Competition Law and Semantic Network Concentration Concerns

1. Introduction

Semantic network concentration refers to a situation in which competition becomes concentrated not merely because one undertaking controls a traditional market, but because it controls the connections, meanings, classifications, data relationships, interoperability layers, search structures, knowledge graphs, APIs, recommendation systems, or other semantic links through which users and businesses discover and interact with products and services.

The concept is particularly relevant to digital ecosystems, search engines, AI systems, online marketplaces, social networks, cloud platforms, operating systems, app stores and data-driven services.

It is important to clarify that “semantic network concentration” is not, by itself, a separate statutory offence in most competition-law systems. It is an analytical concept that can help identify conventional competition-law problems such as:

• abuse of dominance;
• self-preferencing;
• discriminatory access;
• refusal to interoperate;
• tying and bundling;
• leveraging;
• foreclosure;
• exclusionary data practices;
• excessive dependence on a platform;
• network-effect-driven entrenchment;
• manipulation of rankings or classifications; and
• acquisitions that reinforce an already concentrated ecosystem.

Modern competition analysis increasingly recognises that digital ecosystems can consist of interconnected products and services whose competitive significance cannot always be understood by examining each product in isolation.

2. Meaning of a Semantic Network

A semantic network is a network in which information is connected according to relationships or meanings.

For example, a digital search or AI platform may internally connect:

user → query → concept → entity → product → seller → advertisement → recommendation → transaction.

Similarly, an e-commerce ecosystem may connect:

consumer → search term → category → ranking → seller → product → review → payment → logistics.

The undertaking controlling these relationships may therefore possess influence over how market participants are discovered, classified, ranked and connected.

Competition-law significance

A company may not need to manufacture the largest number of products to exercise substantial market power.

It may instead control the architecture through which competitors reach consumers.

This creates a distinction between:

Traditional concentration

Market share → market power → competitive effects

and:

Semantic/network concentration

Data + classification + interoperability + network effects + user relationships + ranking + ecosystem integration → cumulative market power

3. How Semantic Network Concentration Develops

A. Data accumulation

A platform collects information about:

• users;
• searches;
• purchases;
• preferences;
• clicks;
• locations;
• product relationships;
• suppliers;
• competitors; and
• consumer behaviour.

The accumulated data can improve the platform's semantic understanding of the market.

B. Network effects

The more users a platform has, the more attractive it may become to sellers.

The more sellers it has, the more attractive it becomes to users.

This creates a feedback loop:

Users ↑ → sellers ↑ → data ↑ → service quality ↑ → users ↑

Such indirect network effects were expressly important in the assessment of Google's Play Store position in India.

C. Semantic feedback loops

A more sophisticated feedback loop can arise:

More users → more queries → more behavioural data → better classification → better recommendations → more users

The resulting advantage may become difficult for new competitors to replicate.

D. Interoperability control

An undertaking may control the technical interface connecting different parts of the ecosystem.

Examples include:

• APIs;
• operating-system interfaces;
• app-store interfaces;
• identity systems;
• payment interfaces;
• search interfaces;
• data portability mechanisms;
• cloud interfaces.

Refusal to provide interoperability can therefore become a competition concern where the interface is strategically important.

4. Major Competition Concerns

4.1 Market Definition Problems

Traditional market definition normally examines substitutability between products.

Semantic networks complicate this because several apparently separate products may be interconnected.

For example:

Operating system → app store → payment → browser → search → advertising

may function as an integrated competitive system.

The European Commission and EU courts have increasingly considered ecosystem characteristics, including network effects, interconnection and complementarity, when analysing digital markets.

Therefore, competition authorities may need to examine:

• individual markets;
• adjacent markets;
• multi-sided markets;
• aftermarkets;
• ecosystem relationships; and
• competition between ecosystems.

5. Network Effects and Concentration

Semantic concentration is especially problematic where network effects are strong.

Suppose Platform A has:

• 70 million users;
• 2 million businesses;
• enormous historical search data; and
• an established recommendation system.

A new entrant with technically superior software may nevertheless have difficulty competing because it lacks the existing network.

This creates dynamic entry barriers.

