Competition Law and Future Governance of Semantic Economies

Introduction

A semantic economy is an economy in which competition increasingly depends not merely on products, prices, or traditional keywords, but on the meaning, interpretation, classification, contextualisation and prediction of information.

In a semantic economy, artificial intelligence, large language models, knowledge graphs, recommendation systems, search engines, digital assistants and autonomous agents may determine:

• what a consumer's query means;
• which products or services are relevant;
• how products are categorised;
• which supplier is recommended;
• which information is considered authoritative;
• how competing offers are compared;
• which transaction an autonomous agent executes; and
• how commercial decisions are translated from natural language into economic action.

Consequently, future competition law must address a shift from control over products and distribution to control over meaning, data, semantic infrastructure and machine-mediated decision-making.

The central competition-law question becomes:

Who controls the informational layer through which consumers and machines understand, rank and transact in markets?

This issue can be analysed through existing doctrines of dominance, exclusionary conduct, tying, self-preferencing, refusal of access, interoperability, data advantages, algorithmic coordination and merger control.

I. Meaning and Characteristics of a Semantic Economy

A semantic economy is characterised by the conversion of raw information into commercially meaningful relationships.

For example:

Raw data → interpretation → classification → recommendation → transaction

A consumer may ask an AI system:

"Find me a reliable electric vehicle suitable for long-distance travel under my budget."

The AI does not merely search keywords. It may interpret:

• "reliable";
• "long-distance";
• "suitable";
• "budget";
• battery performance;
• charging availability;
• warranty;
• consumer reviews;
• geographic availability; and
• financing conditions.

The system therefore becomes an economic intermediary between consumer intention and market supply.

This creates a new competition problem: a firm may possess enormous power even where it does not directly sell the underlying product.

II. Main Competition-Law Risks

1. Semantic Gatekeeping

The entity controlling the semantic interface may determine which competitors are visible to the consumer.

Traditional search competition asks:

Who appears on page one?

Semantic competition asks:

Which suppliers does the machine consider relevant enough to mention at all?

This makes ranking and recommendation algorithms potentially important competitive infrastructure.

2. Semantic Self-Preferencing

A dominant platform could design its AI system to interpret its own products more favourably.

For example:

Consumer query → AI interpretation → ranking → platform's own product

Potential concerns include:

• preferential inclusion;
• preferential ranking;
• favourable descriptions;
• greater visibility;
• exclusion of rivals;
• favourable quality classifications;
• preferential access to recommendation slots.

The problem resembles traditional self-preferencing but operates at a deeper informational level.

The Google Shopping litigation is particularly relevant because the EU General Court examined Google's favouring of its own specialised search service in general search results.

III. Semantic Dominance

Traditional dominance analysis normally examines market shares, barriers to entry, network effects and countervailing buyer power.

Semantic economies require additional indicators.

Possible semantic-dominance indicators

1. Control over consumer queries.
2. Control over training data.
3. Control over proprietary knowledge graphs.
4. Control over recommendation infrastructure.
5. Control over semantic taxonomies.
6. Access to real-time behavioural data.
7. Ability to determine relevance.
8. Ability to control interoperability.
9. Dependence of downstream businesses on semantic APIs.
10. Ability to influence autonomous purchasing decisions.

A company might therefore possess semantic bottleneck power without having the largest market share in every downstream product market.

IV. Data as a Semantic Competitive Asset

Raw data becomes more valuable when converted into semantic information.

For example:

Search queries + clicks + purchases + reviews + location + context

can be transformed into:

consumer intention + product relevance + predicted preferences + purchasing probability.

This can create a feedback loop:

More users
↓
More queries
↓
More behavioural data
↓
Better semantic interpretation
↓
Better recommendations
↓
More users

This is a form of semantic network effect.

The competition concern is that a dominant firm may become increasingly difficult to challenge because rivals cannot reproduce the informational feedback loop.

