Competition Law and Competition Implications of Cognitive Marketplaces

1. Introduction

A cognitive marketplace is a market in which artificial intelligence, machine learning, predictive analytics, recommendation engines, automated decision systems, generative AI, or other computational systems play a substantial role in matching buyers and sellers, ranking products, setting prices, allocating attention, predicting demand, personalising offers, evaluating users, or making commercial decisions.

Unlike a conventional marketplace, where consumers and businesses make most competitive decisions directly, a cognitive marketplace may delegate significant parts of market functioning to algorithms. Examples include:

• AI-powered e-commerce marketplaces;
• search and recommendation platforms;
• AI travel and hotel marketplaces;
• digital advertising exchanges;
• financial and credit marketplaces;
• app stores;
• cloud-service marketplaces;
• AI model and data marketplaces;
• autonomous procurement platforms;
• algorithmic labour and gig-work platforms.

Competition law therefore has to examine not only who has market power, but also who controls the cognitive infrastructure through which market participants obtain information and make decisions.

The principal competition concerns involve algorithmic dominance, self-preferencing, exclusionary access rules, data concentration, algorithmic collusion, discriminatory ranking, tying, interoperability restrictions, switching costs, network effects and acquisitions of emerging AI competitors.

2. Meaning of a Cognitive Marketplace

A cognitive marketplace can be understood through five interconnected layers:

A. Data layer

The platform collects:

• consumer behaviour;
• transaction data;
• search histories;
• pricing information;
• location information;
• product performance data;
• supplier information;
• behavioural signals.

B. Intelligence layer

AI systems analyse the data to predict:

• consumer demand;
• willingness to pay;
• likely purchases;
• supplier performance;
• customer churn;
• competitor behaviour.

C. Decision layer

The system may automatically determine:

• rankings;
• recommendations;
• prices;
• visibility;
• advertising placement;
• access to customers;
• creditworthiness;
• commissions.

D. Intermediation layer

The marketplace connects:

Consumers → AI system → suppliers

The intermediary may therefore become the principal gateway to market demand.

E. Feedback layer

Every transaction generates additional data, which improves the algorithm:

More users → More transactions → More data → Better prediction → More users → Greater market power.

This creates a potentially powerful data-network-feedback loop.

3. Competition-Law Framework

Cognitive marketplaces can raise issues under several branches of competition law.

4. Relevant Markets

Traditional market definition becomes more complicated in cognitive marketplaces.

A single platform may simultaneously operate in several markets:

1. consumer search;
2. online retail;
3. digital advertising;
4. payment services;
5. cloud computing;
6. AI services;
7. data analytics;
8. app distribution;
9. recommendation services.

Competition authorities may therefore examine multi-sided markets rather than treating the platform as operating in one conventional product market.

Example

An AI marketplace could simultaneously connect:

consumers + merchants + advertisers + payment providers + AI developers.

A conduct that appears harmless on one side can nevertheless restrict competition on another side.

5. Key Competition Implications

A. Algorithmic Self-Preferencing

A dominant cognitive marketplace may use its algorithm to favour its own products.

For example:

Marketplace operates AI ranking system → marketplace launches its own product → ranking algorithm gives greater visibility to that product.

This may disadvantage independent suppliers even where the platform claims that its algorithm is neutral.

The competition-law question is whether the platform is using control over an important intermediary to distort downstream competition.

6. Data as a Competitive Advantage

Data can become an important competitive input.

A dominant cognitive marketplace may possess:

• enormous transaction datasets;
• behavioural profiles;
• supplier performance data;
• pricing information;
• search data;
• consumer preference information.

Competitors may be unable to reproduce these datasets at commercially meaningful scale.

This can create a data-based entry barrier.

The legal issue is not simply possession of data. Competition authorities must examine:

• whether the data is commercially significant;
• whether competitors can obtain equivalent data;
• whether access is technically feasible;
• whether the incumbent uses data obtained from competitors to compete against them;
• whether exclusion from the data materially affects competition.

