Competition Law and Future Regulation of Ontology-Driven Markets

1. Introduction

Ontology-driven markets are markets in which competition is substantially influenced by the way economic information is classified, structured, related, interpreted, and made machine-readable.

An ontology is a formal system for defining categories, concepts, relationships, attributes, and rules within a particular domain. In digital markets, an ontology may determine, for example:

• what counts as a product;
• whether two products are treated as substitutes;
• how businesses are categorised;
• which entities are considered related;
• how users, transactions, locations or services are connected;
• which data points are treated as relevant;
• how an AI system understands a market;
• which results, recommendations or competitors are surfaced.

The competition-law significance is that control over the classification architecture can become economically equivalent to control over an important layer of market infrastructure.

Traditional competition law generally examines price, output, market shares, access, exclusion and consumer welfare. Ontology-driven markets require additional attention to classification power, semantic interoperability, knowledge-graph control, data portability, algorithmic interpretation and the ability to define the categories through which competitors are discovered.

2. Meaning of an Ontology-Driven Market

An ontology-driven market can be represented as:

Raw Data → Ontology → Classification → Relationships → Algorithmic Interpretation → Market Visibility → Commercial Outcome

For example, suppose a digital health platform controls the ontology used by hospitals and insurers.

It may determine that:

"Cardiac monitoring device A" belongs to category X,

while:

"AI-enabled cardiac monitoring service B" belongs to category Y.

If reimbursement, search ranking, procurement or insurance comparison depends on those categories, the platform's classification decision can materially affect competition.

Thus, competition may occur not merely within categories, but also over who controls the categories themselves.

3. Core Competition-Law Problem

The central issue can be expressed as:

Who controls the semantic infrastructure through which competitors become visible, comparable and interoperable?

This produces several potential competition concerns.

A. Ontology foreclosure

A dominant undertaking may design its classification system so that competing products are:

• excluded from important categories;
• placed in inferior categories;
• treated as non-comparable;
• prevented from appearing in searches;
• denied interoperability;
• classified as substitutes only when commercially convenient.

B. Semantic self-preferencing

A platform may classify its own products in a manner that gives them:

• broader category recognition;
• greater relevance;
• additional attributes;
• preferred placement;
• greater compatibility with other services.

C. Ontology lock-in

Businesses may invest heavily in adapting their databases, software and operations to one proprietary ontology.

Switching to a competing ontology may then involve substantial:

• migration costs;
• data-conversion costs;
• retraining;
• API restructuring;
• compliance costs;
• loss of historical relationships.

D. Ontology-based exclusion

A competitor may technically have access to data but be unable to compete effectively because it cannot use the dominant platform's semantic structure.

This produces a distinction between:

Data access and meaningful semantic access.

E. Algorithmic discrimination

The same underlying information may receive different competitive treatment depending upon how the dominant ontology labels it.

4. Why Existing Competition Law Is Relevant

Ontology-driven markets do not necessarily require an entirely new competition-law doctrine.

Existing principles can be extended to examine:

1. abuse of dominance;
2. refusal of access;
3. tying and bundling;
4. discriminatory treatment;
5. self-preferencing;
6. interoperability restrictions;
7. exclusionary product design;
8. data advantages;
9. ecosystem leveraging;
10. merger-related entrenchment.

The difficult question is whether existing doctrines can adequately recognise semantic infrastructure as a source of market power.

5. Relevant Market Definition

Traditional market definition may become difficult because ontology-driven platforms can influence the boundaries of the market itself.

For example, an AI platform might classify:

• autonomous-driving software;
• vehicle operating systems;
• mapping data;
• driver-assistance systems;
• fleet-management software

as separate products.

Another ontology might treat several of them as components of a single integrated ecosystem.

Consequently, regulators may need to examine:

5.1 Functional substitutability

Are products functionally interchangeable?

5.2 Semantic substitutability

Are products recognised by the relevant platform as belonging to the same category?

5.3 Data substitutability

Can another ontology produce equivalent commercial information?

5.4 Ecosystem substitutability

Can a business move to another ecosystem without losing commercially important relationships?

6. Six Major Competition-Law Case Laws

Because "ontology-driven markets" are an emerging analytical category, there are not yet six leading cases expressly decided under that label. The following cases provide important doctrinal precedents that can be applied by analogy.

