discovery reveals about Rainmaker. The plaintiffs need evidence supporting their allegations about data pooling, recommendation logic, automatic uploads, and hotel acceptance. Internal documents and technical records will carry more weight than marketing language.

Evidence that each hotel’s confidential data materially affected recommendations sent to rivals would strengthen the plaintiffs’ theory. Evidence showing strict data separation, meaningful aggregation, or independent hotel decisions would weaken it.

Discovery will also test the alleged market effects. Plaintiffs point to higher rates despite declining occupancy and synchronized movement among participating hotels. Defendants can offer alternative explanations based on demand, costs, events, property differences, or ordinary revenue management.

The second signal is how other courts use the opinion. The Third Circuit covers federal courts in Pennsylvania, New Jersey, Delaware, and the Virgin Islands. Its decision binds those courts but does not control the Ninth Circuit or other federal appellate regions.

Plaintiffs in housing, healthcare, fuel, and hospitality cases will cite its treatment of confidential data and recommendation acceptance. Defendants will emphasize its pleading posture and fact-specific language. Later decisions will show which allegations courts consider essential.

A split among appellate courts could eventually attract Supreme Court attention, especially if judges apply different standards to shared algorithms. No such definitive split should be assumed from the hotel cases alone because their claims and records differ.

The third signal is how software vendors redesign their products. Providers can reduce risk by limiting current competitor data, increasing aggregation, introducing time delays, and separating customer models. They can also expose clearer controls for human review.

Design changes would show that the ruling has influenced the market before any final liability decision. Vendors might publish clearer data-governance documentation or offer configurations that exclude competitor-derived inputs.

The outcome could also affect contracting. Customers may demand audit rights, data-use restrictions, model documentation, and indemnification provisions. Vendors may respond by narrowing promises about automated revenue optimization.

Academic research will remain relevant, but simulations cannot answer what happened inside a particular commercial platform. Some studies find that learning algorithms can sustain elevated prices under controlled conditions. Other work suggests improved forecasting can increase incentives to undercut rivals.

That disagreement reinforces the court’s fact-specific approach. Antitrust law does not punish a model because a laboratory experiment produced collusion elsewhere. Investigators need evidence connecting a system’s actual data and behavior to an agreement among real competitors.

The Google News headline marks a wake-up call, not a final verdict. The Third Circuit has made confidential data, operational control, and competitor expectations central to the next stage.

Companies using algorithmic pricing should now ask a direct question: can they prove that each business still makes an independent decision? If the answer depends only on an unused override button or a vendor’s marketing claim, the decision has already identified the weakness.