Competition Law and Machine-Curated Marketplaces and Market Power

1. Introduction

A machine-curated marketplace is a digital marketplace in which algorithms, artificial intelligence (AI), recommendation systems, automated ranking tools, or machine-learning models determine, influence, or personalize what products or services consumers see and purchase.

Examples include:

• AI-generated product rankings;
• personalized search results;
• algorithmic recommendations;
• automated seller rankings;
• machine-generated discounts;
• AI-selected suppliers;
• automated product bundling;
• personalized advertisements;
• algorithmic visibility or “featured seller” systems;
• autonomous purchasing agents selecting products for consumers.

These systems can improve consumer choice and reduce search costs. However, when a marketplace becomes sufficiently powerful, machine curation itself can become a source of market power or a mechanism for exercising existing market power.

The central competition-law question is:

When does algorithmic curation improve competition, and when can it be used to create, maintain, or exploit market power?

2. Meaning of Machine-Curated Marketplaces

Traditional marketplaces allow consumers to compare products relatively directly.

A machine-curated marketplace introduces an algorithmic intermediary:

Seller → Marketplace algorithm → Ranking/recommendation → Consumer

The algorithm may determine:

• which products appear first;
• which sellers receive visibility;
• which products are recommended;
• which advertisements are displayed;
• which sellers are “preferred”;
• which products are bundled;
• which prices are highlighted;
• which customers receive particular offers.

Consequently, control over the curation layer may become economically significant.

3. Why Machine Curation Matters to Competition Law

Machine curation can affect competition in at least five ways.

1. Visibility

A marketplace controls which competitors consumers see.

2. Ranking

Algorithmic ranking can determine which seller receives the first opportunity to make a sale.

3. Access

A marketplace can determine which sellers qualify for recommendation or search visibility.

4. Data

The marketplace can collect information about:

• consumer demand;
• seller prices;
• inventory;
• conversion rates;
• product performance.

5. Consumer dependence

Consumers may increasingly rely upon the marketplace's algorithm rather than independently searching for alternatives.

This can transform a marketplace from a simple intermediary into a competitive gateway.

4. Market Power in Machine-Curated Marketplaces

Market power means the ability of an undertaking to behave to an appreciable extent independently of competitive constraints.

Traditional indicators include:

• market share;
• barriers to entry;
• customer dependence;
• switching costs;
• network effects;
• availability of substitutes.

For machine-curated marketplaces, additional indicators may be relevant:

• control over search visibility;
• number of active users;
• seller dependence;
• data advantages;
• algorithmic superiority;
• network effects;
• ecosystem integration;
• switching costs;
• access to consumer attention;
• control over recommendation infrastructure.

5. Market Definition

Before determining market power, authorities normally need to determine the relevant market.

Possible markets include:

A. General online marketplace

Different digital marketplaces compete for buyers and sellers.

B. Product-specific marketplace

For example, an online marketplace specifically serving travel, food delivery, or electronics.

C. Intermediation services

The relevant market may concern the service provided to sellers rather than consumers.

D. Attention or discovery

In some circumstances, consumer discovery and product-ranking services may become competitively significant.

E. Two-sided market

A marketplace often serves two groups:

consumers ↔ marketplace ↔ sellers

This makes competition analysis more complicated because conditions on one side can affect the other.

6. Network Effects

Machine-curated marketplaces frequently benefit from network effects.

More consumers attract more sellers.

More sellers attract more consumers.

This creates:

More users → more sellers → more products → more users → more data → better algorithms → more users

This feedback loop can strengthen market power.

7. Data as a Source of Competitive Advantage

Machine-curated marketplaces can collect enormous amounts of information.

For example:

• search history;
• purchase history;
• prices;
• seller performance;
• consumer preferences;
• product demand;
• click-through rates;
• conversion rates.

The marketplace can use this information to improve its algorithm.

This can create a data feedback loop:

More users → more data → better algorithm → better curation → more users.

Data alone does not establish dominance, but its strategic importance may contribute to competitive advantages and entry barriers.

8. Algorithmic Ranking and Foreclosure

Suppose a marketplace operates its own competing product.

It may control the ranking system:

Independent Seller A → Rank 25
Independent Seller B → Rank 31
Marketplace's own product → Rank 1

The competition-law question is whether the ranking is based on legitimate factors or whether the marketplace is using its dominant position to disadvantage competitors.

