The phenomenon of fake reviews in the digital space has evolved from a marginal image issue to a central focus of market supervision authorities and EU legislators. The contemporary ontology of this phenomenon extends beyond primitive content fabrication to encompass any form of communication that, by distorting the actual consumer experience, misleads the recipient, directly influencing their decision-making process. Legally, a fake review is considered not only a completely false message, but also one that, by omitting important facts or manipulating context, creates a false impression of the quality of a product or the reliability of a seller. This practice is classified as unfair commercial activity if its nature causes or is likely to cause the average consumer to make a transactional decision they would not otherwise make, thus violating the fundamental principles of fair dealing.

The typology of activities considered unfair rests on several fundamental pillars, the most blatant of which is direct fabrication. This involves posting or commissioning the creation of false recommendations from specialized external entities, such as marketing agencies, which directly violates regulations on combating unfair market practices. Another mechanism is selective manipulation, in which a business intentionally manages the visibility of reviews by removing, concealing, or delaying the publication of negative reviews while favoring positive ones. Such action distorts the image of actual customer satisfaction and is considered misleading regarding the essential characteristics of a product or service. An equally significant aspect is feigned verification, i.e., declaring that reviews come from real buyers without implementing proportionate and reasonable steps to verify their authenticity, which constitutes a direct violation of the disclosure obligations imposed by the Omnibus Directive.

Contemporary market practices have also evolved more subtle forms of manipulation, such as astroturfing, which involves creating artificial social support through employees or store owners posing as independent consumers. These activities often involve the manipulation of user profiles, where images generated by artificial intelligence algorithms are used to authenticate fictitious accounts, creating false social proof. Each of these practices, regardless of their technological sophistication, is subject to strict scrutiny by competition and consumer protection authorities.

The role of the President of the Office of Competition and Consumer Protection and the responsibility of management boards

The President of the Polish Office of Competition and Consumer Protection (UOKiK) serves as a central regulator in the legal system, endowed with rigorous powers to counteract violations of collective consumer interests. The main disciplinary instrument at the authority’s disposal is an administrative fine, which can be imposed in the amount of 10% of the turnover achieved by the entrepreneur in the financial year preceding the year of issuance of the decision. The amount of the fine is not determined arbitrarily, but rather results from precisely defined criteria, which include, above all, the scale of the violation, its duration, and the degree of intentionality of the perpetrator. Importantly, this fine is intended to serve not only a repressive function but, above all, a preventive and deterrent one, discouraging other market participants from engaging in similar unfair practices involving the manipulation of reviews or misleading as to the authenticity of reviews.

The enforcement procedure in consumer matters is designed to ensure high effectiveness of supervisory activities. A business subject to a sanction is obligated to settle the fine within 14 days of the decision becoming final, which directly contributes to the state budget. A crucial procedural element is the prejudicial nature of the decisions of the President of the Office of Competition and Consumer Protection (UOKiK), which means that the authority’s findings regarding violations of the law are binding on common courts in compensation cases brought by injured customers. This legal structure significantly facilitates consumers in pursuing civil claims, as they do not have to prove the illegality of the store’s actions, focusing solely on demonstrating the damage suffered. The office’s activity in recent years, reflected in numerous proceedings against e-commerce leaders, confirms that protecting the transparency of reviews has become a regulatory priority, translating into real and severe financial consequences for violators.

The contemporary model of liability in consumer protection law departs from a concept focused solely on the business entity, shifting the burden of sanctions also to individuals who actually manage the enterprise. The President of the Office of Competition and Consumer Protection (UOKiK) has the authority to impose a personal fine of up to PLN 2,000,000 on a manager. This liability is triggered by demonstrating that the manager has intentionally allowed – through their actions or conscious omissions – the company to violate collective consumer interests. In case law, the degree of management involvement in decision-making processes regarding marketing and communications is crucial. This liability may therefore affect a management board member who approves a budget for obtaining reviews from external opinion farms or ignores the lack of implementation of verification procedures under the Omnibus Directive, despite being aware of such deficiencies.

It should be emphasized that the responsibility of managers is autonomous and independent of any penalty imposed directly on the entrepreneur. This is intended to provide a strong incentive for management to build internal compliance structures and actively oversee the entity’s operational ethics. In the era of digitalization of trade, where algorithms and automation of marketing processes can generate violations on a massive scale, the personal financial risk of managers is intended to compel prioritizing compliance as the foundation of business strategy. Therefore, the systemic fight against false reviews is implemented not only through sanctions against corporate structures but also by disciplining those who actually shape companies’ market policies. This, according to the legislature, is intended to ensure long-term improvement in integrity standards in electronic trading.