Case 1: Google Search (Shopping) — European Commission

Principle

The Google Shopping decision is highly relevant to ontology-driven markets because it concerned the manner in which a dominant search engine structured and displayed competing services.

Google was found to have given preferential treatment to its own comparison-shopping service in general search results while applying less favourable treatment to competing comparison-shopping services.

Relevance to ontology-driven markets

A future ontology platform could similarly determine:

which category a service belongs to and whether it deserves visibility within that category.

The distinction is important.

A conventional search-ranking system decides where a result appears.

An ontology-driven system may decide something even earlier:

whether the result is conceptually recognised as belonging to the relevant category at all.

Competition principle

Dominance combined with discriminatory treatment of competing services can raise exclusionary concerns where the conduct affects competition.

Future application

A regulator could investigate whether a dominant knowledge graph:

• systematically misclassifies competitors;
• assigns inferior semantic attributes to rivals;
• gives its own services richer ontology relationships;
• prevents rivals from appearing in commercially significant categories.

7. Case 2: Google Android — European Commission

The Google Android proceedings provide another important analogy.

The case involved Google's conduct concerning Android, including restrictions associated with the licensing and distribution of Google's mobile applications and services.

Ontology relevance

Modern operating systems increasingly function as semantic ecosystems.

An ecosystem owner may control:

• application categories;
• device compatibility;
• service relationships;
• default associations;
• APIs;
• metadata;
• search and recommendation structures.

If access to a commercially important ontology is conditioned upon accepting unrelated restrictions, this could resemble leveraging or tying.

Future issue

Suppose a dominant AI ecosystem tells manufacturers:

"To obtain access to our device ontology, you must also adopt our search, advertising or payment infrastructure."

Competition law would need to examine whether ontology access is being used to extend dominance into neighbouring markets.

Principle

Dominance in one technological layer may create competitive concerns where contractual or technical restrictions extend that power into adjacent markets.

8. Case 3: Microsoft — European Commission

The Microsoft cases are particularly relevant to interoperability.

Microsoft's conduct concerning interoperability information demonstrated the importance of access to technical information needed by competing products to function effectively with a dominant platform.

Ontology connection

Future competition disputes may move from:

technical interoperability

to:

semantic interoperability.

Technical interoperability asks:

Can two systems communicate?

Semantic interoperability asks:

Can two systems understand the information in substantially equivalent ways?

For example:

Platform A may provide:

"Customer = high-value enterprise account."

Platform B may receive only:

"Customer = account."

Technically, the data has been transferred.

Commercially, however, the second platform may be deprived of important semantic information.

Future doctrine

Competition authorities may therefore need to distinguish:

Access to data

from

access to the meaning encoded in data.

9. Case 4: IMS Health v NDC Health

Principle

The IMS Health litigation is important for understanding the relationship between intellectual property, essential information structures and competition.

The dispute concerned the use of a pharmaceutical data structure involving geographical segmentation.

The European Court of Justice developed strict conditions under which refusal to license an intellectual-property right could constitute an abuse of dominance.

Ontology relevance

A proprietary ontology could become commercially indispensable if:

• an industry universally adopts it;
• switching becomes impracticable;
• downstream businesses require it;
• competitors cannot reasonably reproduce the same semantic structure.

For example, suppose virtually every pharmaceutical procurement platform uses one dominant ontology for:

• medicines;
• active ingredients;
• therapeutic classes;
• manufacturers;
• dosage forms.

A competing platform might technically possess all the underlying data but still be unable to participate effectively without compatibility with the dominant ontology.

Future issue

The question could become:

When does a proprietary semantic architecture become sufficiently indispensable to trigger competition-law intervention?

IMS Health provides an important starting point for answering that question.

10. Case 5: Magill

Principle

The Magill litigation established an important framework concerning exceptional circumstances in which refusal to license intellectual property can amount to abuse of dominance.

The case involved information concerning television programme listings.

Ontology-driven significance

The deeper relevance is the transformation of information into a commercially indispensable structure.

A future ontology may combine:

• facts;
• relationships;
• classifications;
• metadata;
• identifiers;
• semantic connections.

The underlying individual facts may not be proprietary, while the structured architecture through which they are organised may be protected or controlled.

