Competition Law and Intelligent Matching Systems

1. Introduction

Intelligent Matching Systems (IMS) are algorithmic or AI-driven systems that match buyers with sellers, consumers with service providers, advertisers with publishers, passengers with drivers, patients with healthcare providers, employers with workers, or products with consumers.

Examples include:

• ride-hailing driver–passenger matching;
• e-commerce product ranking and seller matching;
• online travel and hotel-booking systems;
• employment and gig-work platforms;
• digital advertising exchanges;
• financial-credit and insurance matching;
• healthcare and telemedicine platforms;
• app-store recommendation and ranking systems;
• AI-driven marketplace recommendation engines.

From a competition-law perspective, the important issue is not merely that a platform uses AI or algorithms. The legal question is whether the design, operation, data inputs, ranking criteria, or outputs of the matching system:

1. facilitates collusion;
2. excludes competitors;
3. discriminates between competing suppliers;
4. self-preferences the platform's own services;
5. exploits or locks in business users;
6. reduces consumer choice;
7. coordinates prices or other competitive parameters;
8. creates or reinforces market power; or
9. makes an otherwise independent competitive decision effectively dependent upon a common algorithm.

Competition authorities increasingly apply traditional competition principles to technologically sophisticated matching systems rather than treating algorithms as a separate legal category. The CMA, for example, recognises that algorithmic systems can produce substantial efficiencies but may also reduce competition where they are used to coordinate or restrict competitive behaviour.

2. Meaning of an Intelligent Matching System

An intelligent matching system generally performs four stages:

Data collection → Algorithmic assessment → Matching/ranking → Allocation or recommendation

For example:

Passenger requests ride → platform analyses location, driver availability, historical data and price → algorithm selects driver → transaction is allocated.

In an e-commerce marketplace:

Consumer searches product → algorithm evaluates sellers/products → ranking system selects or recommends offers → consumer is directed toward particular sellers.

The sophistication of the system may involve:

• machine learning;
• predictive analytics;
• real-time pricing;
• behavioural profiling;
• recommendation engines;
• natural-language processing;
• automated ranking;
• reinforcement learning;
• dynamic allocation;
• automated bidding;
• personalised recommendations.

The competition-law significance arises because the algorithm may control access to demand.

3. Competition-Law Framework

A. India

The principal provisions are:

Section 3, Competition Act 2002

Section 3 prohibits agreements that cause or are likely to cause an appreciable adverse effect on competition.

Relevant forms include:

• price fixing;
• market allocation;
• limiting supply;
• bid rigging;
• vertical restraints.

An algorithm can therefore become the technical mechanism through which an agreement is implemented.

Section 4

Section 4 concerns abuse of dominant position.

An intelligent matching platform may raise Section 4 issues where a dominant platform:

• manipulates rankings;
• discriminates between suppliers;
• denies market access;
• imposes unfair conditions;
• leverages dominance into another market;
• favours its own services;
• uses data obtained from dependent suppliers against them.

B. European Union

The principal provisions are:

• Article 101 TFEU — agreements, decisions and concerted practices;
• Article 102 TFEU — abuse of dominant position.

In addition, the Digital Markets Act creates specific obligations for designated gatekeepers concerning ranking, self-preferencing, interoperability and access.

C. United States

Important provisions include:

• Section 1 Sherman Act — agreements restraining trade;
• Section 2 Sherman Act — monopolisation and attempted monopolisation;
• Section 7 Clayton Act — mergers and acquisitions;
• FTC Act provisions concerning unfair methods of competition.

An algorithm therefore does not become immune from antitrust law simply because the coordination or exclusion occurs through software.

4. Major Competition Concerns

4.1 Algorithmic Collusion

The most important concern is whether competing businesses use a common intelligent system to coordinate their conduct.

Suppose:

100 competing hotels provide pricing and occupancy data to the same AI platform.

The system learns from all of that data and recommends similar prices to all hotels.

Even if the hotel owners never meet, the system may reduce independent decision-making.

The legal inquiry therefore becomes:

Was there an agreement or concerted practice, or merely parallel conduct produced independently by software?

