Competition Law and Competition Concerns in Predictive Ecosystems

1. Introduction

A predictive ecosystem can be understood as a digital commercial environment in which firms use large datasets, algorithms, artificial intelligence, machine learning and real-time information to predict consumer behaviour, demand, prices, market conditions or competitors' likely actions.

Examples include online marketplaces, digital advertising systems, travel platforms, mobile operating systems, financial platforms, property-management systems and other businesses in which algorithms continuously analyse information and influence commercial decisions.

Predictive technologies can produce important benefits. They can improve demand forecasting, reduce costs, improve matching between buyers and sellers, detect fraud, optimise inventories and provide consumers with more relevant products.

However, competition concerns arise when prediction is combined with market power, exclusive access to data, common algorithms, network effects or control over important digital infrastructure. Competition authorities may then have to consider whether predictive technology is facilitating collusion, excluding competitors or reinforcing an already dominant ecosystem.

The main legal questions generally fall under rules dealing with:

• anti-competitive agreements and concerted practices;
• abuse or monopolisation by dominant firms;
• algorithmic price coordination;
• exchange of competitively sensitive information;
• self-preferencing;
• tying and bundling;
• discriminatory access to data;
• interoperability restrictions;
• exclusionary ecosystem design; and
• acquisitions of data, technology or emerging competitors.

2. Predictive Ecosystems and Market Power

Traditional competition analysis often concentrates on price, output and market shares. Predictive ecosystems add another important competitive asset: data combined with computational capability.

A platform may collect information concerning searches, purchases, location, advertising responses, transactions or user interactions. Machine-learning systems can convert these observations into predictions.

This can create a feedback mechanism:

More users → more data → better predictions → better service or targeting → more users → still more data.

Such a cycle is not automatically unlawful. Successful firms are normally entitled to benefit from superior products, innovation and legitimate economies of scale.

The competition concern arises where the cycle becomes difficult for competitors to challenge because the incumbent controls indispensable inputs, distribution channels or ecosystem access and then uses that position to exclude rivals rather than merely competing through a better product.

3. Algorithmic Collusion

One of the most important concerns is that algorithms can facilitate coordination between competitors.

Historically, a cartel normally required competitors to communicate about matters such as prices, production or customers. Digital systems can make coordination faster and easier.

For example, competitors could use software capable of:

1. monitoring rivals' prices;
2. detecting price reductions immediately;
3. automatically responding to those reductions;
4. processing confidential competitor information; or
5. recommending similar prices to several competing firms.

Competition law does not generally create an exemption merely because an agreement is implemented through software.

The difficult issue concerns autonomous algorithms. Suppose competing firms independently introduce AI systems that learn that matching one another's prices is more profitable than aggressive competition. Establishing an unlawful agreement can become much more complicated if there was no communication or agreement between the firms.

Therefore, competition law must distinguish conscious parallel behaviour from actual coordination or an unlawful agreement.

Important Case Laws

4. United States v. David Topkins

This is one of the clearest early cases connecting algorithms with traditional price fixing.

David Topkins was involved in selling posters through an online marketplace. The US Department of Justice charged him with participating in an agreement with another seller to fix prices. The participants used algorithm-based pricing software to implement their arrangement.

The algorithm therefore did not independently create the cartel. Instead, human competitors allegedly reached an agreement and used computer code to carry it out.

Topkins agreed to plead guilty.

Competition-law significance

The case established an important practical principle:

Using an algorithm to implement price fixing does not transform illegal coordination into lawful conduct.

The relevant competition-law inquiry remains whether competitors agreed to restrict competition. Technology can simply become the mechanism through which that agreement operates.

The case is particularly important for predictive ecosystems because automated systems can implement coordinated behaviour continuously and much faster than humans could manually.

5. Eturas UAB and Others v Lithuanian Competition Council — Case C-74/14

The Eturas case concerned Lithuanian travel agencies participating in a common online booking system.

The system administrator sent a message concerning restrictions on discounts available through the platform, and a technical modification placed a cap on discounts.

The Court of Justice of the European Union examined whether travel agencies using the system could be regarded as participating in a concerted practice.

The Court did not establish a rule that merely using the same computer platform automatically proves collusion. Questions concerning awareness, evidence and the presumption of innocence remained important.

Importance for predictive ecosystems

Eturas demonstrates why a common technological intermediary can become relevant to competition analysis.

