Competition Law and Machine-Mediated Purchasing Ecosystems

1. Introduction

Machine-mediated purchasing ecosystems are commercial systems in which AI, algorithms, automated agents, recommendation engines, procurement software, smart contracts, or other machine-based technologies substantially influence or execute purchasing decisions.

Examples include:

AI shopping assistants;

automated procurement platforms;

algorithmic B2B purchasing systems;

autonomous purchasing agents;

digital marketplaces;

smart-contract marketplaces;

AI-driven supplier-selection systems;

automated subscription platforms;

machine-controlled inventory purchasing.

Competition law becomes important because the machine may control a significant part of the relationship between buyers, sellers, suppliers, platforms, advertisers, payment providers and logistics providers.

The central question is:

Does machine-mediated purchasing improve competition and efficiency, or does control over the purchasing system create market power, exclusion, coordination, discrimination or foreclosure?

2. Meaning

A machine-mediated purchasing ecosystem can be defined as:

An interconnected commercial environment in which algorithms, AI systems, automated agents or other computational mechanisms substantially determine, recommend, negotiate or execute purchasing transactions between buyers and sellers.

The ecosystem may involve:

Buyer → AI Purchasing Agent → Marketplace → Seller → Payment System → Logistics

The machine may perform some or all of the following:

search;

product comparison;

supplier selection;

price negotiation;

recommendation;

purchasing;

payment;

delivery selection;

subscription renewal;

complaint processing.

3. Why Competition Law Matters

Traditional purchasing decisions are generally made by humans.

Machine-mediated purchasing can change the competitive structure because one algorithm may influence thousands or millions of transactions.

For example, an AI purchasing system might decide:

“Buy only from suppliers registered with Platform X.”

If Platform X is dominant, this could potentially disadvantage competing suppliers.

Similarly, an AI marketplace may recommend its own products more frequently than competing products.

This can create:

self-preferencing;

foreclosure;

exclusion;

tying;

discriminatory access;

algorithmic coordination;

switching costs;

network effects.

4. Basic Competition-Law Formula

Machine-Mediated Purchasing Competition =

Market Power + Automated Decision-Making + Control of Purchasing Channel + Network Effects + Data + Exclusion/Coordination + Competitive Effects

The use of AI itself is not an antitrust violation.

The legal concern depends upon the conduct and its competitive effects.

5. Main Participants

A machine-mediated purchasing ecosystem may contain:

1. Buyers

Consumers or businesses purchasing products.

2. AI purchasing agents

Systems that search, compare and purchase products.

3. Marketplaces

Platforms connecting buyers and sellers.

4. Suppliers

Businesses offering goods or services.

5. Payment providers

Companies processing transactions.

6. Logistics providers

Companies delivering products.

7. Data providers

Entities supplying information used by purchasing algorithms.

8. Infrastructure providers

Cloud, computing, API and software providers.

6. AI Purchasing Agents

An AI purchasing agent may:

receive the buyer's requirements;

search multiple suppliers;

compare prices;

evaluate quality;

negotiate;

select a supplier;

execute the transaction.

This can improve consumer choice.

But if a dominant platform controls the AI agent, it may influence the entire purchasing journey.

7. Gatekeeper Power

A purchasing ecosystem can become a gatekeeper where sellers must access it to reach customers.

The platform may control:

search visibility;

rankings;

commissions;

advertising;

payment;

delivery;

customer data.

This creates a potential bottleneck.

The more purchasing decisions flow through the system, the greater the importance of competitive neutrality.

8. Algorithmic Self-Preferencing

Suppose an AI shopping assistant belongs to a company that sells its own products.

The system may recommend:

Platform's Product → first

while competing products appear later.

Competition concerns can arise where the platform uses its control over purchasing decisions to advantage its own products.

This is particularly important in digital marketplaces.

9. Google Shopping and Machine-Mediated Purchasing

The Google Shopping proceedings are highly relevant.

The European Commission examined Google's treatment of competing comparison-shopping services.

The case involved the use of Google's search results and ranking mechanisms to favour its own comparison-shopping service over competitors.

Competition principle

Control over an important digital access point can become a competition concern where a dominant undertaking uses that position to disadvantage competing services.