The Omnibus Directive and the blacklist of market practices

The implementation of the Omnibus Directive into the Polish legal system significantly redefined transparency standards in e-commerce, introducing mechanisms that directly address the systemic manipulation of consumer reviews. A key instrument in this regard is the so-called blacklist of market practices, which constitutes a catalog of behaviors considered unfair in all circumstances, eliminating the need for supervisory authorities to conduct a case-by-case analysis of the consequences of a given action. Classifying these market torts as unfair practices aims to eliminate evidentiary difficulties, as their mere existence exaggerates the entrepreneur’s wrongdoing. This legal framework not only strengthens the consumer’s position but, above all, simplifies the evidentiary process, making the fight against e-commerce abuse more effective and predictable for market participants. The foundation of the new regulations is an absolute prohibition on manipulating the verification and authenticity of product recommendations, which imposes an active obligation on sellers to implement procedures to verify the origin of reviews.

Under the current wording of the regulations, it is considered an unfair market practice for a trader to claim that product reviews were posted by consumers who actually used or purchased the product, in situations where reasonable and proportionate steps were not taken to verify their authenticity. This practice violates the consumer’s right to reliable information, which is essential for making an informed decision about purchasing the product, and violating it constitutes conduct contrary to good practice. The law prohibits not only posting completely false reviews, but also commissioning third parties to create them, or transferring recommendations between products with different parameters, which is referred to as review hijacking. Other offenses listed in the catalog are treated equally severely, such as using false quality certificates without appropriate authorization or using surreptitious advertising, which involves using editorial content to promote a product without clearly identifying the paid nature of the communication. Aggressive techniques are also considered particularly burdensome, including mass spamming and forced selling, which involves demanding payment for products delivered to the consumer without their prior order.

The blacklist also eliminates techniques such as bait advertising and direct persuasion of children to purchase, which aims to protect the integrity of the consumer decision-making process from manipulation. This protection of minors stems from their particular vulnerability to advertising messages and their inability to critically assess the persuasive nature of commercial offers. Expanding the list to include a ban on posting or commissioning another person to post false reviews for the purpose of promoting products significantly complements the system, preventing brands from using agencies that fabricate social evidence. It is emphasized that any form of distortion of the actual image of a product’s popularity constitutes a violation of the collective interests of consumers, which entitles the President of the Office of Competition and Consumer Protection (UOKiK) to intervene under public law as soon as a threat to the interests of all market users arises.

A particularly significant and painful consequence of these unfair techniques for entrepreneurs is a specific civil law sanction in the form of an extended right of withdrawal from the contract. If an e-store engages in practices listed in the prohibited catalog or fails to comply with information obligations regarding review verification, the statutory return period granted to the buyers is extended from 14 days to a full 12 months. This mechanism is a direct consequence of the assumption that, in the absence of reliable information, the consumer could not have expressed a fully informed intention to purchase, which suspends the running of standard mandatory deadlines. Systematic combating of review fraud and the use of black market practices is therefore becoming not only a matter of business ethics but the foundation of legal security and stability for every entity operating in the e-commerce sector. Neglect in transparency can lead to mass claims for refunds, posing a real threat to the operational liquidity of the company.

Manipulation Architecture and Platform Obligations under the Digital Services Act (DSA)

The phenomenon known as dark patterns constitutes a sophisticated form of interference in the user’s decision-making process, based on the deliberate use of interface architecture to distort their autonomy of will. Manipulative design patterns are not merely a manifestation of aggressive marketing, but a systematic designer’s action aimed at inducing a specific cognitive bias in the consumer, which ultimately leads to a purchase decision they would not have made in conditions of full transparency. The psychological foundation of these actions is the use of heuristics, i.e., simplified rules of reasoning and automatic thinking, which in the fast-paced environment of e-commerce transactions make the user susceptible to subliminal suggestions. This phenomenon has evolved from simple forms of persuasion to advanced interface manipulation, where the line between inducement and fraud is deliberately blurred to maximize conversion at the expense of the interests of the weaker party in the legal relationship.

A particularly significant area of application of these practices is the system for presenting reviews and suggesting their authenticity, where manipulation takes the form of so-called interface interference. Businesses often employ patterns involving selective content display, which in practice means deliberately hiding negative reviews on subsequent pages of the website while simultaneously highlighting only enthusiastic reviews on the product’s home page. This practice violates the model of the average consumer, who has the right to expect that the image presented of a product’s popularity and quality is reliable and has not been subjected to arbitrary filtering. Manipulation in the sphere of social evidence also includes fabricating popularity indicators, such as false messages about the number of people viewing a given product at a given time or false offer duration counters, which create an artificial sense of scarcity in the user and pressure them to immediately close the transaction. Under the Polish Act on Combating Unfair Market Practices, these activities may be classified as misleading because they distort the actual market conditions, preventing a rational comparison of offers.

Another dimension of manipulation is the technique known as confirmation shaming, which in the sphere of opinion writing involves the use of evaluative and emotional language to coerce users into specific behaviors, for example, through unsubscribe buttons suggesting a lack of consumer awareness. These practices are closely related to the “roach motel model”, where the process of issuing a favorable review is simplified to the maximum extent, while editing, reporting an error, or deleting content requires navigating a complex subpage structure, which is intended to discourage users from correcting false information. In the legal context, such procedural barriers are considered burdensome impediments that violate good practice and the principle of commercial fairness. An analysis of case law and the positions of supervisory authorities indicates that an interface that deliberately hinders users from exercising their rights or changing their minds loses its neutrality and becomes a tool for harming consumer interests.