The European Commission's 2026 DMA measures concerning access to anonymised Google Search data illustrate the emerging regulatory significance of search data as a competitive input, including for AI services offering search functionality.

V. Essential Semantic Infrastructure

Future competition law may need to recognise certain semantic infrastructures as economically indispensable.

Potential examples include:

• search indexes;
• knowledge graphs;
• product taxonomies;
• semantic APIs;
• AI recommendation systems;
• identity-resolution systems;
• machine-readable catalogues;
• digital maps;
• transaction ontologies;
• interoperability protocols.

A refusal to provide access could potentially raise issues analogous to essential facilities or refusal-to-deal cases, although the strict legal requirements of those doctrines would still have to be satisfied.

VI. Semantic Interoperability

Interoperability becomes critical where several AI systems must communicate.

Imagine:

Consumer AI → Search AI → Retailer AI → Payment AI → Logistics AI

If one dominant firm controls the semantic protocol, it may prevent rival systems from functioning effectively.

Future competition rules may therefore require:

• API access;
• common semantic standards;
• data portability;
• interoperability;
• machine-readable product information;
• non-discriminatory access;
• transparent interface rules.

The EU's current DMA framework already illustrates this direction. The Commission has required measures concerning effective interoperability between competing AI services and Android features, while separately addressing access to search data.

VII. Six Important Case Laws

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

Principle

Google Shopping is highly relevant to semantic economies because it concerned the relationship between:

• general search;
• specialised search;
• ranking;
• visibility; and
• self-preferencing.

The General Court upheld the essential finding concerning Google's favouring of its own specialised comparison-shopping service over competing services.

Importance for semantic economies

A future AI system may perform the same function at a more sophisticated level.

Instead of:

"Here are ten results."

the AI may say:

"Based on your requirements, these are the three products you should consider."

That recommendation can have substantially greater competitive significance.

Lesson: Competition law must scrutinise not merely access to a platform but the conditions under which the platform determines relevance and visibility.

2. Google Android — Google LLC and Alphabet Inc. v European Commission, Case C-738/22 P

The Google Android litigation concerned Google's conduct involving:

• Android;
• Google Search;
• mobile applications;
• pre-installation;
• exclusivity-related payments;
• Android forks; and
• tying.

The Court of Justice addressed the competitive effects of contractual restrictions associated with Google's position in mobile ecosystems.

Semantic-economy significance

The future equivalent may involve:

Operating system → AI assistant → semantic interpretation → commercial recommendation

If the dominant operating-system provider gives its own AI assistant superior technical access while restricting competing assistants, the conduct may raise competition concerns.

The relevant competitive asset would no longer simply be the operating system. It would be the ability of the AI system to understand and act across the user's digital environment.

3. Microsoft — Commission Decision concerning Microsoft's tying of Internet Explorer

The Microsoft litigation historically addressed Microsoft's integration of Internet Explorer with Windows.

The broader competition-law principle concerns the possibility that a firm possessing dominance in one layer of a technological ecosystem may use that position to strengthen another product.

Semantic-economy application

The modern equivalent could be:

Dominant operating system → mandatory AI assistant → preferred semantic service

or:

Dominant cloud platform → proprietary AI model → privileged access to applications and data.

The lesson is that competition law must examine ecosystem leverage, not merely isolated products.

4. Microsoft — Commission Decision concerning interoperability

The Microsoft interoperability litigation also provides an important foundation for semantic economies.

The underlying concern was whether a dominant undertaking could restrict access to information necessary for competitors to achieve effective interoperability.

Semantic significance

Semantic systems increasingly depend upon interoperability.

For example:

AI model ↔ operating system ↔ cloud ↔ application ↔ database

If competitors cannot obtain the information or interfaces necessary to make their systems interoperable, the dominant firm can potentially transform technological compatibility into a competitive barrier.

Principle

Interoperability information can become a competitive input.

This principle is especially important for future AI ecosystems.

5. United States v. Google

The U.S. Google search litigation concerns Google's conduct in maintaining its position in general search through distribution arrangements and related mechanisms.