The relevant question becomes not merely:

“Can another company technically enter?”

but also:

“Can another company realistically reproduce the network, data and semantic relationships necessary to compete?”

6. Self-Preferencing

A platform may control both:

1. the infrastructure through which products are discovered; and
2. products competing within that infrastructure.

It may then preferentially position its own products.

For example:

Search engine → ranking algorithm → platform's own service

If the platform gives its own service preferential visibility, competitors may suffer even without an explicit contractual exclusion.

This was central to Google Shopping.

7. Six Major Case Laws

1. Google Search (Shopping) — Google and Alphabet v Commission, T-612/17

Facts

Google operated a general search engine while also operating a comparison-shopping service.

The European Commission found that Google had systematically given prominent placement to its own comparison-shopping service while demoting competing comparison-shopping services in its general search results.

The General Court upheld the Commission's decision in substantial part.

Competition-law principle

The case demonstrates that a dominant undertaking controlling a major information-discovery network may create competitive concerns by manipulating the conditions under which competing services are discovered.

Relevance to semantic networks

Search results are fundamentally semantic:

query → interpretation → classification → ranking → result

Consequently, control over search relevance can become control over competitive visibility.

Principle

Control over the information-discovery layer can influence competition in adjacent markets.

This makes Google Shopping particularly relevant to AI search, knowledge graphs and semantic ranking systems.

2. Google Android — Google and Alphabet v Commission, T-604/18

Facts

The European Commission found several forms of abusive conduct involving Google's Android ecosystem, including restrictions concerning search, browser distribution and Android fragmentation.

The General Court substantially upheld the Commission's findings, while modifying the fine.

Competition-law principle

A dominant position in one layer of a digital ecosystem can potentially be leveraged into adjacent markets.

The ecosystem consisted of interconnected components such as:

• Android;
• Google Play;
• Google Search;
• Chrome;
• device manufacturers;
• application developers; and
• consumers.

Relevance

The case illustrates ecosystem concentration.

Control over one essential gateway may reinforce concentration across other interconnected layers.

The Court recognised the importance of the interconnected structure of Google's products and services.

Principle

A dominant digital infrastructure can generate leverage into neighbouring markets when contractual or technical arrangements reinforce the ecosystem.

3. Microsoft v Commission — T-201/04

Facts

Microsoft was found to have abused its dominant position by restricting interoperability information necessary for competing work-group server operating systems to achieve effective interoperability with Windows PCs and servers.

The case also involved tying concerning Windows Media Player.

Competition-law principle

Interoperability can become a competition-law issue where a dominant undertaking controls an important technical interface.

Semantic-network relevance

Modern semantic systems frequently depend upon:

• APIs;
• protocols;
• schemas;
• metadata;
• technical documentation;
• interoperability standards.

If access to such interfaces is withheld, competitors may be unable to participate effectively in the surrounding network.

Principle

Control of a technically indispensable interface can reinforce concentration in an interconnected ecosystem.

The Microsoft decision remains one of the foundational EU precedents concerning interoperability and refusal to supply/access information.

4. IMS Health v Commission — C-418/01

Facts

IMS Health developed a sophisticated system for organising pharmaceutical sales data into geographical segments.

Competitors sought access to the structure because it had become an important industry standard.

The case concerned whether refusal to license intellectual property could constitute abuse of dominance.

Competition-law principle

The Court established stringent conditions for compelling access to intellectual property under Article 102 TFEU.

The exceptional circumstances included considerations such as:

• indispensability;
• elimination of effective competition;
• absence of objective justification; and
• prevention of the emergence of a new product for which consumer demand exists.

Semantic-network relevance

IMS Health is particularly important because the dispute concerned the architecture used to organise and interpret market information.

In modern terminology, a proprietary:

• taxonomy;
• database structure;
• ontology;
• data schema; or
• knowledge architecture

could potentially perform a similar competitive function.

Principle

Competition law does not automatically require sharing of proprietary information architecture, but exceptional circumstances can justify intervention.

5. Magill — Joined Cases C-241/91 P and C-242/91 P

Facts

Television broadcasters possessed copyright over their programme listings.