A fundamental change in the regulatory sphere was brought about by the entry into force of the EU Digital Services Act (DSA), which, in Article 25, explicitly prohibits online platform providers from designing, organizing, and operating interfaces in a way that misleads or manipulates service users. This regulation is overarching and complements the existing consumer protection framework by introducing a direct obligation to maintain neutrality in choice architecture and prohibiting structures that significantly impede users’ ability to make free and informed decisions. Violation of this prohibition entails not only civil law risks but also severe administrative sanctions, which can amount to a significant percentage of the business’s global turnover.

In the sphere of law enforcement, the key role is played by the model design of the average consumer, who is observant and cautious but lacks specialized knowledge of the psychological mechanisms used in interface design. This protection is preventative and abstract in nature, meaning the President of the Office of Competition and Consumer Protection (UOKiK) can intervene in situations where the mere existence of a manipulative pattern poses a real risk of distorting market behavior, without having to wait for measurable financial damage to a specific individual. Effectively combating dark patterns requires businesses not only to comply with the law but, above all, to shift to a design model focused on reliability, where all product information, including opinions, is presented free from coercive mechanisms. Ultimately, interface transparency is becoming a prerequisite for maintaining trust in the digital economy, and the use of sophisticated forms of manipulation is perceived as highly harmful to society, subject to strict assessment in light of the principles of social coexistence.

New obligations for marketplaces regarding moderation and transparency

The entry into force of Regulation 2022/2065, known as the Digital Services Act (DSA), represents a fundamental shift in the liability paradigm for intermediary service providers, particularly marketplaces. This regulation shifts the emphasis from passive content hosting to active oversight of the transparency and security of the digital system, introducing rigorous operational standards aimed at eliminating illegal content while respecting users’ fundamental rights. A key pillar of this reform is the formalization of moderation processes, which until now were often subject to arbitrary internal platform decisions and are now subject to strict procedural rigors contained in the notice-and-action mechanism. Under the DSA, each platform is required to provide easily accessible and user-friendly tools for identifying potentially illegal content, including fake reviews or infringing offers. The mere receipt of a report obliges the provider to promptly and objectively address it.

The evolution of moderation obligations is inextricably linked to the requirement for transparency in decisions, which is achieved through the justification mechanism provided for in the EU regulation. When a marketplace decides to remove content, limit its visibility, or suspend a user’s account, the user is absolutely obligated to provide clear and specific reasons for such action, which is intended to prevent abuse by blocking reliable reviews that are unfavorable to the seller. This system is complemented by a mandatory internal complaint handling system, which allows users to appeal moderation decisions free of charge within a period of at least six months. This constitutes an important procedural guarantee and allows for the correction of potential algorithmic errors. It is indicated that such a legal framework is necessary to counteract the fragmentation of consumer protection, which previously relied primarily on general national clauses that were unsuitable for the scale of operations of global digital entities.

A significant innovation introduced specifically for trading platforms is the “Know Your Business Customer” (KYBC) principle, regulated in the chapter on marketplace transparency. These entities are charged with collecting and verifying information about traders offering their products through their interfaces, including registration data, payment account numbers, and declarations of commitment to offer goods in compliance with EU law. This mechanism aims to eliminate the phenomenon of anonymous sellers, who often promote defective products using fabricated reviews and, after raising capital, disappear from the market, avoiding legal liability. The platform is obligated to suspend services for sellers who fail to submit the required documents, making the marketplace an active guardian of the legality of trade, rather than merely a passive intermediary in trade.

The scope of transparency obligations extends beyond relationships with individual users to include public reporting through the periodic publication of transparency reports. These documents must include detailed data on the number of orders received from national authorities, statistics on content moderation initiated by the platform itself, and information on the use of automated tools in verification processes. For very large online platforms, these rigors are even stricter, including the obligation to conduct annual audits and systemic risk assessments, including analysis of the interface’s vulnerability to manipulation that could negatively impact public safety or consumer protection. The systemic fight against disinformation and unfair market practices is therefore anchored in the full transparency of operational processes, which allows supervisory authorities to continuously monitor the effectiveness of implemented security measures.

Supervision of compliance with these obligations is based on a new institutional architecture, in which national digital services coordinators, working closely with the European Commission, play a central role. The enforcement system for the adopted regulations is based on fines of up to 6% of a provider’s global turnover, which compels compliance with specific cybersecurity standards. This control system is designed to ensure that marketplaces not only implement the required procedures but also apply them reliably and uniformly across the European Union, which is crucial for building consumer confidence in cross-border trade. The introduction of these standards ends the phase of full regulatory freedom for platforms, imposing on them real responsibility for shaping the environment in which the modern exchange of goods and services takes place.

