Competition Law and Governance of Recommendation-Driven Markets

Introduction

Recommendation-driven markets are markets in which commercial visibility, consumer choice, transaction opportunities, or access to customers are significantly influenced by algorithmic recommendations. Examples include search engines, e-commerce marketplaces, app stores, streaming services, travel platforms, social-media feeds, food-delivery platforms, financial-product comparison services, and digital advertising systems.

In traditional markets, competition largely depends on price, quality, distribution, advertising and consumer preference. In recommendation-driven markets, an additional competitive variable becomes critical:

Who controls the mechanism that determines what the consumer sees first, what is suggested, and what remains practically invisible?

A platform may therefore possess market power even without directly fixing prices. Control over rankings, recommendations, personalised feeds, default suggestions, search results, product placement and automated matching can influence demand and consequently affect competitors.

Competition law must consequently address not merely market concentration, but also the governance of recommendation systems.

I. Meaning of Recommendation-Driven Markets

A recommendation-driven market exists where an intermediary or technological system uses data, algorithms or platform rules to determine or influence the products, services, sellers or content presented to users.

Typical examples

1. E-commerce
   • recommended products;
   • "frequently bought together";
   • sponsored rankings;
   • personalised search results.
2. Search engines
   • search-result ordering;
   • specialised search results;
   • local recommendations;
   • shopping recommendations.
3. App stores
   • app discovery;
   • featured applications;
   • recommended apps;
   • default placement.
4. Travel platforms
   • hotel rankings;
   • preferred-partner placement;
   • personalised accommodation recommendations.
5. Food-delivery platforms
   • restaurant rankings;
   • promoted restaurants;
   • personalised recommendations.
6. Digital advertising
   • advertiser recommendations;
   • automated targeting;
   • ranking of advertisements.
7. Streaming platforms
   • recommended films, programmes and music;
   • personalised playlists;
   • algorithmic promotion.
8. Financial platforms
   • recommended insurance, loans or investment products;
   • comparison rankings;
   • personalised financial offers.

II. Why Recommendation Systems Create Competition Concerns

Recommendation systems can transform visibility into market power.

A seller may technically remain available on a platform while being commercially disadvantaged because the algorithm rarely recommends its products.

This produces an important distinction:

Traditional exclusion

"The competitor is prohibited from entering."

Algorithmic exclusion

"The competitor remains technically present but is systematically deprived of visibility."

The second form may be considerably more difficult to detect.

III. Legal Framework

Recommendation-driven markets can engage several areas of competition law.

1. Abuse of Dominant Position

A dominant platform may abuse its position through:

• self-preferencing;
• discriminatory ranking;
• exclusionary recommendations;
• demotion of competitors;
• preferential treatment of affiliated products;
• tying;
• leveraging dominance from one market into another;
• discriminatory access to recommendation infrastructure.

The central question is generally whether the conduct distorts competitive conditions rather than merely improving the platform's own product or service.

IV. Self-Preferencing

One of the most important concerns is self-preferencing.

A vertically integrated platform may operate both:

• the marketplace or intermediary; and
• products competing on that marketplace.

The platform then controls the recommendation mechanism.

For example:

Platform → controls ranking → owns competing product → recommends its own product → rival receives less visibility.

This creates a potential conflict between the platform's role as market organiser and its role as competitor.

V. Case Law

1. Google Shopping — Google Search (Shopping), European Union

Case: Google and Alphabet v European Commission, C-48/22 P, judgment of the Court of Justice of the European Union, 10 September 2024.

This is one of the most significant authorities for recommendation and ranking-driven markets.

The European Commission had found that Google had given preferential positioning and display to its own comparison-shopping service while applying less favourable treatment to competing comparison-shopping services.

The EU courts ultimately upheld the finding of abuse.

Competition-law significance

The case demonstrates that a dominant digital platform can potentially distort competition through the architecture of visibility.