Semantic-economy significance

Search is becoming increasingly semantic.

Traditional search:

Query → links.

AI-mediated search:

Query → interpretation → synthesis → recommendation → action.

Consequently, control over the starting point for consumer information can become even more significant when consumers increasingly rely upon a single AI-generated answer instead of visiting numerous competing websites.

The case therefore provides a conceptual foundation for examining distribution, default positioning and informational bottlenecks in semantic markets.

6. Meta Platforms Inc. v Bundeskartellamt

The German Facebook/Meta case is particularly important for the relationship between data and competition law.

The German competition authority considered whether Meta's combination of data from different sources constituted abusive conduct by a dominant undertaking.

The subsequent EU judicial proceedings became significant for the relationship between competition law and data-protection principles.

Semantic-economy significance

Semantic systems depend heavily on data.

The relevant competitive advantage may therefore arise not merely from possessing a large dataset but from the ability to:

collect → combine → interpret → predict → personalise.

The case demonstrates that competition law may need to consider how data practices reinforce market power.

VIII. Additional Relevant Authorities

Other important authorities that contribute to the legal architecture include:

7. United Brands v Commission

Relevant to the concept of dominance and the ability of an undertaking to behave independently of competitors, customers and consumers.

8. Hoffmann-La Roche v Commission

Important for exclusionary conduct and loyalty-inducing arrangements.

9. Bronner v Mediaprint

Important for refusal-to-supply and the strict conditions associated with compelling access to an infrastructure.

10. Intel v Commission

Important for analysis of exclusionary rebates and competitive effects.

11. Slovak Telekom v Commission

Relevant to access restrictions and margin-squeeze analysis in network industries.

12. Amazon Marketplace investigations and litigation

Relevant to the use of platform data, marketplace competition and the relationship between a platform's intermediary role and its own commercial activities.

These cases collectively help construct a doctrinal bridge from traditional competition law toward semantic markets.

IX. Algorithmic Coordination in Semantic Economies

Semantic systems may create new forms of coordination.

Traditional cartel:

Human A ↔ Human B → agreement → price coordination.

Algorithmic coordination:

Algorithm A ↔ Algorithm B → continuous observation → adaptive pricing.

Semantic coordination could go further:

AI system interprets market conditions → predicts competitor behaviour → adjusts strategy → competitor AI responds.

The danger is not necessarily an explicit human agreement.

Potential theories include:

• hub-and-spoke coordination;
• conscious parallelism;
• algorithmic facilitation;
• information exchange;
• tacit coordination;
• collusive signalling;
• common algorithm providers.

Competition authorities will therefore need to distinguish legitimate autonomous optimisation from unlawful coordination.

X. Semantic Mergers

Merger control may also change.

Traditional merger analysis asks whether two businesses combine.

Future analysis may need to ask whether a merger combines:

• large datasets;
• AI models;
• search infrastructure;
• semantic taxonomies;
• recommendation engines;
• consumer identity systems;
• cloud infrastructure;
• distribution networks.

Example

Suppose:

Company A: dominant consumer AI assistant

Company B: dominant product-information database.

The merger may not produce a conventional horizontal overlap.

Nevertheless:

AI + product data + recommendation infrastructure

could create a powerful semantic bottleneck.

Therefore, merger authorities may need to consider data and semantic capabilities as competitive assets.

XI. The Problem of Semantic Bias

Competition law traditionally focuses on price, output and market access.

Semantic economies create another possibility:

biased interpretation.

A recommendation engine could systematically describe one supplier as:

• safer;
• more reliable;
• more relevant;
• environmentally superior;
• better value;
• more compatible.

Even without changing prices, such semantic treatment can influence demand.

The competition question becomes:

Is the ranking or interpretation genuinely based on neutral relevance criteria, or is it being manipulated to disadvantage competitors?

XII. Transparency and Explainability

Complete disclosure of AI source code is generally neither necessary nor always appropriate.