Magill sought to publish comprehensive television programme information.

The Court recognised exceptional circumstances under which refusal to license intellectual property could amount to abuse.

Competition-law principle

The case established an important foundation for the exceptional-circumstances approach to compulsory access.

Relevance to semantic networks

Programme listings were not simply raw information; they were structured information capable of being transformed into a commercially valuable information product.

The case therefore has conceptual relevance to:

• proprietary databases;
• information aggregation;
• structured datasets;
• semantic classification; and
• information intermediaries.

Principle

Intellectual-property rights cannot automatically be used to exclude competition where exceptional circumstances satisfying the Article 102 framework are present.

6. Bronner v Mediaprint — C-7/97

Facts

Bronner sought access to Mediaprint's newspaper home-delivery system.

The Court rejected the claim because the strict conditions required for an obligation to provide access were not satisfied.

Competition-law principle

A dominant undertaking does not automatically have to provide competitors access to every facility that would make competition easier.

The facility must be genuinely indispensable under the applicable test.

Relevance

This principle provides an important limitation on semantic-network theories.

A competitor cannot simply argue:

“Your database/API/semantic system would help me compete, therefore you must provide access.”

The legal threshold is considerably higher.

Principle

Useful infrastructure is not necessarily indispensable infrastructure.

This protects incentives for investment while preventing strategically important bottlenecks from being used abusively.

7. Alphabet / Android Auto — Case C-233/23

This more recent case is especially relevant to semantic interoperability.

Facts

Enel X developed an electric-vehicle charging application and sought interoperability with Google's Android Auto platform.

Google initially refused interoperability.

The Italian competition authority considered the refusal abusive, and the matter reached the Court of Justice.

The Court held that continued activity by competitors did not automatically establish that the refusal could not have anticompetitive effects. The assessment must consider all relevant circumstances.

Significance

This case shows the movement from traditional infrastructure toward digital interface infrastructure.

Android Auto functions as an intermediary layer connecting:

vehicle → operating system → application → consumer

Control over that interface can therefore affect competitive access.

Principle

Digital interoperability can be competitively significant even where rival applications remain active in the underlying market.

8. Meta / Facebook Marketplace

The European Commission's Facebook Marketplace case provides another important ecosystem example.

Meta's social-networking ecosystem could potentially provide advantages to Marketplace because the platform already possessed a large user network.

The Commission examined the relationship between Facebook's social-networking services and Marketplace and considered the possibility of leveraging advantages from one service into another.

Relevance

The case demonstrates how:

existing network → user access → data → complementary service → competitive advantage

can create ecosystem concentration.

9. CCI — Umar Javeed & Others v Google LLC

The Indian example is particularly useful.

The Competition Commission of India considered Google's position in the market for app stores for Android mobile operating systems.

The CCI placed importance on factors including:

• indirect network effects;
• Google's large user base;
• application developers' dependence on Play Store;
• integration with Google Play Services;
• lack of substitutability; and
• entry barriers. 

Relevance

This illustrates how Indian competition law can analyse ecosystem-based market power, rather than relying exclusively on conventional market-share analysis.

10. Comparative Case Table

11. Semantic Concentration and Self-Preferencing

One of the most important risks is semantic self-preferencing.

Imagine an AI search engine that controls:

• the underlying knowledge graph;
• entity classification;
• search ranking;
• recommendation engine;
• advertising;
• its own competing products.

The platform could theoretically structure the semantic system so that its own products receive:

• better categorisation;
• higher relevance scores;
• greater visibility;
• more prominent recommendations;
• richer structured results; or
• preferential access to users.

The competition issue would not necessarily be a simple contractual exclusion.

It could arise from the architecture of the information system itself.

12. Semantic Data Advantages

Large semantic networks can generate a data advantage.

For example:

100 million searches
↓
billions of relationships
↓
improved entity recognition
↓
improved predictions
↓
improved recommendations
↓
more users
↓
more searches

This produces a potentially self-reinforcing competitive advantage.

Data can therefore function simultaneously as:

• an input;
• a competitive asset;
• a source of network effects; and
• a barrier to entry.

Academic literature similarly identifies data collection and processing as important sources of market power within digital ecosystems.