The relevant competitive harm does not necessarily require:

• an outright prohibition on competitors;
• exclusion from the platform;
• a traditional price squeeze.

Instead, differential treatment in a strategically important ranking system may affect competitive opportunities.

Principle

Where a dominant platform controls an important channel through which users discover products, systematic preferential treatment of its own competing service can raise Article 102 TFEU concerns.

Relevance to recommendation-driven markets

Google Shopping demonstrates that:

Ranking can be an economic resource.

The algorithmic position assigned to a service may substantially influence traffic and commercial opportunities.

2. Google Android — Google and Alphabet v Commission

Case: Google and Alphabet v European Commission, C-738/19 P, Court of Justice, 10 September 2024.

The case concerned Google's Android ecosystem and contractual arrangements involving search, browser and application distribution.

The broader competition issue involved Google's ability to use control over one part of a digital ecosystem to reinforce its position in related markets.

Relevance

Recommendation-driven ecosystems frequently contain multiple interconnected layers:

Operating system → app store → search → browser → recommendations → advertising.

A company controlling several layers may have the ability to influence which services users encounter.

Competition principle

Competition authorities may therefore examine whether contractual or technical arrangements cause:

• foreclosure;
• ecosystem leveraging;
• reduced opportunities for competing services;
• reinforcement of an existing dominant position.

Importance for recommendation markets

The case demonstrates why recommendation governance cannot always be examined in isolation.

The relevant competitive effects may occur across an interconnected ecosystem.

3. Google Search (AdSense) — Google and Commission

Case: Google and Alphabet v Commission, C-293/21 P, Court of Justice of the European Union, 10 September 2024.

The case concerned Google's conduct in online search advertising and contractual restrictions imposed on publishers.

Although not a pure recommendation case, it is important because advertising and recommendation systems increasingly overlap.

Competition relevance

A platform can influence competition by controlling:

• access to users;
• advertising placement;
• search traffic;
• publisher relationships;
• commercial visibility.

The case illustrates the importance of examining how a dominant intermediary's contractual arrangements affect competing intermediaries.

Broader principle

In recommendation-driven markets, competition may be harmed not merely through the recommendation itself but through restrictions surrounding the infrastructure that generates user traffic.

4. Booking.com — German Competition Authorities

The Booking.com proceedings in Germany provide an important example of platform governance in the hotel-booking market.

The German competition authorities examined Booking.com's best-price clauses, which restricted hotels' ability to offer different prices through other distribution channels.

The German Federal Court of Justice ultimately upheld the competition-law concerns regarding broad price-parity restrictions.

Relevance to recommendation-driven markets

Hotel platforms do not merely provide a neutral directory.

They commonly combine:

• rankings;
• reviews;
• recommendations;
• preferred-partner status;
• commissions;
• price information;
• visibility mechanisms.

A platform's contractual rules can therefore influence how accommodation providers compete for consumer attention.

Principle

Platform rules that appear commercially neutral may nevertheless affect competition when they influence the ability of suppliers to compete through alternative channels.

Importance

The case shows that recommendation governance and platform contractual governance are interconnected.

5. HRS — German Hotel Booking Platform Case

Case: HRS-Hotel Reservation Service, German Federal Cartel Office proceedings concerning narrow best-price clauses.

The Bundeskartellamt found that HRS's wide-ranging best-price clause restricted competition between hotel booking channels.

Competition relevance

Hotels depend heavily on platforms for:

• consumer discovery;
• rankings;
• booking traffic;
• reviews;
• recommendations.

Consequently, restrictions imposed by a dominant intermediary can influence the competitive environment surrounding recommendation systems.

Key lesson

A recommendation platform may exercise market influence through a combination of:

ranking + contractual restrictions + data + consumer traffic.

Competition analysis should therefore not consider algorithmic recommendations independently from platform rules.

6. Microsoft — Internet Explorer

Case: Microsoft v Commission, T-201/04, General Court of the European Union.