7. Algorithmic Ranking and Discrimination

Cognitive marketplaces frequently determine the visibility of businesses through algorithms.

A platform may rank one supplier above another because of:

• commission rates;
• advertising payments;
• historical sales;
• platform participation;
• proprietary data;
• algorithmic predictions.

If the criteria are manipulated to disadvantage rivals, the conduct may raise exclusionary-abuse concerns.

The difficulty is that algorithmic discrimination can be difficult to detect because the discriminatory rule may not be expressly written into the platform's terms.

8. Algorithmic Collusion

One of the most significant emerging issues is whether algorithms can facilitate coordination.

Traditional cartel law generally requires some form of agreement, concerted practice, or other legally relevant coordination.

AI systems can, however:

• observe competitors' prices;
• predict their reactions;
• change prices automatically;
• repeatedly interact;
• identify profitable pricing patterns.

This creates several possibilities:

Explicit algorithmic collusion

Competitors deliberately agree to use an algorithm to coordinate prices.

Algorithm-mediated collusion

Businesses communicate or coordinate through a common algorithmic intermediary.

Tacit algorithmic coordination

Algorithms independently learn that maintaining higher prices is profitable without an explicit agreement.

The third category creates a difficult doctrinal question: competition law normally cannot prohibit mere parallel conduct simply because algorithms make it easier.

9. Personalised Pricing

AI can predict an individual's:

• willingness to pay;
• urgency;
• purchasing probability;
• income proxies;
• price sensitivity.

A marketplace could therefore potentially present different prices to different consumers.

Competition concerns arise where personalised pricing:

• exploits market power;
• discriminates between similarly situated users;
• facilitates exclusion;
• prevents effective price comparison;
• creates discriminatory access conditions.

Personalisation itself is not necessarily anti-competitive. Its competitive significance depends on the market structure and the conduct involved.

10. Network Effects

Cognitive marketplaces often display strong network effects.

More consumers attract more suppliers.

More suppliers attract more consumers.

More transactions produce more data.

More data improves the AI.

Better AI attracts still more users.

Thus:

Users → Data → Better AI → Better matching → More users

This feedback mechanism can produce rapid concentration.

Once a platform becomes sufficiently large, a new competitor may face a substantial data-and-network entry barrier.

11. Tying and Bundling

A dominant technology company could require users to adopt one cognitive service as a condition for accessing another.

Examples could include:

• AI assistant tied to an operating system;
• AI search tied to a browser;
• AI payment service tied to a marketplace;
• AI cloud service tied to hosting;
• generative AI tied to productivity software.

Competition analysis should consider:

1. whether the undertaking is dominant;
2. whether separate products exist;
3. whether customers are forced or incentivised to take both;
4. whether rivals are foreclosed;
5. whether legitimate efficiencies justify the arrangement.

12. Interoperability and API Restrictions

Cognitive marketplaces frequently depend upon APIs.

A dominant platform may control access to:

• application interfaces;
• model APIs;
• payment APIs;
• data APIs;
• identity systems;
• interoperability protocols.

Restricting access may make competing services less effective.

The competition question becomes particularly significant where interoperability is necessary or highly important for effective competition.

13. Switching Costs and Ecosystem Lock-In

Users may accumulate:

• behavioural profiles;
• transaction histories;
• AI preferences;
• customised models;
• recommendation histories;
• stored prompts;
• proprietary workflows.

If these cannot easily be transferred to another platform, users may become locked into the incumbent.

High switching costs can reduce:

• consumer mobility;
• multi-homing;
• competitive entry;
• price competition.

14. Merger and Acquisition Concerns

Cognitive marketplaces create special merger-control problems.

A large platform might acquire:

• an AI start-up;
• a recommendation engine;
• a data analytics company;
• a foundation-model developer;
• an AI infrastructure company;
• an emerging marketplace.

The target may have little current revenue but significant future competitive potential.