Technological verification mechanisms and modern operating models

Authenticity Suggestion and Pressure Mechanisms

The evolution of digital market oversight has led to the development of mechanisms in which traditional legal instruments are increasingly being replaced by algorithmic jurisdictions based on advanced artificial intelligence systems. The phenomenon known as AI exclusion is a modern form of sanction that, for e-commerce entities, can prove more severe than traditional financial penalties imposed by administrative bodies. The foundation of this process is the integration of data on the credibility of reviews directly with positioning parameters in ranking systems, which means that transparency is no longer merely an ethical obligation but a condition for the technical visibility of an offer. Recommendation algorithms operating within platforms such as Google and Amazon constantly analyze behavioral and linguistic patterns to identify anomalies suggesting manipulation of social evidence. These systems are currently capable of recognizing the structure of texts generated by LLM language models, which are characterized by a specific repetition of phrases and a lack of emotional details typical of authentic consumer experiences. An additional risk factor subject to automatic verification is the so-called review growth rate, where a sudden jump in the number of positive ratings without correlation with actual website traffic or sales volume is interpreted by AI as a warning signal initiating restrictive procedures.

The consequences of an online store being classified by AI systems as posing a high risk of manipulation are immediate and often irreversible in the short term. This mechanism, known in market practice as shadow banning or de-indexing, leads to a drastic decline in visibility in search results and the blocking of offers in advertising systems, effectively cutting the entrepreneur off from key customer acquisition channels. Under the provisions of the Digital Services Act, providers of very large online platforms are required to maintain particular transparency regarding the parameters used in recommendation systems. Article 27 of the aforementioned regulation requires platforms to clearly define in their regulations the key parameters determining information ranking, which aims to limit algorithmic arbitrage and enable entrepreneurs to understand the reasons for a potential decline in their market exposure. It is worth noting that modern risk assessment systems may be classified as high-risk systems within the meaning of the Artificial Intelligence Regulation, which imposes strict requirements on their creators regarding human oversight and the prevention of algorithmic discrimination.

In parallel to restrictive systems, a paradigm known as agentic commerce is developing, in which purchasing processes are carried out by autonomous AI assistants acting directly on behalf of the consumer. In this model, traditional product reviews cease to serve as persuasive texts for humans and become raw input data for machines that filter the market in search of offers with the highest level of verified trust. A key element of this new commerce architecture is the so-called trust layer, built on protocols such as the Universal Commerce Protocol promoted by Google or the Agentic Commerce Protocol developed by OpenAI. These systems are guided not only by price or availability of goods but above all by the certified credibility of the seller’s data, automatically rejecting offers from entities that lack a clear digital traceability of their recommendations. The collaboration of AI assistants with secure payment systems, such as the Agent Payments Protocol, creates a closed ecosystem in which offers at risk of manipulation are excluded at the initial algorithmic selection stage, before they are even presented to the user.

In the era of agent-based commerce, the role of modern shopping assistants is becoming dominant, forcing businesses to redefine their credibility-building strategies. The Context Protocol model and other open-source solutions enable the exchange of context between various AI models and commerce systems, allowing information about unfair practices by a single store to be instantly shared across the entire assistant network. The doctrine suggests that this systematic approach to eliminating abuse is a natural response to the technological ease of fabricating content online. For an e-commerce store, losing its trustworthy status in the eyes of Google or OpenAI algorithms means the modern equivalent of server shutdown, as AI assistants, protecting the interests of their users, will systematically bypass offers that generate manipulative signals. Thus, the fight for authenticity is no longer a mere compliance issue but an existential foundation in the new, automated e-commerce environment, where barriers to entry into the trust layer are becoming increasingly difficult for entities employing pressure mechanisms and suggesting false authenticity.

Compliance as a Service and the Digital Feedback Path

The rapid evolution of the e-commerce market and the increasing professionalization of unfair market practices have forced entrepreneurs to abandon a reactive reputation management model in favor of proactively building a digital immune system. The scale of the challenge facing modern e-commerce is illustrated by analyses of the systematic erosion of trust in the digital sector, pointing to the prevalence of fake reviews and consumer concerns about the mass implementation of generative artificial intelligence for opinion fabrication. This state of affairs creates decision paralysis, where an overabundance of unreliable information, instead of supporting the purchasing process, becomes an insurmountable barrier.

The economic impact of the lack of reliable content verification is directly measurable and translates into tangible operational losses for businesses. The literature emphasizes that exposure to manipulated reviews drastically reduces purchase intentions and brand trust, generating measurable financial losses. The information vacuum filled with false enthusiasm also leads to a phenomenon known as post-purchase dissonance, in which a product that fails to meet expectations is returned to the seller as a complaint or contract withdrawal. Consequently, the lack of investment in transparent review processes generates hidden logistical and operational costs that, in the long run, may outweigh the gains achieved through the temporary increase in conversions driven by manipulation.