This distinction is crucial.

5. Case Law

Case 1 — Eturas UAB and Others v Lietuvos Respublikos konkurencijos taryba, C-74/14

This is one of the most important European cases concerning technology-assisted coordination.

Several travel agencies used the common E-TURAS electronic booking system.

The system administrator communicated a message concerning reduction of online discounts and technically modified the system so that discounts were effectively capped.

The European Court of Justice held that, where the agencies were aware of the message and the technical implementation, participation in a concerted practice could be presumed, subject to rebuttal. The Court also emphasised that the mere existence of a technical restriction was not automatically sufficient where awareness of the communication could not be established.

Principle

A common digital system can constitute an instrument of coordination.

The important factors include:

• communication;
• knowledge;
• technical implementation;
• conduct following the communication;
• ability to distance oneself from the practice.

Relevance to intelligent matching

An AI marketplace cannot necessarily be treated as a neutral technical intermediary where its system is being used to communicate or implement a common competitive strategy.

6. Case 2 — Samir Agrawal v Competition Commission of India

The case concerned Ola and Uber and allegations that their algorithmic pricing systems facilitated price coordination among drivers.

The Supreme Court considered allegations concerning price fixing under Section 3 of the Competition Act.

The underlying competition question was whether drivers were independently determining prices or whether the platform's algorithm effectively coordinated prices.

Importance

The case demonstrates that:

algorithmic price determination does not automatically establish an antitrust agreement.

The existence of a common algorithm must be connected to the legal requirements for an agreement, arrangement or concerted practice.

Principle for matching systems

Where a platform:

• controls matching;
• determines prices;
• allocates customers;
• controls driver incentives;

the authority must still establish the necessary competition-law elements rather than assuming that algorithmic intermediation itself constitutes unlawful coordination.

This distinction is particularly important in two-sided markets.

7. Case 3 — United States v RealPage Inc.

The RealPage litigation represents one of the most significant modern examples of algorithmic coordination.

The U.S. Department of Justice alleged that competing landlords supplied non-public, competitively sensitive information to RealPage and that its software used this information to generate rental-price recommendations.

The DOJ alleged violations of Sections 1 and 2 of the Sherman Act.

The allegations included:

• sharing competitors' sensitive information;
• algorithmic processing of that information;
• pricing recommendations;
• mechanisms encouraging adherence to recommendations;
• reduced independent price competition.

The DOJ subsequently pursued settlements involving several landlords and RealPage concerning algorithmic coordination and information sharing.

Competition principle

The important issue is not whether the algorithm itself "agrees" with competitors.

The issue is whether human businesses supplied competitively sensitive information and used a common algorithm in a manner that substitutes coordinated decision-making for independent competition.

Application to matching systems

An intelligent matching platform that receives confidential bids, prices, capacities or strategic information from competing suppliers may create serious competition concerns if that information is used to coordinate outcomes.

8. Case 4 — Google Shopping

The European Commission's Google Shopping decision is highly relevant to intelligent ranking and matching systems.

Google operated a general search engine while also providing its own comparison-shopping service.

The Commission found that Google systematically positioned and displayed its comparison-shopping service more favourably in search results while demoting competing comparison-shopping services through its general search algorithms.

The case therefore illustrates how an algorithmic ranking or matching mechanism can become a means of self-preferencing. The Commission's findings concerned Google's dominance in general search and the treatment of competing comparison-shopping services.

Principle

A dominant platform's algorithmic ranking system may raise Article 102 concerns when the platform:

1. controls access to consumers;
2. operates a downstream service;
3. determines ranking;
4. gives preferential treatment to its own service; and
5. thereby disadvantages competing services.

Relevance

The same principle may apply to:

• food-delivery rankings;
• hotel rankings;
• app recommendations;
• financial-product recommendations;
• healthcare-provider matching;
• marketplace seller rankings.

9. Case 5 — Amazon Marketplace / Buy Box

The European Commission investigated Amazon's marketplace practices concerning:

• third-party seller data;
• the Buy Box;
• Prime eligibility;
• ranking and selection of offers.