This is closely connected with self-preferencing.

9. Google Shopping Case

Google and Alphabet v Commission, Case T-612/17 (General Court, 2021)

This is one of the most important cases for machine-curated digital marketplaces.

The European Commission found that Google had given favorable positioning and display to its comparison-shopping service while competing comparison-shopping services were subject to less favorable treatment.

The General Court upheld the central finding of abuse.

Importance

The case illustrates that a dominant digital platform's control over search visibility and ranking can have competitive significance.

Machine-curation relevance

A future AI marketplace could similarly control:

• product ranking;
• AI recommendations;
• seller visibility;
• autonomous purchasing suggestions.

The competition concern would depend upon the specific conduct and market circumstances.

10. United States v Microsoft

United States v Microsoft Corp., 253 F.3d 34 (D.C. Cir. 2001)

Microsoft's conduct concerning its dominant operating-system platform remains an important example of platform power and exclusionary strategies.

The case demonstrated how control over an important technological platform can influence competition in adjacent markets.

Relevance

A machine-curated marketplace can become a similar gateway:

Consumers → algorithm → products → sellers

If the gateway is controlled by a powerful undertaking, control over access and visibility can influence competition downstream.

11. United Brands v Commission

United Brands Company v Commission, Case 27/76 (1978)

This foundational European case established important principles concerning dominance and abusive conduct.

The case demonstrates that an undertaking with substantial market power has greater responsibilities regarding conduct that can affect competitive conditions.

Application

A machine-curated marketplace with substantial market power may face greater scrutiny when its algorithms:

• discriminate against rivals;
• impose exclusionary conditions;
• restrict market access;
• exploit dependent trading partners.

Dominance itself, however, is not unlawful.

12. Hoffmann-La Roche v Commission

Hoffmann-La Roche & Co. AG v Commission, Case 85/76 (1979)

This is a leading European case on abuse of dominance, particularly loyalty-inducing practices.

The Court examined arrangements capable of tying customers to a dominant undertaking.

Machine-marketplace relevance

Algorithmic loyalty mechanisms can potentially produce similar effects.

For example, a dominant marketplace could design an algorithm that gives substantial advantages to sellers who:

• use its logistics;
• use its payment service;
• purchase advertising;
• accept exclusivity;
• avoid competing platforms.

If such arrangements foreclose competitors, competition concerns may arise.

13. Intel v Commission

Intel Corp. v Commission, Case C-413/14 P (2017)

Intel concerned conditional rebates and the assessment of their potential exclusionary effects.

The judgment emphasized the importance of examining the economic circumstances and effects of rebate practices where appropriate.

Machine-curated marketplace application

An algorithm might automatically provide preferential visibility to sellers meeting particular conditions.

For example:

“Preferred sellers” receive higher rankings if they purchase the marketplace's advertising services.

Such practices may require examination of:

• coverage;
• duration;
• market position;
• foreclosure effects;
• alternative channels;
• efficiencies.

The presence of an algorithm does not change the fundamental competition-law inquiry.

14. Amazon Marketplace Concerns

Amazon's marketplace model provides an important real-world example of the issues that can arise when a platform simultaneously operates as:

1. marketplace operator;
2. seller;
3. logistics provider;
4. advertiser;
5. data collector;
6. product recommender.

Competition authorities have investigated various aspects of Amazon's marketplace conduct in different jurisdictions.

Potential concerns have included:

• use of seller data;
• treatment of competing sellers;
• marketplace access;
• ranking;
• advertising;
• preferential treatment;
• competition between Amazon and independent sellers.

These should be distinguished from a final judicial finding of unlawful conduct in any particular proceeding.

15. Self-Preferencing

Self-preferencing occurs when a platform gives its own products or services preferential treatment over competing products.

In machine-curated marketplaces, this could occur through:

• ranking;
• recommendations;
• search results;
• default placement;
• AI-generated answers;
• product badges;
• automated bundles.

For example:

Search query → AI recommendation → platform's own product

The legal issue is not simply that the platform recommends its own product.

Authorities may need to examine:

1. dominance;
2. market definition;
3. discriminatory treatment;
4. foreclosure;
5. effects on competition;
6. objective justification;
7. efficiencies.