In response to increasing regulatory rigor, including the Omnibus Directive, the Digital Services Act (DSA), and the AI Act framework, an operational model known as Compliance as a Service (CaaS) has emerged in market practice. It involves fully outsourcing compliance processes to specialized technology providers who take over the burden of monitoring and verifying content in accordance with current regulations. CaaS allows for the automation of data oversight, which is essential in an environment where the volume of incoming reviews precludes manual oversight without risking accusations of disproportionality. In this approach, compliance ceases to be merely an administrative cost and becomes a component of a strategy for building brand value by guaranteeing the authenticity of every customer touchpoint.

The foundation of the Compliance as a Service model is the maintenance of clean data and the generation of an indisputable digital trace of the review’s provenance. Every published review should be accompanied by a log containing metadata regarding the specific transaction, a unique order number, and delivery status, creating auditable proof of authenticity that can be presented during inspections by supervisory authorities such as the President of the Office of Competition and Consumer Protection. This digital reconstruction of the review process provides the most effective legal shield for businesses, eliminating the risk of allegations of unfair market practices. In the era of algorithmic jurisdiction, where ranking systems favor content supported by digital evidence, having a certified trace of data provenance is becoming a prerequisite for maintaining the market visibility of an offer.

Parallel to technical verification, modern review management systems integrate mediation mechanisms that allow for the amicable resolution of disputes before they are publicly expressed. Market experience suggests that implementing structured review processes allows for the amicable resolution of a significant portion of consumer disputes, effectively preventing the publication of negative reviews resulting from logistical errors. This approach aligns with the principles of reliability and good market practices, building customer relationships based on dialogue rather than solely on the one-way transmission of ratings.

Transaction verification is now becoming the market standard, replacing open, abuse-prone review sections with a system of unique invitations sent only after a purchase is completed. The literature emphasizes that restricting the review process to those who actually purchased the product is the simplest and most effective way to comply with the obligations imposed by the Omnibus Directive. This not only minimizes the risk of severe financial penalties, but above all, provides AI shopping assistants with reliable input data, which, in the new agent-based commerce paradigm, will determine the viability of each entity in the e-commerce ecosystem.

Artykuł Faking reviews in e-commerce – analysis of new legal regulations, algorithmic mechanisms and market practices in the e-commerce sector pochodzi z serwisu KIELTYKA GLADKOWSKI LEGAL | CROSS BORDER POLISH LAW FIRM RANKED IN THE LEGAL 500 EMEA SINCE 2019.

Just a few years ago, online store owners primarily had to ensure terms and conditions, privacy policies, and efficient order processing. Today, this is clearly not enough. EU regulations such as the Omnibus Directive and the Digital Services Act (DSA), as well as the increasing role of artificial intelligence in assessing store credibility, force businesses to consider their platforms much more broadly. It is no longer just about regulatory compliance, but also about building digital trust, which influences a store’s visibility, legal security, and customer purchasing decisions. Below, we present a practical checklist of the most important actions to implement to reduce the risk of sanctions and increase the credibility of an online store.

Practical guidelines for online store owners

I.            Avoiding UOKiK fines and compliance with the Omnibus Directive

a. Implement transactional verification: You should configure your feedback system so that each review you post is technically linked to the unique order number and email address of the customer who actually completed the purchase.

b. Updating the content of the regulations: In the “Rules for publishing opinions” section, the verification procedure should be described in detail, whether all opinions (including critical ones) are published and how the average product rating is calculated.

c. Transparent labeling: Each review should have a clear status indication (e.g., “Purchase confirmed”). If a benefit is provided in exchange for reviews (e.g., a discount code), this information must be clearly and prominently displayed within the review text.

d. Lowest price mechanism: In accordance with the requirements of price transparency, each discount must display the lowest price of the product that was valid in the 30 days prior to the introduction of the discount.

Legal basis: Act of 30 May 2014 on consumer rights ( Journal of Laws of 2024, item 1796, as amended); Directive (EU) 2019/2161 of the European Parliament and of the Council of 27 November 2019 amending Council Directive 93/13/EEC and Directives 98/6/EC, 2005/29/EC and 2011/83/EU of the European Parliament and of the Council as regards the better enforcement and modernisation of Union consumer protection rules (OJ EU L 328 of 2019, No. 328, p. 7, as amended); Act of 23 August 2007 on counteracting unfair market practices ( i.e. Journal of Laws of 2023, item 845).

II.            Ensuring compliance with the Digital Services Act (DSA)

a. Implementing a “report content” mechanism: Every review or user-generated content must have an easily accessible button to report suspected illegality or manipulation of the content.

b. Procedure for justifying decisions: In the event of deletion of an opinion or blocking of a user account, the platform is obliged to send the author a detailed justification indicating a specific violation of the regulations or legal provisions.

c. Internal Complaints Process: Users must be able to appeal moderation decisions for a period of at least 6 months from the date the platform takes action.

d. Designation of a contact point: The entrepreneur must designate an electronic contact point for supervisory authorities and users, enabling efficient communication on matters relating to digital security.