If numerous competitors use the same platform, algorithm or decision-making infrastructure, authorities may examine:

• what information the participants receive;
• whether they know how the common system operates;
• whether the system restricts independent commercial decisions;
• whether competitors are aware of the coordination mechanism; and
• whether firms distance themselves from potentially anti-competitive conduct.

Thus, digital coordination does not necessarily require traditional face-to-face cartel meetings.

6. United States and Plaintiff States v. RealPage, Inc.

The RealPage litigation is especially important for modern predictive competition law.

The US Department of Justice alleged that competing landlords supplied RealPage with non-public, competitively sensitive information concerning matters including rents and leasing conditions.

RealPage's revenue-management software processed information and generated rental-price recommendations.

The government alleged that this system reduced independent competition between landlords and also challenged RealPage's conduct under US monopolisation rules.

The case developed significantly after the original 2024 complaint. The DOJ subsequently added landlord defendants, and proposed settlements were reached with RealPage and several property managers. As of September 2026, the DOJ's case page records settlements or proposed judgments involving RealPage, Cortland, Greystar, LivCor, Willow Bridge and Pinnacle, while related enforcement continued.

Importance

RealPage illustrates the hub-and-spoke problem in algorithmic markets.

Instead of competitors directly exchanging information with each other, they may provide information to a common intermediary. The intermediary's algorithm can then process that information and provide recommendations back to participating firms.

This raises several questions:

• Is non-public competitor information being pooled?
• Are recommendations based upon rivals' current sensitive information?
• Do participating businesses continue to make genuinely independent decisions?
• Does the software encourage users to follow common recommendations?
• Does the system reduce incentives to discount?
• Does control over the accumulated dataset create additional entry barriers?

The DOJ's 2025 proposed settlement with RealPage sought, among other things, restrictions on using competitors' non-public competitively sensitive information in rental-price determination and changes to features alleged to align pricing.

RealPage therefore represents one of the strongest contemporary examples of competition authorities examining algorithmic coordination through a common predictive platform.

7. Google and Alphabet v Commission — Google Android, T-604/18 and C-738/22 P

The Google Android proceedings concerned Google's Android ecosystem and restrictions involving Android mobile devices.

The European proceedings examined several practices involving Google's mobile ecosystem, including arrangements relating to Google Search, Chrome, the Play Store, exclusivity and restrictions concerning alternative versions of Android.

The General Court delivered its judgment in 2022.

The subsequent appeal, Google and Alphabet v Commission, C-738/22 P, was decided by the Court of Justice on 2 July 2026.

Importance for predictive ecosystems

The case demonstrates that competition analysis may need to examine the ecosystem as a connected commercial structure, rather than looking at each product completely separately.

An ecosystem may combine:

• an operating system;
• an application store;
• search services;
• browsers;
• advertising;
• user information; and
• relationships with device manufacturers.

Control over several complementary layers can potentially allow market power in one layer to influence competition in another.

For predictive ecosystems this matters because data obtained through one service can improve predictions in another service. Better predictions can improve products and benefit users, but contractual restrictions or exclusionary practices can potentially make it harder for competing ecosystems to obtain sufficient users, distribution or data.

The Android litigation therefore provides an important framework for analysing ecosystem leverage, tying, exclusivity and foreclosure.

8. Google Shopping — Google and Alphabet v Commission, T-612/17

The Google Shopping litigation is another major digital competition case.

The dispute concerned Google's comparison-shopping service and the treatment of competing comparison-shopping services in general search results.

The European Commission's theory centred on the favourable treatment given to Google's own comparison-shopping service and the corresponding treatment of competing services.

The litigation became highly important for the concept commonly described as self-preferencing.

Relevance to predictive ecosystems

A predictive platform may control both:

1. the mechanism determining rankings or recommendations; and
2. products or services competing for those rankings.

That creates a possible conflict between operating the ecosystem and competing inside it.

For example, an ecosystem's algorithm might determine:

• which seller appears first;
• which advertisement receives prominence;
• which product is recommended;
• which application receives visibility; or
• which supplier receives access to customers.

Competition concerns may arise where a dominant platform designs or operates those mechanisms in a manner capable of favouring its own services and excluding rivals.

The broader lesson from Google Shopping is that apparently technical ranking mechanisms can have major competitive consequences where access to users depends heavily upon those rankings.