Although the case predates modern recommendation systems, it remains important for understanding digital ecosystem leveraging.

Microsoft was found to have abused its dominant position through conduct involving Internet Explorer and Windows.

Relevance to recommendation-driven markets

Modern recommendation ecosystems can similarly use a dominant product as a gateway to related products.

For example:

Operating system → default application → recommendation → user adoption → data accumulation → stronger platform position.

Competition principle

The case demonstrates why competition law examines whether control over an important digital gateway is used to strengthen a position in an adjacent market.

7. United States v Google — Search and Search Advertising

Case: United States et al. v Google LLC, U.S. District Court for the District of Columbia, 2024 liability judgment.

The U.S. litigation concerned Google's conduct relating to search distribution and default arrangements.

Although the case is not a pure algorithmic-recommendation case, it is highly relevant to the governance of consumer attention and search access.

Competition significance

Search markets involve an important competitive relationship:

Default access → user queries → search results → commercial visibility → advertising revenue.

A platform controlling the gateway through which users reach recommendations and search results can influence the competitive environment.

Relevance

Recommendation-driven markets demonstrate that competition can be affected before a consumer even sees the recommendation.

The distribution mechanism itself can determine which recommendation environment receives user traffic.

VI. Recommendation Algorithms as Competitive Infrastructure

Recommendation systems should increasingly be understood as a form of digital infrastructure.

Traditional infrastructure includes:

• ports;
• railways;
• electricity grids;
• telecommunications networks.

Digital infrastructure increasingly includes:

• search engines;
• app stores;
• marketplaces;
• ranking systems;
• recommendation engines;
• identity systems;
• payment systems.

Where a recommendation system becomes indispensable for commercial visibility, access to that system may acquire essential competitive importance.

VII. Algorithmic Discrimination

A recommendation algorithm can treat similarly situated businesses differently.

Potential discriminatory factors include:

• commission payments;
• advertising expenditure;
• platform ownership;
• preferred-partner status;
• data contribution;
• contractual arrangements;
• algorithmic quality scores.

Competition authorities may therefore investigate whether supposedly neutral ranking criteria conceal commercially discriminatory treatment.

VIII. Self-Preferencing and Vertical Integration

Self-preferencing becomes especially significant where the platform has a dual role.

Structure

Platform

↓

Controls recommendation algorithm

↓

Controls consumer data

↓

Controls ranking

↓

Competes with third-party sellers

This produces an inherent competition concern.

The platform may have an incentive to recommend its own products even where competing products might otherwise receive greater visibility.

However, competition law must distinguish between:

Legitimate product improvement

For example:

• better quality;
• better delivery;
• lower prices;
• improved customer satisfaction.

and

Potential exclusionary conduct

For example:

• artificially elevating affiliated products;
• systematically demoting competitors;
• manipulating ranking criteria;
• denying comparable access to recommendation mechanisms.

IX. Data as an Input to Recommendations

Recommendation systems depend heavily upon data.

Relevant data may include:

• search histories;
• purchases;
• clicks;
• reviews;
• location;
• browsing behaviour;
• transaction history;
• seller performance;
• customer preferences.

The platform can use this information to improve recommendations.

But a dominant platform may also have the ability to use information generated by third-party competitors to strengthen its own competing products.

This creates a potential:

data accumulation → recommendation improvement → user engagement → additional data accumulation

cycle.

Such feedback loops can reinforce market power.

X. Network Effects

Recommendation-driven markets often have strong network effects.

More users produce more behavioural data.

More data can produce better recommendations.

Better recommendations can attract more users.

More users attract more sellers.

More sellers generate more transactions.

More transactions generate more data.

This produces a self-reinforcing cycle:

Users → Data → Better Recommendations → More Users → More Sellers → More Data

Competition authorities may therefore examine whether a dominant firm can use this cycle to make entry or expansion increasingly difficult.

XI. Personalisation and Competition

Personalisation creates another competition issue.

Two consumers may receive completely different recommendations for the same query.