However, competition authorities may require sufficient information to establish:

• ranking criteria;
• access conditions;
• discrimination;
• preferential treatment;
• changes to algorithms;
• exclusionary effects;
• data-access arrangements.

A future regulatory model could therefore use:

Procedural transparency rather than complete algorithmic disclosure.

This may protect legitimate trade secrets while enabling competition authorities to investigate discriminatory conduct.

XIII. Data Portability and Semantic Portability

Traditional data portability allows users to move their data.

Semantic economies may require something more sophisticated:

semantic portability.

This means preserving not merely the underlying data but the meaningful relationships attached to it.

For example:

Customer → preferences → purchasing history → product relationships → recommendations

If users move from Platform A to Platform B but lose the semantic structure generated around their data, switching costs may remain high.

Future competition policy could therefore examine whether portability must include:

• machine-readable data;
• metadata;
• preference structures;
• recommendation histories;
• product mappings;
• interoperability standards.

XIV. Consumer Choice in Semantic Markets

Consumer autonomy becomes more complicated where AI systems make decisions on behalf of consumers.

Traditional consumer:

Search → compare → choose.

Semantic economy:

Tell AI objective → AI interprets objective → AI compares → AI selects → AI purchases.

The competitive importance of the intermediary consequently increases.

A consumer may never directly encounter the excluded competitor.

This produces a particularly important concept:

Invisible exclusion

A competitor does not merely receive a lower ranking.

It may be:

semantically invisible.

XV. Autonomous Agents and Competition Law

Future autonomous agents may negotiate with other agents.

For example:

Consumer agent → retailer agent → logistics agent → payment agent

The agent could automatically:

• request quotations;
• negotiate prices;
• compare warranties;
• select suppliers;
• execute contracts;
• change suppliers.

This raises new competition questions.

Potential issues

1. Can agents form unlawful coordination?
2. Who is responsible for algorithmic conduct?
3. Can an agent discriminate between suppliers?
4. Can a dominant platform exclude third-party agents?
5. Can agents manipulate rankings?
6. Can an AI intermediary become a dominant gatekeeper?
7. Can autonomous contracting create collective market power?

XVI. Future Doctrine of Semantic Dominance

Competition authorities may eventually need a framework consisting of several layers.

Layer 1 — Traditional economic power

• market share;
• barriers to entry;
• pricing power.

Layer 2 — Data power

• volume;
• quality;
• exclusivity;
• real-time access.

Layer 3 — Semantic power

• classification;
• interpretation;
• ranking;
• recommendation.

Layer 4 — Infrastructure power

• APIs;
• cloud;
• operating systems;
• search;
• identity systems.

Layer 5 — Agentic power

• ability to initiate transactions;
• negotiate;
• select suppliers;
• execute contracts.

This produces a broader concept:

Semantic dominance = the ability to materially influence how market information is interpreted and converted into commercial decisions.

This should be treated as an analytical concept rather than a standalone legal test under present competition law.

XVII. Future Governance Model

A future competition framework for semantic economies could contain the following components.

XVIII. Ex Ante and Ex Post Competition Regulation

Future governance will probably require both.

Ex post regulation

Used after suspected anti-competitive conduct occurs.

Examples:

• abuse of dominance;
• cartel;
• refusal to deal;
• discriminatory access;
• exclusionary tying.

Ex ante regulation

Rules imposed before harmful conduct occurs.

Examples:

• interoperability;
• data access;
• anti-self-preferencing obligations;
• portability;
• transparency;
• gatekeeper obligations.

The EU Digital Markets Act represents an important example of the movement toward ex ante regulation. The Commission currently designates major digital platforms and imposes obligations concerning areas such as interoperability, ranking and access to certain data.

XIX. Key Legal Challenges

1. Defining the relevant market

Should the market be:

• AI assistants?
• search?
• semantic search?
• product recommendations?
• digital advertising?
• autonomous commerce?

Traditional market-definition tools may become less reliable where one AI system performs multiple functions.