9. Meta Platforms and Others v Bundeskartellamt — Case C-252/21

The Meta/Bundeskartellamt proceedings concerned the relationship between competition law and the processing of personal data by a dominant digital platform.

The Court of Justice considered circumstances in which a competition authority examining abuse of dominance could also consider compliance with data-protection rules as part of its assessment.

Importance for predictive ecosystems

Predictive AI frequently depends upon enormous datasets.

Consequently, data practices may affect competition because data can serve simultaneously as:

• an input for algorithms;
• a source of predictive accuracy;
• a means of improving advertising;
• a mechanism for personalisation; and
• a potential barrier facing smaller competitors.

The case demonstrates that competition analysis in digital ecosystems may interact with other regulatory fields, particularly data protection.

However, competition law and data-protection law remain distinct legal regimes. A data-protection issue does not automatically establish a competition-law infringement.

The central competition question remains whether conduct by a dominant undertaking constitutes an abuse under the applicable competition rules.

10. United States v. Google — Advertising Technology Litigation

US litigation concerning Google's advertising-technology business provides another important example of ecosystem competition.

Digital advertising involves interconnected services connecting publishers, advertisers and advertising exchanges. Algorithms make extremely rapid decisions about auctions, advertisement placement and pricing.

The litigation examined Google's conduct across different parts of this technological chain.

Predictive-ecosystem relevance

Advertising technology illustrates how control over several connected layers can produce concerns about:

• conflicts of interest;
• discriminatory treatment;
• interoperability;
• access to information;
• self-preferencing;
• switching barriers; and
• foreclosure of competing technologies.

In September 2026, a federal court ordered behavioural remedies concerning Google's advertising-technology practices rather than requiring the structural breakup sought by the government. The remedies included measures directed at interoperability, discriminatory treatment and publisher access to information; Google has indicated that aspects of the underlying liability ruling are being appealed.

The case illustrates that competition remedies in predictive ecosystems may concern not only prices but also technical architecture and access conditions.

11. Main Competition Concerns

These cases demonstrate several recurring competition problems.

A. Algorithmic price coordination

Competitors may use common algorithms to coordinate prices directly or indirectly.

The risk becomes particularly significant when the system processes competitors' confidential and current commercial information.

The CMA has similarly warned that common pricing systems can facilitate exchanges of confidential information and coordination instead of independent competition.

B. Data concentration

Predictive accuracy frequently improves with access to more or better-quality information.

A dominant platform possessing exceptionally large datasets may therefore have an advantage that smaller competitors find difficult to reproduce.

Data concentration itself is not necessarily unlawful. Competition concerns become stronger when data advantages are combined with exclusionary behaviour.

C. Feedback loops

Predictive ecosystems can develop reinforcing feedback loops:

Users → Data → Better Algorithm → Better Predictions → More Users.

Such effects can make entry difficult.

Authorities therefore need to distinguish legitimate growth based on superior performance from artificial barriers produced through exclusionary conduct.

D. Self-preferencing

A vertically integrated platform may operate a marketplace while simultaneously competing against businesses using that marketplace.

Its predictive systems could potentially influence ranking, recommendations or visibility.

The legal question is whether the conduct satisfies the relevant requirements for abuse or monopolisation rather than whether self-preferencing exists in the abstract.

E. Information exchange

AI systems may consume:

• prices;
• inventories;
• demand information;
• capacity;
• future business plans; and
• customer behaviour.

Where competing firms contribute sensitive information to a shared system, competition authorities may investigate whether the arrangement reduces uncertainty that should normally exist between competitors.

F. Personalised pricing

Predictive technology can estimate an individual consumer's willingness to pay.

Personalised pricing is not automatically an antitrust violation.

Nevertheless, competition concerns can arise where it operates alongside dominance, exclusion, collusion, misleading practices or exploitation prohibited by the relevant legal regime.

The UK's CMA has identified personalised pricing and algorithmic choice architecture as areas capable of producing consumer harm.

G. Tying and bundling

A powerful ecosystem may connect several products together.

For example:

Operating system + app store + search + browser + advertising + AI assistant.

Integration can create genuine efficiencies.

However, competition concerns can arise where a dominant undertaking uses contractual or technical restrictions to foreclose competing products.

The Google Android litigation provides an important example of how these issues can be analysed in a multi-sided ecosystem.

H. Interoperability restrictions

Predictive ecosystems often depend upon APIs, databases, operating systems and technical standards.