This makes traditional market transparency more difficult.

For example:

Consumer A:

Product X → Product Y → Product Z

Consumer B:

Product Z → Product Q → Product X

The platform may therefore exercise considerable influence without imposing an obvious uniform ranking.

Competition authorities may need to examine:

• personalised ranking;
• user segmentation;
• recommendation criteria;
• experimentation;
• dynamic pricing;
• sponsored placement;
• behavioural targeting.

XII. Dark Patterns and Recommendation Governance

Competition concerns can overlap with consumer-protection concerns.

Examples include:

• artificially urgent recommendations;
• default selections;
• hidden sponsored recommendations;
• misleading "best choice" labels;
• repeated recommendations;
• friction against alternative providers;
• personalised nudging.

Not every dark pattern constitutes an antitrust violation.

However, where such mechanisms reinforce market power or exclude competitors, competition law and consumer law may become complementary.

XIII. Transparency and Explainability

A major regulatory problem is that recommendation algorithms can be extremely complex.

A competition authority may need to determine:

1. Why was a product recommended?
2. Why was another product demoted?
3. Were competitors treated differently?
4. Was the platform's own product favoured?
5. Did advertising payments affect ranking?
6. Did commercial agreements affect recommendations?
7. Were consumers segmented?
8. Did the algorithm change after competitors entered?

Therefore, algorithmic governance increasingly requires:

• audit trails;
• ranking logs;
• model documentation;
• testing records;
• access to relevant data;
• explanation of ranking variables;
• monitoring of algorithmic changes.

XIV. Algorithmic Collusion

Recommendation-driven systems can also facilitate coordination between competitors.

Algorithms may monitor:

• prices;
• inventory;
• competitors' recommendations;
• demand;
• consumer behaviour.

If competing algorithms repeatedly respond to each other's conduct, there may be a risk of:

• tacit coordination;
• algorithmic price alignment;
• reduced competitive experimentation.

Competition law therefore needs to distinguish between:

Independent algorithmic optimisation

and

Algorithmically facilitated concerted conduct.

XV. Ranking Manipulation

Ranking manipulation occurs when a platform deliberately changes the order of recommendations to produce a commercial outcome.

Potential examples include:

• pushing platform-owned products upward;
• demoting independent sellers;
• favouring high-commission products;
• rewarding exclusive suppliers;
• penalising sellers using competing platforms.

The legal assessment depends upon factors such as:

• market power;
• intent;
• effects;
• legitimate business justification;
• discriminatory treatment;
• foreclosure;
• duration;
• importance of the platform.

XVI. Governance Models

Competition-law governance of recommendation-driven markets can employ several mechanisms.

1. Ex Post Antitrust Enforcement

Authorities investigate conduct after suspected harm occurs.

Advantages:

• case-specific;
• flexible;
• based on evidence.

Limitations:

• investigations can be lengthy;
• algorithmic markets can change rapidly.

2. Ex Ante Regulation

Certain large digital platforms may be subject to obligations before competition harm occurs.

Potential obligations include:

• non-discrimination;
• transparency;
• interoperability;
• restrictions on self-preferencing;
• data portability;
• auditing.

The EU Digital Markets Act represents an important example of this broader regulatory approach.

3. Algorithmic Auditing

Authorities may require platforms to demonstrate that recommendation systems comply with competition obligations.

Audits can examine:

• ranking outcomes;
• treatment of rivals;
• self-preferencing;
• sponsored recommendations;
• data use;
• algorithmic changes.

4. Data Access and Portability

Competition may be improved where users and businesses can move data between platforms.

This can reduce switching costs and make entry easier.

5. Interoperability

Interoperability can prevent a dominant recommendation ecosystem from becoming completely closed.

For example:

Platform A → compatible with competing services

rather than:

Platform A → closed ecosystem → users cannot effectively migrate.

XVII. Remedies

Competition authorities can potentially employ:

Structural remedies

• divestiture;
• separation of business units.