This creates a potential competition problem:

Can an undertaking monopolise the architecture through which otherwise accessible information becomes commercially usable?

Future application

Regulators may examine:

• indispensability;
• absence of reasonable alternatives;
• market creation;
• exclusion of downstream innovation;
• consumer demand;
• interoperability.

11. Case 6: Bronner v Mediaprint

Principle

In Bronner, the European Court of Justice adopted a restrictive approach to refusal-to-deal claims and the essential-facilities concept.

The Court emphasised the importance of demonstrating that the facility was indispensable and that there was no viable alternative.

Ontology relevance

The case is significant because future regulators may be tempted to declare every important ontology an "essential facility."

That would be problematic.

Not every widely used:

• database;
• taxonomy;
• API;
• knowledge graph;
• metadata standard;
• AI model

should automatically be considered essential.

Future test

A regulator would need to examine:

1. Is the ontology genuinely indispensable?
2. Are reasonable alternatives available?
3. Can competitors develop compatible semantic structures?
4. Would access be technically feasible?
5. Would compulsory access undermine legitimate innovation incentives?
6. Does denial eliminate effective competition?

Bronner therefore provides an important limiting principle against excessive intervention.

12. Additional Relevant Case: United Brands

United Brands v Commission remains relevant to the broader concept of dominance and market power.

The case illustrates that dominance concerns the ability of an undertaking to behave to an appreciable extent independently of competitors, customers and consumers.

Ontology application

An undertaking may possess substantial semantic power even without extremely high conventional market shares if competitors depend upon its classification architecture for:

• discovery;
• interoperability;
• procurement;
• advertising;
• authentication;
• data exchange.

Thus, future market-power analysis may need to consider semantic dependence alongside market share.

13. Additional Relevant Case: Google AdSense

The Google AdSense decision is relevant to leveraging and contractual restrictions.

Ontology relevance

An ontology owner may impose restrictions preventing publishers or businesses from using alternative semantic classification systems.

For example:

access to the dominant advertising ontology may be conditional upon using the platform's own advertising exchange.

This could reinforce ecosystem dependence.

14. Emerging Forms of Ontology-Based Abuse

A. Semantic self-preferencing

A platform gives its own products:

• more categories;
• richer metadata;
• stronger relationships;
• greater search relevance;
• superior compatibility.

Competitors remain technically listed but are semantically disadvantaged.

B. Ontology discrimination

Two economically comparable businesses receive materially different classifications.

For example:

The classification itself may influence visibility and consumer choice.

C. Semantic tying

Access to one ontology may be conditional upon purchasing or adopting another service.

Examples include:

• AI ontology + cloud;
• healthcare ontology + payment system;
• retail ontology + advertising;
• mobility ontology + mapping;
• financial ontology + identity services.

15. Ontology Lock-In

Ontology lock-in may become one of the most important future competition issues.

Consider:

Company A → Ontology A → Historical Data → AI Models → Customer Relationships

After ten years, migrating to Ontology B may require:

• remapping millions of records;
• retraining models;
• rebuilding APIs;
• rewriting software;
• revalidating compliance systems.

The customer therefore becomes economically dependent upon the ontology.

This may create semantic switching costs.

16. Knowledge Graphs and Competition

Knowledge graphs are particularly important because they can encode relationships such as:

Company → owns → subsidiary

Product → manufactured by → manufacturer

Person → associated with → organisation

Service → competes with → service

Technology → compatible with → platform

A dominant knowledge-graph provider may therefore possess information about relationships between market participants, not merely information about individual entities.

This can produce a new form of competitive advantage:

Relational data advantage.

The competitive value lies not merely in knowing:

"Company X exists."

but in knowing:

"Company X competes with Y, supplies Z, uses technology A, operates in market B and is connected to customer group C."

17. AI and Ontology-Driven Competition

Generative AI and agentic AI systems substantially increase the significance of ontology.

AI agents need structured representations of:

• products;
• suppliers;
• prices;
• businesses;
• risks;
• capabilities;
• geographic relationships;
• regulatory requirements.

An AI agent may consequently rely upon one dominant ontology to decide:

"Which suppliers should I recommend?"