Legal basis: Regulation (EU) 2022/2065 of the European Parliament and of the Council of 19 October 2022 on the single market for digital services and amending Directive 2000/31/EC (Digital Services Act) (OJ EU L 277, 2022, No. 277, p. 1, as amended), in particular Articles 16, 17 and 20.

III.            Reducing the risk of “algorithmic exclusion”

a. Design Patterns (UX) Audit: Eliminate so-called dark patterns, such as asymmetric selector buttons, hard-to-close pop-ups, or mechanisms that make it difficult to unsubscribe. Supervisory algorithms treat such practices as signals of poor interface quality.

b. Data Certification for AI: Ensure structured review data is provided, allowing shopping assistants and crawlers to properly verify the “digital provenance” of the data.

c. Filtering synthetically generated content: It is worth implementing tools that monitor review language for bot-like patterns (unnatural correctness, lack of detail) to avoid indexing false enthusiasm that results in lower trust rankings.

Legal basis: REGULATION (EU) 2022/2065 OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL of 19 October 2022 on the single market for digital services and amending Directive 2000/31/EC (Digital Services Act) (OJ EU L 277, 2022, p. 1, as amended) – Article 25 (prohibition of deceptive interfaces)

IV.            Proper management of data and opinions (CaaS model)

a. Digital Audit Trail: It is recommended to store logs containing transaction metadata related to opinions for a period enabling verification of data reliability (e.g. 12-24 months).

b. Active mediation systems: Instead of deleting negative feedback, use complaint management systems that document the process of resolving customer disputes. Resolving a problem is treated by ranking systems as evidence of high-quality service.

c.       “Know Your Business Customer” principle: When running a marketplace model, it is essential to verify the identity of sellers before allowing them to offer goods, collecting registration numbers and contact details.

Legal basis: REGULATION (EU) 2022/2065 OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL of 19 October 2022 on the single market for digital services and amending Directive 2000/31/EC (Digital Services Act) (OJ EU L of 2022, No. 277, p. 1, as amended) – Article 30; Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation) (OJ EU L of 2016, No. 119, p. 1, as amended).

Artykuł Is Your Online Store Ready for the New Era of Control? A Practical Guide to E-Commerce Responsibilities in 2026 pochodzi z serwisu KIELTYKA GLADKOWSKI LEGAL | CROSS BORDER POLISH LAW FIRM RANKED IN THE LEGAL 500 EMEA SINCE 2019.

The dynamic development of artificial intelligence-based technologies is revolutionizing not only the commercial sector but also the area of state oversight of the digital market. The implementation of multi-agent systems by the Office of Competition and Consumer Protection (UOKiK) opens a new era in consumer rights enforcement, enabling the mass and automated identification of unfair market practices. With the Digital Services Act (DSA) and the Omnibus Directive in force, traditional control methods are giving way to algorithmic interface analysis aimed at eliminating so-called dark patterns and price manipulation. However, the use of “digital controllers” raises fundamental questions for legal science and business practice about the limits of automated decision-making processes in public administration. Although AI agents significantly improve the effectiveness of detecting violations, their legal status as a source of evidence remains the subject of heated debate. The main thesis is that while AI can be a powerful auxiliary tool for regulatory bodies, the ultimate responsibility for determining the facts and assessing the legitimate interests of a party must rest with humans, which is the foundation of a fair procedure in a state governed by the rule of law.

Dark Patterns: Legal and Ethical Aspects of Prohibiting Manipulation in Digital Interfaces

A key obligation of internet platform providers in light of modern regulations is to design interfaces in a transparent and ethical manner. The prohibition of manipulation, formulated, among others, in the Digital Services Act (Article 25), directly affects the structure of so-called deceptive interfaces (dark patterns). Websites and applications cannot be designed in a way that limits the recipient’s cognitive autonomy, interferes with their ability to rationally assess the situation, or forces them to make a purchasing decision that they would not have made under other circumstances.

One of the most glaring examples of such violations is the asymmetry in the contract conclusion and termination process, particularly evident in subscription models. This mechanism relies on extreme simplification of the purchase path while simultaneously mounting procedural barriers when attempting to cancel the service. Visual techniques are used here, among other things: payment activation buttons are highlighted with bright colors and a central location, while contract termination options are deliberately hidden at the bottom of the page, written in small font or masked with colors that blend with the background. Furthermore, canceling a subscription on online platforms often requires multiple selections or confirmation of the desire to cancel, despite the consumer’s prior explicit choice. Artificial intelligence algorithms, analyzing the page structure and visual hierarchy of elements, can pinpoint these disparities with mathematical precision, creating a list of violations that serves as hard evidence.