16. Algorithmic Discrimination Between Sellers

A machine-curated marketplace may rank sellers differently.

This can be legitimate.

For example, ranking could depend upon:

• quality;
• delivery performance;
• customer satisfaction;
• price;
• reliability.

But discriminatory ranking may become problematic if the dominant platform uses opaque or manipulated criteria to disadvantage rivals.

Possible concerns include:

• artificial demotion;
• unexplained exclusion;
• discriminatory access;
• retaliation against sellers;
• preferential treatment for affiliated companies.

17. Algorithmic Exclusion

A marketplace could potentially use automated systems to exclude competitors through:

A. De-ranking

Competitors receive reduced visibility.

B. De-listing

Competitors are removed from search results.

C. Algorithmic demotion

Rivals are systematically placed below the platform's own products.

D. Data restrictions

Competitors cannot access necessary information.

E. API restrictions

Competitors cannot integrate effectively.

F. Automated discrimination

Algorithms apply different conditions to different sellers.

18. Loyalty and Machine Curation

Algorithms can create sophisticated loyalty systems.

Instead of traditional loyalty contracts, a marketplace can use:

customer behaviour → algorithmic profile → personalized offer → increased dependence.

For example, customers who purchase frequently through one marketplace may receive:

• lower prices;
• faster delivery;
• exclusive products;
• personalized recommendations.

Such loyalty mechanisms are not automatically unlawful.

The competition concern increases where the system is used by a dominant undertaking to exclude competing platforms.

19. Customer Lock-In

Machine curation can create substantial switching costs.

Consumers may accumulate:

• purchasing histories;
• preferences;
• saved payment details;
• loyalty benefits;
• personalized recommendations.

Sellers may accumulate:

• customer ratings;
• transaction histories;
• platform reputation;
• advertising data;
• fulfilment integration.

The greater the switching cost, the more difficult it may be for competitors to attract users.

20. Data Advantages and Market Power

A machine-curated marketplace can have a major information advantage.

Suppose the platform observes:

Search → click → purchase → return → review → repeat purchase.

The platform can use this information to improve its algorithm.

A smaller competitor may not have comparable data.

This may create a data-driven competitive feedback loop.

However:

Large data holdings alone do not prove unlawful market power or abuse.

The legal assessment must examine whether the data advantage actually creates or reinforces significant competitive barriers.

21. Vertical Integration

Machine-curated marketplaces may be vertically integrated.

For example:

Marketplace → seller → payment → logistics → advertising → cloud → AI

This can create potential foreclosure concerns.

A marketplace could theoretically disadvantage competing sellers by:

• increasing their fees;
• limiting visibility;
• restricting data;
• withholding advertising opportunities;
• offering better conditions to affiliated sellers.

Again, the legality depends on the applicable law and factual evidence.

22. Algorithmic Pricing

Machine-curated marketplaces often combine recommendation systems with automated pricing.

This creates another competition concern.

Algorithms may independently adjust prices based on:

• demand;
• inventory;
• competitor prices;
• consumer behaviour.

Independent algorithmic pricing is not automatically a cartel.

However, competition concerns may arise where firms use algorithms as part of an agreement or coordinated strategy to stabilize prices.

The crucial distinction is:

independent algorithmic adaptation ≠ automatically unlawful coordination.

23. Algorithmic Collusion

Machine-curated marketplaces may make coordination easier because algorithms can:

• monitor competitors continuously;
• detect price changes instantly;
• respond automatically;
• punish deviations;
• exchange market information.

Competition authorities therefore increasingly need to consider whether algorithmic systems facilitate coordinated conduct.

Traditional cartel principles, however, generally require an appropriate legal basis for finding coordination; parallel algorithmic behaviour alone should not automatically be treated as proof of an agreement.

24. Ranking Transparency

Another issue concerns algorithmic opacity.

If sellers do not understand:

• why they are ranked;
• why they are demoted;
• how recommendations are generated;

they may find it difficult to compete effectively.

Competition law may therefore intersect with:

• platform regulation;
• consumer protection;
• digital-services regulation;
• transparency requirements.

But transparency obligations must be carefully designed because complete disclosure of algorithms could expose trade secrets and facilitate gaming.