If the ontology systematically excludes a competitor, that competitor could become commercially invisible to millions of automated purchasing decisions.

This creates a possible future concept of:

Algorithmic semantic foreclosure

where exclusion occurs not through a traditional contractual restriction but through the architecture used to interpret economic information.

18. Competition Concerns in Autonomous AI Markets

Imagine that an AI procurement agent uses a dominant commercial ontology.

The ontology defines:

• "supplier reliability";
• "premium product";
• "green supplier";
• "enterprise-grade";
• "high-risk supplier."

If those definitions are controlled by one dominant company, the company may indirectly influence market allocation.

The competition problem is therefore not merely:

Who controls the data?

but:

Who controls the categories through which the AI interprets the data?

19. Merger Control and Ontology Markets

Future merger analysis may need to examine acquisitions involving:

• knowledge graphs;
• industry taxonomies;
• identity systems;
• metadata providers;
• data-standard companies;
• AI ontology providers;
• interoperability platforms.

A merger may create competitive harm even where the acquired company has relatively little revenue.

Its strategic value may lie in its semantic position.

Example

A large AI platform acquires the leading ontology provider for:

healthcare + pharmaceuticals + insurance.

The acquisition could give the buyer control over a semantic gateway through which numerous downstream applications operate.

Traditional turnover-based analysis may underestimate this strategic importance.

20. Killer-Acquisition Concerns

An incumbent could acquire an emerging ontology company before it becomes a significant competitor.

The target may possess:

• superior semantic architecture;
• open interoperability;
• innovative classification technology;
• cross-platform standards;
• better entity resolution.

Therefore, future merger regulation may need to consider:

future semantic competition, not merely present revenue.

21. Interoperability as a Regulatory Remedy

One possible remedy is mandatory semantic interoperability.

A dominant platform could be required to provide:

• ontology documentation;
• mappings;
• APIs;
• common identifiers;
• schema information;
• conversion tools;
• machine-readable definitions.

However, regulators would need to avoid forcing companies to reveal genuine trade secrets unnecessarily.

22. Data Portability vs Ontology Portability

Traditional data portability may not be enough.

Consider:

User exports 10 million records from Platform A.

If Platform B cannot understand the relationships and classifications attached to those records, portability is technically satisfied but commercially ineffective.

Therefore, future regulation may need to consider:

Data portability

plus

Semantic portability.

Semantic portability could involve:

• portable identifiers;
• relationship metadata;
• classification mappings;
• schema documentation;
• machine-readable definitions;
• ontology translation mechanisms.

23. Standardisation and Competition

Open standards can promote interoperability.

However, standard-setting can itself create competition risks.

Participants might:

• exclude rival technologies;
• manipulate standards;
• delay adoption;
• impose discriminatory licensing;
• create interoperability barriers.

Therefore, competition law may need to scrutinise ontology standard-setting organisations in the same way it examines other standard-setting environments.

24. Algorithmic Collusion Through Shared Ontologies

A particularly novel concern arises where competing companies use the same commercial ontology.

Suppose competing retailers use one dominant ontology and AI pricing system.

The ontology standardises:

• product categories;
• competitors;
• price signals;
• demand indicators;
• inventory conditions.

AI systems could potentially use the common semantic infrastructure to coordinate market behaviour without direct human communication.

The competition question becomes:

Can a common ontology facilitate algorithmic coordination even where competitors do not directly communicate?

This could extend traditional concerns surrounding algorithmic collusion.

25. Consumer Protection and Competition

Ontology design can also influence consumer choice.

For example, an online platform may classify:

• "refurbished";
• "open-box";
• "used";
• "certified pre-owned"

differently.

If the platform deliberately designs its ontology so that its own products appear in a more attractive category, competition and consumer-protection concerns may overlap.

26. Proposed Future Regulatory Framework

A future regulatory framework could contain eight pillars.

Pillar 1 — Semantic transparency

Dominant platforms should disclose material classification criteria affecting commercial visibility.

Pillar 2 — Semantic interoperability

Competitors should have reasonable mechanisms for exchanging and interpreting structured information.

Pillar 3 — Anti-discrimination

Equivalent businesses should not receive unjustifiably different ontology treatment.

Pillar 4 — Anti-self-preferencing

Dominant platforms should not manipulate semantic classifications to advantage their own downstream products.