Behavioural remedies

• non-discrimination;
• ranking neutrality;
• transparency obligations;
• access obligations;
• prohibition of self-preferencing.

Technical remedies

• interoperability;
• data portability;
• API access;
• algorithmic monitoring.

Contractual remedies

• removal of restrictive parity clauses;
• prohibition of exclusivity;
• restrictions on discriminatory contractual conditions.

XVIII. Challenges in Competition-Law Enforcement

1. Black-box algorithms

Authorities may not know precisely why an algorithm generated a particular recommendation.

2. Constant algorithmic modification

The relevant algorithm may change before an investigation concludes.

3. Legitimate optimisation

A platform may argue that ranking changes improve:

• relevance;
• quality;
• consumer welfare;
• fraud prevention;
• user experience.

4. Measurement difficulties

It can be difficult to determine how much a ranking change affected:

• sales;
• traffic;
• entry;
• prices;
• innovation.

5. Multi-sided markets

The platform simultaneously interacts with:

• consumers;
• sellers;
• advertisers;
• developers;
• service providers.

Effects on one side may produce benefits or harms on another.

XIX. Relationship Between Competition Law and AI

AI substantially increases the importance of recommendation governance.

AI systems can generate recommendations regarding:

• products;
• investments;
• insurance;
• medical services;
• travel;
• employment;
• news;
• entertainment.

The system may learn from enormous quantities of behavioural data.

Consequently, future competition disputes may increasingly involve:

Who controls the model?

Who controls the training data?

Who controls the recommendation interface?

Who determines which competitors the AI presents to consumers?

Can rivals obtain meaningful access to the recommendation ecosystem?

XX. Six Core Competition-Law Questions

When analysing a recommendation-driven market, authorities should examine:

1. Who controls the recommendation mechanism?

Is it a dominant intermediary, an independent service or a vertically integrated firm?

2. What determines ranking?

Are recommendations based on:

• relevance;
• quality;
• price;
• consumer behaviour;
• advertising;
• commission;
• affiliation?

3. Are competitors treated equally?

Similar businesses should be examined for differential treatment.

4. Does the platform compete with recommended businesses?

Vertical integration substantially changes the competitive analysis.

5. Can competitors effectively reach consumers without the platform?

If not, the recommendation system may constitute a strategically important competitive gateway.

6. Does the conduct create foreclosure?

The ultimate competition concern is whether rivals are prevented from competing effectively.

XXI. Conceptual Model

The governance structure can be represented as:

DATA

↓

ALGORITHMIC PROCESSING

↓

PERSONALISATION

↓

RECOMMENDATION

↓

CONSUMER ATTENTION

↓

TRANSACTIONS

↓

REVENUE

↓

MORE DATA

This creates a feedback loop:

More transactions → more data → better recommendations → greater consumer engagement → more transactions.

Where one firm controls the entire loop, competition authorities may need to examine whether the system creates durable barriers to entry.

XXII. Key Case-Law Principles

Conclusion

Competition law in recommendation-driven markets is increasingly concerned with a shift from control over products and prices to control over visibility and consumer attention.

The central competition-law problem is not simply whether a platform sells its own product. It is whether a platform that controls the recommendation infrastructure uses that control to distort the competitive process.

The major issues are therefore:

• self-preferencing;
• discriminatory ranking;
• algorithmic foreclosure;
• data advantages;
• personalised recommendations;
• platform parity restrictions;
• ecosystem leveraging;
• algorithmic coordination;
• lack of transparency;
• interoperability;
• access to recommendation infrastructure.

The Google Shopping litigation is particularly important because it demonstrates that the manner in which a dominant platform positions and presents competing services can itself become a central competition-law issue.

The future governance of recommendation-driven markets is consequently likely to combine traditional abuse-of-dominance principles, merger control, platform regulation, algorithmic auditing, data governance, interoperability and ex-ante digital regulation.