Authorities may therefore examine:

• innovation competition;
• data assets;
• technology capabilities;
• potential competition;
• access to computing resources;
• interoperability;
• ecosystem effects.

This is particularly relevant to killer-acquisition theories in digital markets.

15. Important Case Laws

1. Google Search (Shopping) — European Commission, 2017

Google Search (Shopping) is one of the leading authorities concerning algorithmic ranking and self-preferencing.

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

Competition principle

A dominant digital intermediary cannot necessarily use control over an important gateway to systematically advantage its own downstream service.

Relevance to cognitive marketplaces

The case is highly relevant where an AI-powered marketplace controls:

• rankings;
• recommendations;
• search results;
• visibility;
• consumer discovery.

It demonstrates that the architecture of an algorithmic marketplace can itself have competitive consequences.

2. Amazon Marketplace — European Commission, 2022

The European Commission's investigation concerning Amazon examined the use of non-public marketplace seller data.

The concern was that Amazon, as both marketplace operator and retailer, could potentially use information generated by independent sellers to inform its own retail activities.

Competition principle

A vertically integrated marketplace may create competitive concerns when the platform has access to commercially valuable information generated by businesses that simultaneously compete with the platform.

Cognitive-marketplace significance

AI dramatically increases the value of such information because algorithms can aggregate and analyse:

• sales;
• demand;
• inventory;
• pricing;
• consumer behaviour.

Thus, information asymmetry can become an important competitive advantage.

3. Google Android — European Commission, 2018

In Google Android, the European Commission examined several contractual practices involving Android, including tying and restrictions affecting competing search and browser services.

Competition principle

Control over an important technological ecosystem can enable a dominant undertaking to extend its market power into adjacent markets.

Cognitive-marketplace relevance

An AI ecosystem can similarly use:

operating system → default service → AI assistant → search → advertising

to create ecosystem-wide competitive effects.

4. Microsoft — Internet Explorer

In the Microsoft Internet Explorer proceedings, competition authorities examined Microsoft's conduct concerning the integration of Internet Explorer with Windows.

The broader Microsoft jurisprudence demonstrates how control over a dominant technological platform can affect competition in complementary markets.

Cognitive-marketplace relevance

The same analytical problem can arise when a dominant operating system integrates an AI assistant, AI search service or recommendation engine and thereby affects competing providers.

5. United States v. Microsoft Corp. — 2001

The U.S. Microsoft litigation concerned Microsoft's conduct in the operating-system and browser markets.

Among the important issues was the use of control over Windows to disadvantage competing technologies.

Competition principle

A dominant platform can potentially use its control over an important technological bottleneck to protect or extend its position in adjacent markets.

Cognitive-marketplace relevance

AI platforms increasingly function as technological bottlenecks. A dominant AI layer could potentially influence:

• search;
• applications;
• advertising;
• productivity;
• commerce.

6. United States v. Apple Inc. — 2024

The U.S. Department of Justice's Apple antitrust litigation concerns allegations relating to Apple's control over the iPhone ecosystem and practices affecting competition.

The case illustrates the importance of examining ecosystem control, including restrictions involving interoperability and access to platform functionality.

Cognitive-marketplace relevance

AI marketplaces may similarly control access to:

• devices;
• APIs;
• identity;
• app distribution;
• payment systems;
• interoperability.

The broader lesson is that competition analysis may need to examine the entire ecosystem rather than an isolated product.

7. Google AdSense — European Commission, 2019

In the Google AdSense case, the European Commission examined contractual restrictions concerning online search advertising intermediation.

The case involved restrictions that limited the ability of publishers to display competing search advertisements.

Cognitive-marketplace significance

Advertising marketplaces are increasingly algorithmic.

An AI advertising intermediary may control:

advertisers → bidding → ranking → placement → publisher revenue.

Restrictions at any point in that chain can potentially affect competition.

8. Google Android Auto — European Commission, 2023

The European Commission's Google Android Auto decision concerned interoperability and access to the Android Auto platform.