Relevance

An AI purchasing agent could become an even more powerful gatekeeper.

Instead of displaying ten search results, it might provide:

“I have selected this product for you.”

If the agent systematically favours affiliated products, the exclusionary effect could be greater because the buyer may never see competing alternatives.

10. Market Definition

Machine-mediated purchasing may require careful market definition.

Potential markets include:

online retail;

AI shopping services;

B2B procurement;

digital marketplaces;

payment services;

advertising;

logistics;

product-specific markets.

The analysis should consider:

substitutability;

buyer behaviour;

switching;

multi-homing;

network effects;

technological development.

11. Two-Sided and Multi-Sided Markets

Purchasing platforms frequently operate on multiple sides.

For example:

Consumers ↔ Marketplace ↔ Sellers

The platform may earn revenue from:

seller commissions;

advertising;

payment services;

subscriptions;

logistics.

Therefore, competition authorities may need to analyse the interactions between different sides.

12. Ohio v. American Express Co.

Case

Ohio v. American Express Co., 585 U.S. 529 (2018)

Facts

American Express operated a two-sided transaction platform connecting merchants and cardholders.

The case concerned contractual restrictions imposed on merchants.

Principle

The Supreme Court emphasized the importance of considering both sides of a two-sided transaction platform when determining competitive effects.

Relevance

Machine-mediated purchasing platforms can also have multiple sides:

buyer;

seller;

advertiser;

payment provider;

logistics provider.

A restriction affecting one side may influence participation on another side.

Lesson

Platform competition cannot always be analysed by examining only one side of the transaction.

13. Exclusive Dealing

An AI purchasing ecosystem may require suppliers to sell exclusively through the platform.

Example:

“Suppliers receiving priority from our AI purchasing agent cannot sell through competing AI purchasing systems.”

Potential concerns increase where:

the platform is dominant;

a large proportion of suppliers are covered;

switching is difficult;

alternative channels are weak.

14. Tying and Bundling

Machine-mediated systems can combine several services.

For example:

Access to an AI procurement system is conditioned on using the platform's payment and logistics services.

This may involve:

Purchasing + Payment + Cloud + Logistics

The legal question is whether the bundle is legitimate and efficient or whether dominance in one market is being leveraged to restrict competition in another.

15. Self-Preferencing

A platform may simultaneously operate as:

Marketplace + Seller

This creates a structural conflict.

The platform controls:

rankings;

search;

customer information;

commissions;

advertising;

recommendations.

It then competes against the sellers dependent upon the platform.

This is sometimes described as the platform-as-referee-and-player problem.

16. Algorithmic Price Comparison

AI purchasing agents can compare prices across thousands of sellers.

This can be highly pro-competitive.

Consumers may benefit from:

lower prices;

better products;

faster comparisons;

reduced search costs.

However, algorithms can also create concerns if they:

facilitate price coordination;

punish discounting;

implement resale-price restrictions;

discriminate between sellers;

make market entry more difficult.

17. Algorithmic Coordination

Suppose several sellers use a common pricing algorithm.

The system monitors competitors and automatically adjusts prices.

Potential competition-law questions include:

Was there an agreement?

Did firms knowingly use the system to coordinate?

Did the algorithm merely respond independently to public market conditions?

Was confidential information shared?

Was the software designed to facilitate coordination?

Similar prices alone do not automatically prove an unlawful agreement.

18. Eturas UAB v Lithuanian Competition Council

Case

Eturas UAB v Lietuvos Respublikos konkurencijos taryba, Case C-74/14

Facts

Eturas operated an online travel-booking system.

A system communication concerned restrictions on discounts available through the platform.

Principle

The case examined competition-law responsibility for conduct communicated and implemented through an electronic platform.

Relevance

It demonstrates that competition law can apply to conduct occurring through software systems.

Modern purchasing ecosystems may transmit:

pricing instructions;

discount restrictions;

supplier requirements;

market information.

Lesson

Software can be the mechanism through which competition-restricting conduct is communicated or implemented.

19. Algorithmic Loyalty

A purchasing platform could reward buyers who purchase exclusively through its ecosystem.