In the context of the Omnibus Directive, the obligation to disclose the lowest price 30 days before the discount has become a market standard, but its implementation is open to abuse. The practice of “empty promotions” involves artificially inflating the base price just before a planned discount or providing a false reference amount. In this area, AI agents demonstrate particular effectiveness, acting as real-time monitoring systems; they can archive the price history of each product, creating an independent database. Comparing this information with the entrepreneur’s declaration visible on the website allows for immediate detection of manipulation of the promotional algorithm.

An equally important area of control is the phenomenon of drip pricing , or hiding the real costs of a transaction until the final stage of the shopping cart. Businesses often employ a “decoy” strategy, presenting an attractive unit price, which, at the time of order finalization, is increased by mandatory, previously undisclosed costs, such as service fees, packaging costs, or payment processing fees. Pursuant to Article 12 of the Consumer Rights Act, businesses are obligated to clearly and understandably inform consumers about, among other things, the total price for the proposed service. Automated control systems are capable of conducting a full simulation of the purchasing process, from product selection to the payment gateway. Any discrepancy between the price presented in the product list and the amount required to complete the contract is reported by AI as an attempt to circumvent disclosure obligations and a direct violation of the collective interests of consumers.

According to Article 5 of the Act on Combating Unfair Market Practices, the key criterion for assessing a trader’s behavior is the impact of their actions on the recipient’s decision-making process. A market practice is considered misleading if “this action in any way causes or is likely to cause the average consumer to make a transactional decision that they would not otherwise have made”. The legislator specifies that both “spreading false information” and “spreading true information in a manner that is likely to be misleading” can constitute an infringement. In the digital environment, these manipulations most often focus on the “existence of a product, its type, or availability.” A common method of exerting unjustified pressure on consumers is the use of social proof mechanisms and an artificial sense of scarcity. This manifests itself in messages such as: “this product is now being viewed by x people,” “x items have already been purchased today,” or displaying timers indicating that “only 30 minutes left until the end of the promotion.” Particularly problematic from the perspective of trade ethics is the use of so-called false advertising. Timers – clocks counting down to the finale of a supposedly unique price opportunity. In reality, these are fake mechanisms, as after the specified deadline, the offer remains active and the product price remains unchanged or becomes even more favorable. This type of activity, a classic example of dark patterns, is designed to induce fear of missing out (FOMO) in customers and induce them to rush into a transaction. Using AI agents allows regulators to serially monitor such counters and prove their cyclical recurrence, providing direct evidence of deceptive practices.

The algorithm as a controller

With millions of transactions taking place across the country in just a few minutes or hours, standard order verification procedures prove insufficient to effectively fulfill the statutory responsibilities of supervisory authorities. Technological advancements in the form of AI algorithms come to the rescue. These algorithms can automatically monitor numerous commercial transactions simultaneously, generating preliminary opinions that are ultimately subject to human review. Such systems not only save significant processing time but, above all, enable oversight of a much broader range of businesses and their online platforms. The AI multi-agents used in this process are virtual “consumer robots” capable of mass-auditing e-commerce websites, simulating the natural behavior of online users to detect irregularities that a human controller would be unable to detect on such a large scale.

To conduct reliable and effective inspections, Polish law already offers supervisory authorities a toolkit in the form of the “mystery shopper” institution. Traditionally, this involves a person unrelated to the inspected company or the inspecting authority making a purchase and then completing a survey regarding specific activities they observe during standard shopping. The implementation of AI technology by the Office of Competition and Consumer Protection (UOKiK) aims to entrust AI multi-agents with the role of such digital “mystery shoppers.” Their task is to interact with the website interface, add a product to the cart, and complete the entire purchasing process without disclosing that this activity is being performed by an algorithm or that it is part of an official inspection procedure. This approach allows for direct verification of whether the entrepreneur is not using prohibited manipulative practices, known as dark patterns. However, it should be emphasized that the activity of AI multi-agents is strictly regulated by legal procedures and cannot be arbitrary. The algorithm operates under the strict supervision of the President of the Office of Competition and Consumer Protection, who, pursuant to Article 105ia of the Act on Competition and Consumer Protection, must always obtain prior consent from the Court of Competition and Consumer Protection. This mechanism serves as a key safeguard against abuse of power. Furthermore, after completing the inspection, the office is obligated to immediately provide the entrepreneur with an official ID and authorization for the inspection. In the age of digital administration, this obligation can be fulfilled electronically immediately after the AI multi-agents withdraw from the sales platform.

The key legal framework for the operation of algorithms commissioned by the regulator is provided by the EU AI Act. According to its provisions, AI systems used by public authorities for control and supervisory purposes should be considered high-risk AI systems. This entails a strict requirement to design them with appropriate transparency, which allows both the controlling and the controlled entities to properly interpret the system’s results and use them fairly. In practice, this means that algorithms must be built in an “explainable” model. A business subject to allegations based on an algorithmic audit has the statutory right to request full insight into the operation of AI tools. This transparency is essential for the controlled entity to understand the basis and criteria on which the authority deemed its online platform unfair or infringing on the collective interests of consumers (Article 24). This balance between the effectiveness of digital supervision and the right to defense is the foundation of a modern rule of law in the age of algorithms.