25. Machine-Curated Marketplaces and Merger Control

Suppose a dominant marketplace acquires:

• a recommendation engine;
• a leading AI shopping assistant;
• a major seller;
• a logistics platform;
• an emerging competitor.

Even where the target has limited current market share, authorities may consider:

• future competition;
• innovation;
• data assets;
• network effects;
• ecosystem expansion;
• potential competition.

This is sometimes described in policy discussions as the “killer acquisition” concern.

But acquisition of an innovative company is not inherently anticompetitive.

26. Competition for Consumer Attention

Machine-curated marketplaces compete not only for purchases but also for consumer attention.

The algorithm controls:

what consumers see → what they consider → what they buy.

This creates an important distinction between:

Product market power

Ability to influence product prices or conditions.

and

Discovery/gateway power

Ability to determine which products consumers discover.

The latter can become increasingly important in AI-driven commerce.

27. AI Shopping Agents

Future marketplaces may involve AI agents that independently:

1. understand consumer preferences;
2. search marketplaces;
3. compare prices;
4. negotiate;
5. select products;
6. execute payments;
7. arrange delivery.

This changes the marketplace structure.

Instead of:

Human → search → marketplace → seller

the model may become:

Human → AI agent → machine-curated marketplaces → automated transaction.

The entity controlling the AI recommendation or purchasing interface could become an important competitive gateway.

28. Market Power Through AI Gatekeeping

A dominant AI marketplace could potentially influence:

• which suppliers are recommended;
• which products are considered;
• which sellers are excluded;
• which prices are shown;
• which payment methods are preferred.

This creates potential AI gatekeeper power.

Competition analysis may therefore need to consider:

• algorithmic neutrality;
• access conditions;
• ranking;
• interoperability;
• data portability;
• self-preferencing;
• switching costs.

29. Objective Justification

Algorithmic differentiation is not necessarily unlawful.

A marketplace may have legitimate reasons to rank one product above another, such as:

• quality;
• safety;
• reliability;
• delivery performance;
• fraud prevention;
• customer satisfaction;
• inventory availability.

Therefore, competition analysis must distinguish:

legitimate algorithmic curation

from

algorithmic exclusion.

30. Consumer Welfare and Innovation

Machine curation can generate substantial efficiencies.

Benefits

• reduced search costs;
• personalized recommendations;
• better product matching;
• reduced transaction costs;
• improved logistics;
• fraud detection;
• lower prices;
• increased product discovery.

Risks

• foreclosure;
• reduced choice;
• seller dependence;
• market concentration;
• discriminatory ranking;
• exploitation of data;
• algorithmic collusion;
• reduced innovation.

Competition law should consider both sides.

31. Essential-Facilities Considerations

A marketplace may become sufficiently important that competitors argue they need access to its infrastructure.

However, the doctrine must be applied cautiously.

Bronner principle

In:

Oscar Bronner GmbH & Co. KG v Mediaprint, Case C-7/97 (1998)

the European Court adopted a demanding approach to compulsory access.

Machine marketplace application

A competitor cannot normally argue:

“Your marketplace is successful, therefore you must provide us access.”

The competitor generally needs to demonstrate the legally relevant conditions for compulsory access.

32. Aspen Skiing

Aspen Skiing Co. v Aspen Highlands Skiing Corp., 472 U.S. 585 (1985)

This case involved a dominant firm's termination of a previously profitable cooperative arrangement.

It is important because it demonstrates that refusal to cooperate can, in exceptional circumstances, form part of exclusionary conduct.

Marketplace relevance

If a dominant marketplace:

• previously allowed a competitor meaningful access;
• benefited from the arrangement;
• deliberately terminates it;
• refuses to deal without legitimate justification;
• and thereby harms competition,

the circumstances could raise concerns.

The case should not be interpreted as creating a general duty to cooperate with competitors.

33. Trinko

Verizon Communications Inc. v Trinko, 540 U.S. 398 (2004)

The U.S. Supreme Court emphasized caution concerning compulsory cooperation.

Importance

Antitrust law generally does not require a company to help its competitors merely because cooperation would make competition easier.

Machine marketplace relevance

A marketplace should normally retain the ability to:

• develop its own algorithm;
• choose its own technology;
• improve its recommendation system;
• protect proprietary information.