Pillar 5 — Ontology portability

Businesses should be able to migrate economically meaningful semantic information.

Pillar 6 — Merger scrutiny

Authorities should consider whether acquisitions consolidate control over strategically important semantic infrastructure.

Pillar 7 — Auditability

Important automated classification systems should maintain records sufficient to investigate discriminatory outcomes.

Pillar 8 — Contestability

Businesses should have mechanisms to challenge erroneous or competitively harmful classifications.

27. Possible Competition-Law Tests

A future regulator could assess ontology-related conduct using the following framework:

Step 1 — Identify the ontology

What classification or semantic system is being controlled?

Step 2 — Identify the economic function

Does the ontology affect:

• search;
• procurement;
• advertising;
• interoperability;
• payments;
• AI recommendations;
• market access?

Step 3 — Establish market power

Does the undertaking possess significant power over the relevant semantic infrastructure?

Step 4 — Identify dependency

How many downstream businesses depend upon the ontology?

Step 5 — Examine alternatives

Can competitors reasonably use another ontology?

Step 6 — Examine conduct

Has the dominant undertaking:

• discriminated;
• self-preferenced;
• denied access;
• tied products;
• degraded interoperability;
• manipulated classifications?

Step 7 — Establish competitive effect

Does the conduct:

• exclude rivals;
• increase switching costs;
• prevent entry;
• reduce innovation;
• reduce consumer choice?

Step 8 — Consider objective justification

Is there a legitimate:

• security;
• privacy;
• technical;
• quality;
• safety;
• intellectual-property

justification?

28. Challenges for Competition Authorities

A. Technical complexity

Regulators will require expertise in:

• knowledge representation;
• AI;
• databases;
• graph technologies;
• machine learning;
• interoperability.

B. Dynamic ontologies

AI systems may continuously change their classifications.

C. Hidden classification effects

A company may not explicitly instruct an algorithm to exclude competitors; the exclusion may emerge from the architecture.

D. Intellectual-property concerns

Regulators must balance competition with legitimate protection of:

• databases;
• software;
• trade secrets;
• proprietary models.

E. False positives

Not every difference in classification constitutes anticompetitive discrimination.

29. Future Case-Law Development

Future courts may have to answer several novel questions:

1. Can an ontology constitute an essential facility?
2. When does semantic interoperability become mandatory?
3. Can misclassification constitute exclusionary abuse?
4. Can a proprietary knowledge graph constitute a bottleneck?
5. Can semantic switching costs establish market power?
6. Can AI-generated classifications constitute discriminatory conduct?
7. Can ontology standardisation facilitate algorithmic collusion?
8. Can acquisition of an ontology provider substantially lessen competition?
9. Can competitors demand access to semantic mappings?
10. What remedies can preserve interoperability without destroying innovation incentives?

30. Key Case-Law Matrix

31. Doctrinal Synthesis

The existing cases collectively suggest several principles relevant to ontology-driven markets:

First, control over an important digital infrastructure can acquire competition-law significance.

Second, exclusion need not necessarily take the form of an outright refusal. It may arise through preferential treatment, interoperability restrictions or discriminatory technical conditions.

Third, indispensability should be treated cautiously. The existence of a widely used ontology does not automatically make it an essential facility.

Fourth, competition authorities may increasingly have to examine semantic rather than merely technical interoperability.

Fifth, merger control may need to recognise that a company's strategic value can arise from its control over a semantic layer even when its current revenues are modest.

32. Conclusion

Ontology-driven markets represent a potential next stage of digital competition law.

Traditional digital competition questions often ask:

Who controls the data?

Ontology-driven markets add a more fundamental question:

Who controls the structure through which the data is understood?

Control over categories, relationships, identifiers and semantic interpretations can determine which businesses are visible, comparable, interoperable and commercially relevant.

The existing jurisprudence of Google Shopping, Google Android, Microsoft, IMS Health, Magill, Bronner, United Brands and Google AdSense provides important building blocks, although none was decided specifically as an "ontology-driven market" case.

Future competition law is therefore likely to move toward a broader conception of market power that considers semantic infrastructure, ontology dependence, knowledge graphs, interoperability, classification discrimination, AI-mediated discovery and semantic switching costs.