Competition principle

Control over a platform interface can have competitive implications where third-party service providers need access to the platform to reach consumers.

Cognitive-marketplace relevance

AI ecosystems similarly depend upon interoperability with:

• operating systems;
• vehicles;
• smart devices;
• cloud infrastructure;
• application platforms.

Restrictions on such interfaces can therefore become significant competition concerns.

9. Intel — European Commission

The Intel abuse-of-dominance proceedings concerned conditional rebates and the possibility that the conduct could foreclose competitors.

Cognitive-marketplace relevance

The underlying principle extends beyond traditional hardware markets.

A dominant cognitive marketplace could potentially provide:

• preferential commissions;
• AI credits;
• ranking advantages;
• rebates;
• access incentives

conditional upon exclusive or loyalty-based arrangements.

10. Qualcomm — European Commission

The Qualcomm cases examined exclusionary practices involving payments and incentives in the semiconductor ecosystem.

Cognitive-marketplace significance

The case demonstrates the importance of analysing incentives that may exclude rivals from technologically important distribution or supply channels.

In AI markets, similar issues may arise around:

• cloud-computing capacity;
• AI chips;
• model access;
• platform distribution;
• exclusive AI partnerships.

16. Indian Competition-Law Relevance

The same issues can be analysed under the Competition Act, 2002, particularly:

Section 3

Prohibits anti-competitive agreements, including arrangements involving:

• price coordination;
• market sharing;
• restrictions on supply;
• concerted practices.

Section 4

Addresses abuse of dominant position, including:

• unfair or discriminatory conditions;
• unfair or discriminatory prices;
• limiting production or technical development;
• denial of market access;
• tying;
• leveraging dominance into another market.

Sections 5 and 6

Provide the framework for regulation of combinations.

This becomes relevant where large technology companies acquire:

• AI start-ups;
• data-intensive companies;
• emerging digital marketplaces;
• algorithmic infrastructure providers.

17. Competition Issues Specific to Cognitive Marketplaces

The principal concerns can be organised into the following taxonomy:

1. Cognitive dominance

Control over the AI decision-making infrastructure.

2. Data dominance

Control over unique and commercially valuable datasets.

3. Algorithmic discrimination

Different treatment of equivalent competitors or customers.

4. Self-preferencing

Algorithmic promotion of the platform's own services.

5. Algorithmic collusion

Algorithms facilitating coordination among competitors.

6. AI-enabled exclusion

Using predictive systems to identify and disadvantage emerging competitors.

7. AI-enabled tying

Forcing adoption of complementary AI products.

8. API foreclosure

Restricting interoperability with competing services.

9. Personalised pricing

Using individual-level information to differentiate prices.

10. Data harvesting

Using marketplace participants' commercially sensitive information.

11. Ecosystem lock-in

Making users dependent on interconnected AI services.

12. Acquisitions of potential competitors

Purchasing emerging AI technologies before they become effective competitors.

18. Essential-Facility Dimension

A particularly important issue arises where a cognitive marketplace controls an infrastructure that competitors cannot realistically reproduce.

Potential examples include:

• dominant AI marketplaces;
• essential API interfaces;
• critical datasets;
• app-distribution infrastructure;
• digital identity systems;
• cloud-AI infrastructure.

A refusal to provide access does not automatically constitute an abuse.

Competition authorities would generally need to consider factors such as:

1. dominance;
2. indispensability;
3. feasibility of alternative access;
4. exclusionary effect;
5. objective justification;
6. proportionality;
7. impact on innovation.

19. Innovation Competition

Cognitive marketplaces can affect competition even before conventional price competition disappears.

A dominant AI marketplace may reduce:

• experimentation;
• development of competing algorithms;
• alternative business models;
• open-source competition;
• technological innovation.

This makes innovation competition particularly important.

A platform could maintain low consumer prices while nevertheless weakening competition through control over:

data + infrastructure + distribution + AI models.