2. Measuring market power

Market share alone may not capture:

• data advantages;
• semantic accuracy;
• network effects;
• switching costs;
• ecosystem dependence.

3. Proving causation

A regulator must determine whether exclusion resulted from:

superior algorithmic performance

or:

anti-competitive manipulation.

That distinction may be technically difficult.

4. Balancing privacy and competition

Compelling data sharing may improve competition but create privacy and cybersecurity risks.

Therefore:

competition access cannot automatically mean unrestricted data access.

The EU's 2026 approach to search-data sharing expressly incorporates anonymisation and data-protection considerations.

5. Protecting innovation

Competition law must avoid treating every superior AI model as an antitrust problem.

A firm should generally be able to benefit from:

• innovation;
• better algorithms;
• superior products;
• legitimate efficiencies.

The legal concern arises where technological superiority is accompanied by exclusionary strategies that distort competitive conditions.

XX. Future Role of Competition Authorities

Competition authorities may increasingly need multidisciplinary capabilities involving:

• economists;
• AI specialists;
• data scientists;
• cybersecurity experts;
• software engineers;
• legal scholars;
• consumer researchers.

Investigations may require examination of:

• model outputs;
• training datasets;
• API logs;
• ranking systems;
• semantic taxonomies;
• recommendation histories;
• model versions;
• access protocols;
• agent interactions.

Competition enforcement therefore becomes partly an algorithmic audit exercise.

XXI. Core Principles for Future Semantic Competition Law

A coherent framework can be expressed through ten principles:

1. Semantic neutrality

Dominant platforms should not manipulate interpretation merely to disadvantage competitors.

2. Interoperability

Competing systems should be able to interact where legally justified.

3. Data contestability

Strategically important data advantages should not automatically become permanent exclusionary barriers.

4. Non-discriminatory access

Critical semantic infrastructure should not arbitrarily discriminate between competitors.

5. Algorithmic accountability

Automated decisions should remain subject to competition-law scrutiny.

6. Agent neutrality

Autonomous commercial agents should not be systematically restricted by dominant intermediaries.

7. Consumer autonomy

Users should retain meaningful ability to choose alternative services.

8. Merger foresight

Authorities should consider future data and semantic capabilities, not merely current revenue overlaps.

9. Innovation protection

Competition law should distinguish innovation from exclusion.

10. Regulatory interoperability

Competition, data protection, AI regulation and consumer law should operate coherently.

XXII. Overall Legal Framework

The future legal architecture can therefore be represented as:

DATA
↓
SEMANTIC PROCESSING
↓
CLASSIFICATION
↓
RANKING / RECOMMENDATION
↓
AI INTERPRETATION
↓
AUTONOMOUS DECISION
↓
TRANSACTION

Competition law must potentially examine each layer, because market power can arise at any point in this chain.

Conclusion

The semantic economy represents an important evolution of digital competition. Market power is increasingly capable of operating through the ability to determine what information means, what is relevant, which alternatives are visible and which commercial action follows from a consumer's intention.

The traditional doctrines of abuse of dominance, tying, refusal to deal, self-preferencing, interoperability, data exploitation, algorithmic coordination and merger control remain applicable, but they may need to be interpreted against this new technological environment.

The most important future competition-law shift is therefore from examining merely:

Who controls the product?

to examining:

Who controls the informational and semantic infrastructure through which consumers and autonomous systems understand, compare and transact in the market?

Cases such as Google Shopping, Google Android, Microsoft interoperability, United Brands, Bronner, Intel and Meta v Bundeskartellamt provide important doctrinal foundations. They do not themselves establish a separate legal doctrine of "semantic economies"; rather, they supply principles that can be adapted to emerging AI-mediated markets.

The future of competition governance is consequently likely to involve a combination of traditional antitrust, ex ante digital regulation, interoperability rules, data governance, algorithmic accountability and merger scrutiny, while preserving sufficient space for legitimate technological innovation.