A dominant platform could potentially restrict interoperability in ways that make rival services less effective.

Competition analysis may therefore consider whether technical restrictions are objectively necessary or whether they improperly exclude competition.

I. Entry barriers

New firms may require substantial quantities of data before their predictive models become competitive.

A new entrant can therefore face a circular problem:

It needs customers to obtain data, but it needs sufficient data to create a service capable of attracting customers.

Network effects, switching costs and ecosystem integration can increase this problem.

12. The Problem of Autonomous AI Collusion

One of the most difficult future questions arises where algorithms independently learn strategies producing coordinated market outcomes.

Consider three competitors that independently deploy AI pricing systems.

None communicates with another.

Nevertheless, each AI discovers that aggressive discounting causes rivals' algorithms to retaliate immediately. The systems eventually learn that maintaining higher prices produces greater long-term returns.

Economically, the result could resemble coordination.

Legally, however, traditional anti-cartel rules commonly require some form of agreement or concerted practice.

Therefore:

Parallel algorithmic behaviour does not automatically establish an unlawful cartel.

Authorities would need to examine evidence concerning communications, system design, information exchange, common providers, instructions given to algorithms and interactions between competitors.

This distinction is essential because competition law should not treat every similar algorithmic decision as proof of collusion.

13. Common Algorithm Providers

Another difficult situation occurs where many competing businesses purchase predictive software from the same supplier.

This can produce a hub-and-spoke structure:

Competitor A → Common Algorithm Provider ← Competitor B

Competitor C → Common Algorithm Provider ← Competitor D

The common provider acts as the hub while competing businesses form the spokes.

The competition risk becomes considerably stronger where competitors knowingly provide current confidential information and the common provider uses that information to influence their commercial decisions.

The RealPage enforcement proceedings demonstrate why authorities are focusing heavily on this model.

14. Efficiencies and Pro-Competitive Benefits

Predictive ecosystems should not be presumed harmful.

They can create substantial efficiencies, including:

• better demand forecasting;
• lower inventory costs;
• faster matching of supply and demand;
• improved fraud detection;
• reduced transaction costs;
• improved product recommendations;
• more efficient logistics;
• better capacity management; and
• faster innovation.

Competition analysis therefore requires careful attention to the actual conduct, market structure and competitive effects.

The UK's competition authority, for example, recognises that pricing algorithms can produce substantial benefits while also warning that they can be misused to restrict competition.

15. Compliance Measures for Businesses

Businesses operating predictive ecosystems should preserve independent commercial decision-making.

Important safeguards include ensuring that competitors' confidential information is not improperly pooled, reviewing the inputs used by pricing algorithms, maintaining human and legal oversight of automated decisions, documenting legitimate reasons for algorithmic design choices, and carefully reviewing arrangements involving common algorithm providers.

Dominant platforms should pay particular attention to tying, exclusivity, ranking mechanisms, interoperability, access to essential ecosystem functions and treatment of businesses that compete with the platform's own services.

Companies should also understand what their AI systems actually do. Delegating a decision to software does not necessarily remove competition-law responsibility.

The recent RealPage enforcement illustrates this point particularly clearly: competition authorities have focused on whether firms continue making independent pricing decisions when algorithms use competitors' sensitive information.

16. Case-Law Summary

17. Conclusion

Predictive ecosystems are transforming competition because businesses increasingly compete not only through prices and products but also through data, algorithms, prediction capability and control over digital infrastructure.

Competition law does not prohibit predictive technology itself. The central concern is how that technology is used.

The most significant risks arise where predictive ecosystems facilitate algorithmic collusion, exchange of sensitive information, self-preferencing, tying, exclusionary ecosystem practices, data-related entry barriers or foreclosure of competitors.

The Topkins and Eturas cases demonstrate that technology can facilitate traditional coordination. RealPage brings the issue directly into modern AI-assisted pricing and common algorithm providers. Google Android and Google Shopping illustrate the significance of ecosystem power, distribution and ranking mechanisms, while Meta/Bundeskartellamt demonstrates the growing importance of data in competition analysis.

The key principle remains straightforward: commercial prediction and automation can improve competition, but firms must still make genuinely independent competitive decisions and dominant firms remain subject to rules preventing exclusionary abuse.

As AI systems become more autonomous, competition law will increasingly have to distinguish legitimate algorithmic optimisation from coordination or exclusion that undermines the competitive process.