13. Semantic Interoperability as a Competition Issue

Interoperability may involve more than merely connecting two technical systems.

There may be semantic interoperability, meaning that two systems understand information in compatible ways.

For example:

Platform A

Customer = User ID 123

Platform B

Customer = Account #456

A technical API may connect the systems, but without compatible semantic interpretation, meaningful interoperability may not exist.

Competition authorities may therefore need to consider:

• common data schemas;
• metadata;
• taxonomies;
• ontologies;
• APIs;
• protocols;
• identifiers;
• authentication;
• portability;
• machine-readable formats.

14. Concentration Through Standards

A semantic standard can become a competitive bottleneck.

Suppose one undertaking controls the dominant:

product classification → seller identification → search taxonomy → recommendation standard

Competitors may be technically capable of entering but unable to reach consumers efficiently unless they conform to the dominant semantic architecture.

This can produce standard-based network effects.

Competition concerns may include:

1. exclusion of alternative standards;
2. discriminatory access;
3. interoperability restrictions;
4. manipulation of standards;
5. tying;
6. excessive dependence; and
7. strategic acquisition of complementary systems.

15. Merger-Control Concerns

Semantic network concentration is also relevant to mergers.

A merger may not produce a large increase in market share in an individual market but could nevertheless combine:

data + users + infrastructure + complementary services + semantic capabilities.

This can increase ecosystem concentration.

Recent EU merger analysis has specifically considered ecosystem theories of harm, including concerns that a transaction can reinforce an existing ecosystem and make entry or expansion more difficult.

Examples of potentially relevant transactions include combinations involving:

• search + AI;
• cloud + AI;
• operating systems + applications;
• social networks + marketplaces;
• e-commerce + logistics;
• payment systems + commerce;
• advertising exchanges + user data.

16. Defensive Foreclosure

A particularly important theory is defensive foreclosure.

An ecosystem leader may restrict or acquire complementary technologies not merely to obtain immediate profits but to prevent another ecosystem from becoming a competitive threat.

The theory can be represented as:

Emerging competitor
↓
complementary technology
↓
potential ecosystem expansion
↓
incumbent responds through exclusion/acquisition
↓
rival ecosystem cannot develop

Recent competition scholarship identifies defensive foreclosure as an important ecosystem-related theory of harm.

17. Lock-In Effects

Semantic networks can produce switching costs.

A user may have accumulated:

• search history;
• recommendations;
• playlists;
• contacts;
• reviews;
• purchases;
• identity credentials;
• preferences;
• application data.

The more information embedded within the ecosystem, the harder it may become to move elsewhere.

This creates:

Data accumulation → personalisation → switching costs → retention → additional data accumulation

Thus, concentration can become self-reinforcing.

18. Single-Homing and Multi-Homing

Competition authorities should examine whether users or businesses can use several platforms simultaneously.

Multi-homing

A seller uses:

• Amazon;
• eBay;
• its own website;
• another marketplace.

This can constrain platform power.

Single-homing

A seller depends almost entirely on one platform.

This can strengthen the platform's bargaining position.

EU ecosystem analysis increasingly considers switching, multi-homing, product integration and cross-product network effects when evaluating ecosystem competition.

19. Possible Competition-Law Theories

Semantic network concentration can potentially generate several conventional competition-law theories:

Article 102 / abuse of dominance

• discriminatory access;
• refusal to interoperate;
• self-preferencing;
• tying;
• leveraging;
• exclusionary conduct.

Article 101 / restrictive agreements

• interoperability restrictions;
• exclusivity arrangements;
• information-sharing agreements;
• standard-setting restrictions;
• algorithmic coordination.

Merger control

• ecosystem entrenchment;
• elimination of nascent competition;
• data concentration;
• complementary-product foreclosure;
• network-effect enhancement.

Indian Competition Act

Comparable issues can arise under:

• Section 3 — anti-competitive agreements;
• Section 4 — abuse of dominant position;
• Section 5 — combinations;
• Section 6 — regulation of combinations.

The Indian approach is increasingly relevant to digital-platform ecosystems, as illustrated by the CCI's analysis of network effects and Google's Play Store ecosystem in Umar Javeed.