Examples:

points;

cashback;

AI-generated discounts;

loyalty tiers;

personalized offers.

Loyalty systems can be legitimate.

But a dominant platform may create foreclosure risks if incentives make it commercially difficult for buyers to use competing purchasing channels.

20. Hoffmann-La Roche

Case

Hoffmann-La Roche & Co. AG v Commission, Case 85/76

Principle

The Court considered loyalty-inducing rebate arrangements used by a dominant undertaking.

The case is foundational for understanding exclusionary loyalty arrangements.

Relevance

Modern purchasing platforms can create machine-personalized loyalty systems.

An AI system could automatically calculate:

individualized discounts;

rebates;

loyalty points;

purchasing thresholds.

If these arrangements have exclusionary effects and are employed by a dominant undertaking, competition law may become relevant.

Lesson

Automation does not remove competition-law scrutiny of exclusionary loyalty mechanisms.

21. Machine-Generated Switching Costs

AI purchasing ecosystems may learn a buyer's:

preferences;

budget;

purchasing history;

suppliers;

inventory;

payment details.

The longer the buyer uses the system, the more personalized it becomes.

This may produce:

More use → More data → Better recommendations → Greater dependence → More use

A competitor may therefore find it difficult to attract the buyer.

22. Data Portability

One response to machine-generated switching costs is data portability.

A buyer may need to move:

purchasing history;

supplier lists;

preferences;

transaction records;

inventory information.

Without portability, switching may become expensive.

Competition analysis may therefore consider whether data portability and interoperability are available.

23. Network Effects

Purchasing ecosystems can have strong network effects.

More buyers attract more sellers.

More sellers attract more buyers.

More transactions produce more data.

More data improves the AI.

Better AI attracts more buyers.

This produces a reinforcing loop:

Buyers → Sellers → Transactions → Data → Better AI → More Buyers

This can create significant barriers to entry.

24. Foreclosure of Sellers

A dominant purchasing platform could potentially exclude sellers by:

reducing ranking;

increasing commissions;

denying API access;

limiting advertising;

restricting payment access;

imposing exclusivity;

withholding transaction data.

This is a classic platform foreclosure problem with a machine-mediated dimension.

25. Foreclosure of Competing Purchasing Agents

The competition problem can also operate in the opposite direction.

A dominant seller might prevent customers from using independent AI agents.

For example:

A major retailer technically blocks competing purchasing agents from accessing its product catalogue.

This could reduce competition between purchasing intermediaries.

The legal analysis would depend on market power, necessity, justification and competitive effects.

26. Bronner v. Mediaprint

Case

Bronner v Mediaprint, Case C-7/97

Principle

The case concerns the exceptional circumstances in which refusal of access to an infrastructure controlled by a dominant undertaking can constitute abuse.

Relevance

A purchasing ecosystem may control:

APIs;

product catalogues;

transaction interfaces;

payment systems;

distribution networks.

Competitors may request access.

Bronner demonstrates that not every refusal to provide access is unlawful.

Lesson

Indispensability and the practical possibility of duplication are important considerations in access cases.

27. Commercial Solvents

Case

Commercial Solvents Corp. v Commission, Joined Cases 6/73 and 7/73

Principle

A dominant firm controlling an important upstream input can potentially abuse its position by restricting supply to a downstream competitor.

Relevance

Imagine:

AI purchasing infrastructure → marketplace → retail services

If the same company controls an essential upstream technological input and competes downstream, restricting rivals' access could potentially create foreclosure concerns.

Lesson

Vertical control over important inputs can affect downstream competition.

28. Machine-Controlled Marketplace Ranking

Ranking is one of the most important competition mechanisms in purchasing ecosystems.

An AI system may rank products according to:

price;

quality;

delivery;

seller reputation;

advertising;

profitability;

platform affiliation.

Competition questions arise if ranking criteria secretly favour affiliated businesses.

29. Transparency

Competition law does not necessarily require every algorithm to be fully disclosed.

However, from a competition perspective, authorities may need sufficient information to determine:

how the algorithm works;

what inputs it uses;

whether discrimination occurs;

whether competitors are disadvantaged;

whether the system implements exclusionary rules.