The Commission raised concerns that Amazon could use non-public information generated by marketplace sellers to benefit Amazon's own retail operations.

It also raised concerns concerning the criteria used to select the prominent Buy Box offer and Amazon's treatment of offers using its fulfilment services.

Amazon subsequently offered commitments concerning non-discriminatory Buy Box criteria and presentation of competing offers.

Competition principle

A matching or ranking system becomes competition-sensitive where the platform determines:

which supplier receives visibility and which supplier receives the consumer.

Thus, the "match" is economically equivalent to allocation of demand.

10. Case 6 — MakeMyTrip-Goibibo / OYO, Competition Commission of India

The CCI examined allegations concerning the online hotel-booking ecosystem involving MakeMyTrip-Goibibo and OYO.

The allegations included:

• price parity;
• room parity;
• preferential treatment;
• denial of market access;
• differential treatment of hotel suppliers.

The CCI examined the online intermediation market and the relationship between MMT-Go and hotel partners, including allegations that hotels were restricted from offering lower prices elsewhere.

Competition relevance

An online travel platform's matching function does more than connect hotels with consumers.

Its:

• ranking;
• visibility;
• pricing rules;
• availability rules;
• recommendation mechanisms;
• contractual restrictions

can determine which hotel receives consumer demand.

Principle

A platform's algorithmic matching power can reinforce contractual restraints.

Consequently, competition analysis should consider both:

contractual restrictions + algorithmic implementation.

11. Case 7 — Amazon Marketplace — Bundeskartellamt

The German Bundeskartellamt investigated Amazon's use of price-parity clauses and required Amazon to abandon them.

The authority noted that Amazon simultaneously operated as marketplace operator and competitor to marketplace sellers.

More recently, the Bundeskartellamt raised concerns regarding Amazon's algorithmic price-control mechanisms, which can affect whether sellers' offers are visible in search results or eligible for the Buy Box.

Importance for intelligent matching

This illustrates a critical distinction:

Price algorithm → price effect

but also:

Ranking algorithm → visibility effect → demand allocation → competitive effect

Therefore, a matching algorithm can produce anticompetitive effects even where it does not directly set prices.

12. Algorithmic Self-Preferencing

Suppose a dominant marketplace operates:

• its own delivery service;
• its own payment service;
• its own retail products; and
• its own sellers.

Its matching algorithm could systematically give its affiliated services:

• higher rankings;
• better recommendations;
• faster matching;
• greater visibility;
• preferred eligibility;
• lower commissions.

This can raise abuse-of-dominance concerns.

The relevant question is:

Are ranking and matching criteria genuinely competition-neutral, or are they structured to favour the platform's own ecosystem?

The European Commission's Amazon proceedings illustrate precisely this concern with respect to Buy Box and Prime eligibility.

13. Data as a Competitive Advantage

Intelligent matching systems are heavily dependent on data.

A dominant platform may possess:

• consumer search histories;
• purchasing histories;
• seller prices;
• conversion rates;
• inventory;
• delivery performance;
• customer preferences;
• competitor information.

The platform can then use the information generated by one side of the market to compete against that side.

This produces a potential data-feedback loop:

More users
↓
More data
↓
Better matching
↓
More transactions
↓
More data
↓
Greater platform advantage.

The competition concern is whether this feedback loop merely reflects legitimate innovation or creates exclusionary barriers that competitors cannot realistically overcome.

14. Network Effects

Intelligent matching systems frequently operate in two-sided or multi-sided markets.

For example:

More drivers → shorter waiting time → more passengers

and:

More passengers → more drivers → greater driver demand

This creates positive network effects.

Network effects can produce substantial consumer benefits, but they can also strengthen incumbent market power.

A dominant matching platform may therefore acquire:

• scale advantages;
• data advantages;
• reputation advantages;
• liquidity advantages;
• switching-cost advantages.

15. Market-Tipping Effects

Matching platforms can experience market tipping.