20. Possible Remedies

Competition authorities could theoretically employ several remedies depending on the specific infringement.

Structural remedies

• divestiture;
• separation of business units;
• prohibition of acquisitions;
• ownership separation.

Behavioural remedies

• non-discriminatory access;
• interoperability obligations;
• data portability;
• API access;
• prohibition of self-preferencing;
• transparent ranking;
• non-discriminatory terms.

Technical remedies

• interoperable APIs;
• common technical standards;
• data export mechanisms;
• open interfaces;
• identity portability;
• semantic compatibility requirements.

Merger remedies

• access commitments;
• licensing;
• data separation;
• interoperability commitments;
• restrictions on combining datasets.

21. Important Limitation: Not Every Semantic Advantage Is Anticompetitive

Competition law should not treat every large semantic network as unlawful.

Large networks may result from:

• innovation;
• superior products;
• consumer preference;
• economies of scale;
• legitimate investment;
• interoperability;
• efficient integration;
• better algorithms.

The critical question is generally whether market power is being maintained or extended through exclusionary or otherwise abusive conduct, rather than whether an undertaking simply possesses a sophisticated network.

This is particularly important because the refusal-to-access cases such as Bronner and IMS Health establish substantial limitations on compulsory access.

22. Emerging AI Dimension

Semantic network concentration becomes especially significant with generative AI and AI search.

An AI platform may control:

training data → embeddings → knowledge representation → retrieval → ranking → generation → recommendation → transaction

This creates a new potential form of vertical integration.

A dominant AI ecosystem could potentially control both:

1. the semantic infrastructure used to understand information, and
2. the commercial services selected through that infrastructure.

Potential competition questions include:

• Who controls the underlying knowledge graph?
• Can competing AI systems access necessary data?
• Can publishers prevent discriminatory crawling?
• Can users transfer their semantic profiles?
• Can competing applications interoperate?
• Can an AI platform favour its own services?
• Can a dominant platform manipulate semantic rankings?
• Can proprietary ontologies become bottlenecks?

These questions extend established competition-law concepts rather than necessarily requiring an entirely new doctrine.

23. Conceptual Flowchart

Data accumulation
↓
Semantic classification
↓
Network expansion
↓
More users and complementors
↓
More data and relationships
↓
Improved ranking/recommendation
↓
Higher switching costs
↓
Greater dependence
↓
Increased ecosystem concentration
↓
Potential foreclosure of competitors

24. Key Legal Tests to Apply

When analysing a semantic network concentration problem, the following questions are useful:

1. Relevant market

What product, service, platform or ecosystem is being examined?

2. Dominance

Does the undertaking possess substantial market power?

3. Network effects

Does additional participation increase the value of the network?

4. Semantic control

Does the undertaking control classification, ranking, identity or information relationships?

5. Interoperability

Can competitors connect meaningfully to the network?

6. Switching

Can users and businesses realistically move to alternatives?

7. Multi-homing

Can users simultaneously participate in competing networks?

8. Foreclosure

Does the conduct make entry or expansion more difficult?

9. Objective justification

Is there a legitimate technical, security, privacy or efficiency justification?

10. Consumer impact

Does the conduct affect:

• price;
• quality;
• innovation;
• choice;
• privacy;
• interoperability; or
• technological development?

25. Conclusion

Semantic network concentration represents an important evolution of digital-market concentration analysis. The central concern is not merely that one company has a large market share, but that it may control the semantic and technical architecture through which different participants discover, interpret, connect and transact with one another.

The principal case-law foundations include:

1. Google Shopping — control of search-based visibility and self-preferencing;
2. Google Android — leveraging across interconnected digital markets;
3. Microsoft — interoperability and control over technical interfaces;
4. IMS Health — exceptional circumstances involving proprietary information structures;
5. Magill — exceptional access to commercially important information;
6. Bronner — limits on compulsory access;
7. Android Auto — modern digital interoperability;
8. Facebook Marketplace — ecosystem/network leveraging; and
9. Umar Javeed v Google — network effects and platform dominance under Indian competition law.