20. Efficiency Defences

AI can also generate substantial efficiencies.

A platform may legitimately argue that algorithmic integration:

• reduces transaction costs;
• improves matching;
• reduces fraud;
• improves logistics;
• lowers search costs;
• improves product recommendations;
• increases consumer choice;
• reduces waste.

Therefore, the mere use of AI does not establish an infringement.

Competition analysis must distinguish between:

legitimate algorithmic optimisation

and

algorithmic optimisation deliberately or materially used to exclude competitors.

21. Regulatory and Enforcement Challenges

A. Explainability

Competition authorities may not understand why an AI system produced a particular ranking.

B. Dynamic markets

AI markets change rapidly, making conventional market definition difficult.

C. Data opacity

Important competitive information may be held privately by the platform.

D. Algorithmic complexity

The platform itself may not be able to explain every machine-learning output.

E. Multi-market effects

Conduct in one market may affect several adjacent markets.

F. Rapid technological change

A market that appears competitive today may become highly concentrated quickly.

G. Potential competition

A small AI firm may represent significant future competitive pressure despite low current revenues.

22. Competition-Law Test for Cognitive Marketplaces

A useful analytical framework is:

Step 1 — Identify the cognitive function

What does the AI system actually control?

Step 2 — Define the relevant market

Identify:

• product market;
• geographic market;
• platform sides;
• complementary markets.

Step 3 — Establish market power

Examine:

• market share;
• network effects;
• data advantages;
• switching costs;
• entry barriers;
• interoperability;
• technological advantages.

Step 4 — Identify the conduct

Determine whether the platform is engaging in:

• self-preferencing;
• tying;
• discrimination;
• exclusion;
• refusal to deal;
• data exploitation;
• algorithmic coordination.

Step 5 — Examine foreclosure

Ask whether competitors are materially prevented from:

• entering;
• expanding;
• accessing consumers;
• accessing data;
• interoperating.

Step 6 — Examine consumer effects

Consider:

• prices;
• quality;
• choice;
• privacy-related competition;
• innovation;
• service quality.

Step 7 — Examine efficiencies

Determine whether the conduct produces objectively verifiable benefits that cannot reasonably be achieved through less restrictive means.

23. Comparative Case-Law Principles

24. Emerging Forms of Cognitive-Marketplace Abuse

Future competition cases may involve:

AI recommendation foreclosure

An AI assistant systematically recommends the platform's own services.

AI supplier scoring

A marketplace's algorithm assigns inferior scores to independent suppliers.

Predictive exclusion

The platform identifies rapidly growing rivals and strategically changes access conditions.

Data feedback foreclosure

The platform uses competitor-generated data to improve its own competing service.

Autonomous pricing coordination

Multiple firms employ systems capable of learning and responding to competitors' prices.

AI model tying

Access to a dominant model is conditioned on using a particular cloud or marketplace.

AI interoperability discrimination

Third-party AI systems receive technically inferior access to a dominant platform.

Algorithmic acquisition strategies

Dominant firms acquire promising AI technologies before they become substantial competitors.

25. Conclusion

Cognitive marketplaces represent an evolution from ordinary digital platforms toward markets in which algorithms themselves perform important competitive functions. The central competition-law concern is therefore not simply the size of an AI platform, but the possibility that control over data, algorithms, interfaces, rankings and decision-making infrastructure can determine which businesses obtain access to consumers.

The established jurisprudence on Google Shopping, Amazon Marketplace, Google Android, Microsoft, Google AdSense, Google Android Auto, Intel, Qualcomm and ecosystem-related platform cases provides useful principles even though many of these cases pre-date today's most advanced AI systems.

The future legal analysis is likely to focus increasingly on five forms of power:

Data power + Algorithmic power + Infrastructure power + Distribution power + Network power

Where these forms of power reinforce one another, competition authorities may need to examine not merely prices, but also access, ranking, interoperability, innovation, data control, switching costs and the architecture of the marketplace itself.