If consumers go where the largest number of suppliers are, and suppliers go where the largest number of consumers are, competitors may struggle to achieve sufficient scale.

This is particularly important in:

• ride-hailing;
• food delivery;
• employment platforms;
• hotel booking;
• online marketplaces;
• digital advertising.

Competition authorities may therefore examine whether an incumbent's matching system creates artificial barriers to entry.

16. Exclusive Matching

An intelligent platform may require suppliers to provide services exclusively through it.

For example:

Driver cannot use competing platform.

or:

Hotel cannot provide lower prices through another OTA.

or:

Seller cannot use another logistics provider.

Such restrictions can raise concerns under vertical-restraint and abuse-of-dominance rules depending upon market power and competitive effects.

The MMT-Goibibo and Amazon cases demonstrate the importance of examining contractual restrictions alongside platform mechanisms.

17. Discriminatory Matching

An algorithm may treat otherwise similar suppliers differently.

For example:

Competition law can become relevant where discrimination is:

• unjustified;
• systematic;
• exclusionary;
• connected with dominance;
• capable of foreclosing competitors.

18. Dynamic Matching and Predatory Conduct

Intelligent systems can change matching decisions in real time.

A platform might temporarily:

• lower prices;
• increase driver incentives;
• subsidise customers;
• prioritise particular sellers;
• increase commissions for rivals;
• reduce competitor visibility.

The legal analysis should distinguish:

legitimate competition through innovation and investment

from

strategic conduct designed to eliminate or weaken competitors.

Evidence concerning duration, cost, intent, market power, recoupment and actual effects can become important depending upon the applicable legal provision.

19. Algorithmic Discrimination and Consumer Choice

Personalised matching may result in different consumers seeing different:

• prices;
• suppliers;
• products;
• advertisements;
• rankings;
• recommendations.

Personalisation itself is not necessarily anticompetitive.

But competition concerns can arise if personalisation is used to:

• exclude rivals;
• prevent comparison;
• exploit switching costs;
• conceal competing offers;
• manipulate consumer choice;
• favour affiliated businesses.

20. Information Exchange Through Matching Platforms

A particularly important problem is indirect information exchange.

Competitors might not communicate directly.

Instead:

Competitor A → Platform
Competitor B → Platform
Platform → Algorithm
Algorithm → A and B

If commercially sensitive information is used to align competitive behaviour, competition law may become applicable.

This is one of the central lessons from Eturas and the RealPage litigation.

21. Intelligent Matching and Essential Facilities

A dominant matching platform may become an important gateway to consumers.

Examples:

• dominant app store;
• dominant hotel platform;
• dominant employment platform;
• dominant digital advertising exchange;
• dominant online marketplace.

Refusal to provide access, discriminatory access, or technically inferior access may raise essential-facility, refusal-to-deal, or discriminatory-access questions depending on the jurisdiction and applicable doctrine.

However, mere commercial importance does not automatically establish an essential facility. The legal requirements of the relevant jurisdiction must still be satisfied.

22. Transparency

Transparency is becoming increasingly important.

A competition authority may ask:

1. What factors determine matching?
2. Who controls the algorithm?
3. What data are used?
4. Does the platform use competitors' data?
5. Are affiliated businesses treated differently?
6. Can suppliers challenge algorithmic decisions?
7. Can suppliers opt out?
8. Can suppliers switch platforms?
9. Are ranking criteria changed unilaterally?
10. Is the algorithm independently audited?

The objective is not necessarily to require disclosure of source code.

Rather, competition law may require examination of competitive effects and discriminatory mechanisms.

23. Explainability and Auditability

An intelligent matching system should ideally permit the platform to demonstrate:

• relevant ranking factors;
• data provenance;
• decision rules;
• changes to algorithms;
• treatment of comparable suppliers;
• reasons for suspension;
• reasons for reduced visibility;
• mechanisms preventing competitor-data misuse.

This becomes especially important when the platform itself is both:

marketplace operator + participant in the marketplace.

24. Compliance Architecture

Businesses operating intelligent matching systems should consider:

1. Competition-law screening

Every major algorithmic change should undergo competition assessment.