Thus, auditability can become important.

30. AI Purchasing Agents and Consumer Choice

AI agents may actually increase competition by:

comparing more suppliers;

identifying cheaper alternatives;

reducing search costs;

helping small sellers reach customers;

negotiating prices;

reducing information asymmetry.

Therefore, regulators should distinguish:

AI as a competition-enhancing intermediary

from

AI as a gatekeeping mechanism used to suppress competition.

31. Machine-Mediated Procurement

The same principles apply to B2B procurement.

Large businesses may use AI to automatically select:

suppliers;

raw materials;

logistics providers;

software;

financial services.

A procurement platform could potentially become powerful enough to determine which suppliers survive.

Potential concerns include:

discriminatory supplier access;

exclusion;

buyer power;

monopsony;

coordinated purchasing;

supplier dependency.

32. Buyer Power and Monopsony

Machine-mediated ecosystems can create buyer-side market power.

A large AI procurement system may control access to a substantial percentage of buyers.

It could potentially:

force suppliers to accept lower prices;

impose restrictive terms;

demand exclusivity;

restrict alternative sales channels.

The competition issue therefore operates on both sides:

Seller power → monopoly

and

Buyer power → monopsony.

33. Machine-Mediated Purchasing and Private Labels

A marketplace may observe which products perform well.

Its AI system then identifies successful third-party products and develops competing private-label products.

This can create concerns when the platform simultaneously controls:

marketplace access;

customer data;

seller data;

product ranking;

private-label products.

The competition analysis should distinguish ordinary competitive imitation from conduct involving unlawful exclusion or misuse of market power.

34. Vertical Integration

Consider:

AI Purchasing Agent → Marketplace → Payment → Logistics → Retailer

One company may control all stages.

Vertical integration can create efficiencies:

faster delivery;

lower costs;

better coordination.

But it may also create foreclosure opportunities.

The relevant question is whether the integration substantially reduces competitors' access to important inputs or customers.

35. India Perspective

In India, machine-mediated purchasing ecosystems can potentially raise issues under the Competition Act, 2002.

Section 3

Potentially relevant to:

anti-competitive agreements;

exclusive arrangements;

vertical restrictions;

coordination.

Section 4

Potentially relevant to abuse of dominance, including:

denial of market access;

discriminatory conditions;

tying/bundling;

leveraging;

exclusionary conduct.

The Competition Commission of India may need to consider:

network effects;

data;

platform dependence;

switching costs;

self-preferencing;

algorithmic conduct.

36. EU Perspective

EU competition law provides an important framework through:

Article 101 TFEU

Anti-competitive agreements.

Article 102 TFEU

Abuse of dominance.

Digital Markets Act

The DMA introduces additional ex-ante obligations for designated gatekeepers.

For machine-mediated purchasing systems, important issues include:

self-preferencing;

interoperability;

data access;

platform neutrality;

tying;

ranking;

switching.

37. UAE Perspective

In the UAE, competition issues involving machine-mediated purchasing ecosystems may be analysed under the federal competition framework, including rules concerning:

restrictive agreements;

abuse of dominant position;

market power;

economic concentration.

Digital purchasing platforms may also require consideration of relevant technology, consumer-protection and data regulations.

38. United States Perspective

In the United States, potential legal frameworks include:

Sherman Act §1;

Sherman Act §2;

Clayton Act;

FTC Act;

state competition statutes.

The precise analysis depends upon whether the conduct involves:

agreement;

unilateral exclusion;

merger;

vertical restraint;

buyer power;

platform conduct.

39. Legitimate Business Justifications

Machine-mediated restrictions may have legitimate purposes.

Security

Preventing fraudulent sellers.

Quality

Ensuring minimum product standards.

Privacy

Protecting consumer information.

Efficiency

Reducing transaction costs.

Reliability

Maintaining delivery standards.

Consumer protection

Preventing unsafe products.

Fraud prevention

Removing suspicious transactions.

These justifications should be tested for:

necessity + objective justification + proportionality.

40. Remedies

Potential remedies include:

1. Behavioural remedies

prohibit discriminatory ranking;

remove unnecessary exclusivity;

prohibit tying.