Intervention becomes more plausible when there is established exclusionary conduct rather than merely a refusal to assist rivals.

34. Remedies

Where unlawful conduct is established, possible remedies include:

1. Non-discrimination

Require comparable sellers to receive comparable treatment.

2. Ranking safeguards

Prevent manipulation of rankings to disadvantage rivals.

3. Data portability

Allow users to transfer relevant information.

4. API access

Require access where legally justified.

5. Interoperability

Permit competing systems to interact.

6. Algorithmic auditing

Independent auditing may be required in appropriate circumstances.

7. Behavioural commitments

The platform may be prohibited from particular exclusionary practices.

8. Structural remedies

In exceptional cases, separation of conflicting business functions may be considered.

35. Challenges for Competition Authorities

Machine-curated marketplaces create several enforcement difficulties.

A. Algorithmic opacity

Authorities may not know precisely how ranking works.

B. Rapid changes

Machine-learning systems can change continuously.

C. Multi-sided markets

Effects occur simultaneously among consumers, sellers and advertisers.

D. Dynamic competition

Current market shares may not capture future competitive threats.

E. Data advantages

It can be difficult to determine whether data creates genuine market power.

F. International operations

Platforms operate across multiple jurisdictions.

G. AI decision-making

Algorithms may make commercially significant decisions with limited human intervention.

36. UAE Perspective

Machine-curated marketplaces are particularly relevant to the UAE's expanding:

• e-commerce;
• fintech;
• logistics;
• AI;
• smart-city;
• cloud-computing;
• digital-platform;
• automated-commerce sectors.

Under UAE competition-law principles, potential concerns may arise where a dominant undertaking uses its market position to:

• exclude competitors;
• impose discriminatory conditions;
• restrict market access;
• abuse seller dependence;
• foreclose competing platforms.

Machine curation should therefore be assessed alongside traditional concepts of:

• relevant market;
• dominance;
• abuse;
• exclusion;
• foreclosure;
• consumer welfare;
• efficiencies.

AI technology does not eliminate traditional competition-law principles; it changes how market power may be exercised.

37. Key Distinction: Curation vs Abuse

It is important to remember:

Machine curation ≠ antitrust violation.

An algorithm may legitimately rank products based on:

• quality;
• relevance;
• price;
• reliability;
• consumer preference.

Competition concerns arise where the curation mechanism is used in a way that satisfies the applicable legal test for exclusionary or exploitative conduct.

38. Summary of Major Case Laws

39. Quick Revision Notes

Machine-curated marketplace:
A marketplace where algorithms or AI determine product visibility, ranking, recommendations or purchasing options.

Main competition concern:
Control over algorithmic curation may become control over consumer access and seller visibility.

Important concepts:

• market power;
• dominance;
• network effects;
• data advantages;
• ranking;
• self-preferencing;
• foreclosure;
• loyalty;
• switching costs;
• algorithmic pricing;
• algorithmic collusion;
• interoperability;
• data portability;
• essential facilities;
• vertical integration;
• merger control.

Important cases:

1. United Brands v Commission — dominance.
2. Hoffmann-La Roche v Commission — loyalty and exclusion.
3. Microsoft v Commission — technological foreclosure.
4. United States v Microsoft — platform power.
5. Intel v Commission — exclusionary incentives and effects.
6. Google Shopping — preferential treatment and digital ranking.
7. Bronner v Mediaprint — indispensable access.
8. Aspen Skiing — exceptional refusal to deal.
9. Trinko — limits on compulsory cooperation.
10. IMS Health — IP and competition.

Conclusion

Machine-curated marketplaces represent a major evolution in digital competition because the algorithm increasingly controls not only transactions but also consumer discovery itself.

A marketplace with substantial market power may potentially use its curation system to influence which sellers succeed, which products receive visibility, and which competing platforms can reach consumers. This can create concerns involving self-preferencing, foreclosure, loyalty, discriminatory access, data advantages, network effects, and algorithmic coordination.

However, algorithmic curation is not inherently anticompetitive. Effective competition-law analysis must distinguish legitimate innovation and personalization from conduct that unlawfully exploits or reinforces market power.

The central long-term principle is therefore:

Competition law should protect the competitive process surrounding algorithmic marketplaces without treating every successful algorithm, recommendation system, or proprietary technology as an antitrust violation.