2. Data separation

Competitively sensitive seller information should not automatically flow into the platform's competing business.

3. Independent ranking criteria

Ranking should be based on objectively defensible criteria.

4. No competitor coordination

The system should not be designed to facilitate coordinated pricing or output restrictions.

5. Audit trails

Maintain records of:

• algorithmic changes;
• training data;
• ranking modifications;
• pricing recommendations;
• compliance reviews.

6. Human oversight

High-risk competitive decisions should not necessarily be left entirely to autonomous systems.

7. Supplier appeal

Business users should have mechanisms to challenge unexplained exclusion or ranking decisions.

25. Distinguishing Legitimate Matching From Anticompetitive Matching

26. Six Core Legal Tests

When analysing an intelligent matching system, the following framework is useful:

Test 1 — Market Power

Does the platform possess substantial market power or dominance?

Test 2 — Control

Does it control access to consumers, suppliers or essential data?

Test 3 — Algorithmic Function

What exactly does the algorithm do?

• match;
• rank;
• price;
• recommend;
• allocate;
• exclude;
• prioritise?

Test 4 — Data

Whose data does the system use?

Test 5 — Competitive Effect

Does the system:

• reduce competition;
• increase barriers to entry;
• foreclose rivals;
• facilitate coordination;
• raise switching costs?

Test 6 — Justification

Does the platform have legitimate efficiency or consumer-welfare reasons for the design?

27. Emerging Issues

Intelligent matching systems raise several future competition-law questions.

A. AI-to-AI competition

What happens when competing firms allow autonomous AI agents to negotiate with one another?

B. Autonomous price matching

Can continuous machine-learning systems converge on supra-competitive prices without explicit human communication?

C. AI agent marketplaces

If AI agents independently select suppliers, platforms could acquire enormous influence over commercial transactions.

D. Predictive exclusion

Algorithms may predict that a new entrant will become a competitor and alter matching behaviour before the entrant becomes significant.

E. Data feedback loops

Large platforms may continuously improve matching because their large user base generates more data than smaller rivals can obtain.

F. Algorithmic discrimination

Different suppliers may receive different levels of consumer exposure without understanding why.

G. Generative AI recommendations

AI assistants may increasingly determine which products, services or suppliers consumers encounter.

This creates a new form of algorithmic intermediation power.

28. Overall Legal Position

The central principle can be stated as follows:

Competition law is technologically neutral but economically sensitive.

An intelligent matching system is not unlawful simply because it is:

• automated;
• AI-driven;
• predictive;
• personalised;
• dynamic; or
• algorithmic.

The competition-law problem arises when the system becomes a mechanism for:

coordination + exclusion + discrimination + self-preferencing + exploitation + foreclosure.

The principal cases illustrate different parts of this spectrum:

1. Eturas — technology-assisted concerted practice;
2. Samir Agrawal — algorithmic pricing and the requirement to establish an actual competition-law agreement;
3. RealPage — algorithmic coordination and use of competitors' sensitive information;
4. Google Shopping — algorithmic self-preferencing and discriminatory ranking;
5. Amazon Buy Box — algorithmic selection and allocation of marketplace visibility;
6. MMT-Goibibo/OYO — platform intermediation, parity and preferential treatment;
7. Amazon/Bundeskartellamt — algorithmic price controls and platform power.

29. Conclusion

Intelligent Matching Systems are becoming a central component of modern digital markets. They do not merely facilitate transactions; they can determine who meets whom, which supplier is visible, which product is recommended, what price is proposed, and ultimately which competitor receives demand.

Consequently, competition law must examine the entire technological and economic architecture of the system.

The most important questions are:

Who controls the algorithm?
What data does it receive?
Whose interests does it serve?
How does it allocate demand?
Does it preserve independent competition?

The developing case law demonstrates that traditional doctrines concerning concerted practices, abuse of dominance, exclusion, discrimination, information exchange, self-preferencing and market foreclosure remain applicable even when competitive conduct is implemented through sophisticated AI and algorithmic matching systems.