2. Interoperability

Allow competing purchasing agents to connect.

3. Data portability

Allow buyers and sellers to transfer relevant data.

4. Algorithmic monitoring

Independent audits of ranking and recommendation systems.

5. Access remedies

Permit reasonable access to essential interfaces where legally justified.

6. Structural remedies

In exceptional cases, separation of platform and competing commercial activities may be considered.

41. Practical Example

Suppose SmartBuy AI becomes the dominant AI purchasing agent.

It:

searches products;

compares prices;

negotiates with sellers;

processes payment;

selects logistics;

maintains customer purchasing histories.

SmartBuy also owns an online marketplace.

Its algorithm begins:

ranking its own products first;

charging competitors higher commissions;

limiting competing AI agents' API access;

requiring suppliers to use SmartBuy payment services;

giving exclusive discounts to sellers who do not use rival marketplaces.

The competition analysis would examine:

Market power → platform role → algorithmic conduct → self-preferencing → exclusion → tying → interoperability → foreclosure → consumer effects → efficiencies → remedies.

42. Competition Risks at Each Stage

43. Compliance Checklist

A machine-mediated purchasing platform should ask:

Is the platform dominant?

What is the relevant market?

Does the platform compete with sellers?

Does its AI favour affiliated products?

Are ranking criteria objective?

Can suppliers multi-home?

Can buyers switch purchasing agents?

Is data portable?

Are competitors given reasonable interoperability?

Are exclusivity clauses necessary?

Does the algorithm receive competitor-sensitive information?

Does the system facilitate price coordination?

Does the platform discriminate against competitors?

Are payment and logistics services improperly tied?

Can the restriction be objectively justified?

Is the measure proportionate?

44. Overall Legal Test

A useful examination framework is:

Step 1 — Identify the purchasing ecosystem

Who are the buyers, sellers and intermediaries?

Step 2 — Identify the machine's function

Is it recommending, ranking, negotiating or executing transactions?

Step 3 — Define the relevant markets

Consider both sides where appropriate.

Step 4 — Determine market power

Examine market share, network effects, data and switching costs.

Step 5 — Identify restrictive conduct

Look for:

exclusivity;

tying;

discrimination;

self-preferencing;

refusal of access;

coordination.

Step 6 — Measure foreclosure

How much of the market is affected?

Step 7 — Assess consumer and supplier effects

Consider:

price;

quality;

innovation;

choice;

supplier access.

Step 8 — Examine efficiencies

Determine whether the machine-driven restriction produces legitimate benefits.

Step 9 — Apply proportionality

Could the objective be achieved through a less restrictive method?

Step 10 — Select remedies

Use appropriate behavioural, interoperability, access or structural measures.

45. Important Case-Law Principles at a Glance

46. Conclusion

Machine-mediated purchasing ecosystems can substantially improve competition by reducing search costs, increasing price transparency, expanding supplier access and allowing consumers or businesses to make more informed purchasing decisions.

At the same time, the entity controlling the purchasing machine may become a powerful commercial gatekeeper.

Competition concerns can arise where the system is used to:

favour affiliated products;

exclude competing sellers;

impose exclusivity;

tie services;

restrict interoperability;

exploit data advantages;

facilitate coordination;

increase switching costs;

leverage dominance across markets.

The key principle is therefore:

Competition law should focus on the competitive structure and effects of machine-mediated purchasing rather than treating automation itself as either inherently pro-competitive or inherently anti-competitive.

Quick Revision Formula

Machine-Mediated Purchasing Ecosystems =

AI Purchasing Agent + Platform Power + Network Effects + Data + Ranking + Self-Preferencing + Exclusivity + Tying + Interoperability + Switching Costs + Algorithmic Coordination + Foreclosure + Consumer Choice + Proportionate Remedies

One-Line Exam Definition

Machine-mediated purchasing ecosystems are digital commercial environments in which AI, algorithms or automated agents substantially control or influence purchasing decisions, creating potential competition-law concerns where platform power, data, network effects or automated conduct restrict independent rivalry, market access, consumer choice or supplier competition.