The opinion of AI multi-agents as evidence in the case

After completing the inspection activities on the entrepreneur’s online platform, the AI algorithm’s role evolves towards an analytical function, consisting of preparing an opinion indicating detected violations. In the context of potential proceedings against an entity employing unfair market practices, the admissibility of using such an analysis as valid evidence becomes a key issue. Pursuant to Article 7 of the Code of Administrative Procedure (hereinafter referred to as the Code of Administrative Procedure), which establishes the principle of objective truth, a public administration body is obligated to take all steps necessary to thoroughly clarify the factual circumstances. This obligation is consistent with Article 75 § 1 of the Code of Administrative Procedure, which introduces an open catalog of evidence, allowing as evidence anything that may contribute to the clarification of the case, provided it is not contrary to the law.

Under these regulations, the results of AI multi-agent work – taking the form of reports, opinions, or analyses generated after conducting an audit with court approval – fully fall within the statutory definition of evidence. However, it should be clearly stated that an AI opinion cannot be equated with an expert opinion within the meaning of Article 84 of the Code of Administrative Procedure. This stems from the fact that an algorithm does not possess the status of a natural person equipped with specialized knowledge, which is a statutory requirement for appointing an expert. Instead, documentation generated by an AI agent should be classified as a private document or so-called “unnamed evidence.”

Practical justification for this position can be found in the case law concerning digital evidence. The judgment of the Court of Appeal in Szczecin of September 19, 2016, I ACa 364/15, LEX no. 2147337 aptly describes this issue, pointing out that evidence in a case may include official and private documents, but also means other than those listed in Articles 305-308 of the Code of Civil Procedure. Electronic evidence, currently increasingly used in civil proceedings, is not explicitly listed in the catalog of means of evidence. However, the Code of Civil Procedure does not contain a closed list of evidence sources; anything relevant to the case may constitute evidence. Although the above ruling was issued in the context of civil procedure, due to the identical approach to the openness of the evidence system, it remains fully applicable to administrative proceedings conducted by the President of the Office of Competition and Consumer Protection.

The key element of algorithmic evidence remains the human factor, which serves as a primary safeguard over the autonomous operation of technology. It’s important to note that AI multi-agents, despite their high sophistication, operate based on statistical probability models, which carries the risk of misinterpreting dynamic website elements. For example, the system may incorrectly classify a standard technical error as intentional dark web activity. patterns or misinterpret the interface’s intentions in a specific cultural or linguistic context. Therefore, opinions generated by AI agents cannot constitute a standalone and final basis for a decision, but should be subjected to thorough, critical review by an official. Only such a comparison of the “raw” algorithmic result with human knowledge and experience allows for avoiding errors that could lead to unjustified penalties. This approach is directly supported by Article 80 of the Code of Administrative Procedure, according to which a public administration body assesses whether a given circumstance has been proven based on the entirety of the evidence. In this process, the “AI opinion” is only one of many components that must be weighed against other evidence and evaluated through the prism of principles of logic and life experience, ultimately guaranteeing the implementation of the principle of objective truth and protecting the entrepreneur from the automaticity of decisions made by the algorithm.

Summary

Multi-agent system implemented by the Office of Competition and Consumer Protection for automatic control of the e-commerce sector poses a significant challenge for entrepreneurs, forcing strict compliance with regulations regarding dark patterns, price transparency (Omnibus Directive, Art. 6a) and information obligations (Consumer Rights Act, Art. 12). These tools are used to mass detect manipulative practices such as drip pricing, fake timers or making it difficult to unsubscribe. Although AI agents perform a function similar to “mystery shoppers,” their activity must meet the rigors of Article 105ia of the Act on Competition and Consumer Protection, including the requirement to obtain court consent for a controlled purchase. What is crucial from a procedural perspective is that the findings made by the algorithm do not have the status of an expert opinion within the meaning of Article 84 of the Code of Administrative Procedure (lack of the status of a natural person with specialist knowledge), but constitute only a private document or “other evidence” subject to the authority’s free assessment (Article 80 of the Code of Administrative Procedure).

Consequently, the official is required to subject AI reports to thorough human review to eliminate the risk of misclassification resulting from so-called “AI hallucinations” or technical errors in the interpretation of the website’s code. The entrepreneur has full rights of defense based on the principle of active participation of the party (Article 10 of the Code of Administrative Procedure) and the principle of objective truth (Article 7 of the Code of Administrative Procedure), which means the right to question the bot’s logic and to access the instructions and parameters of the AI system, in accordance with the “explainability” requirement enshrined in the AI Act (Article 13). Any decision based solely on the automated generation of conclusions, without providing the party with an opportunity to comment on the evidence (Article 81 of the Code of Administrative Procedure), constitutes a gross violation of administrative procedure and may constitute an effective basis for challenging the authority’s decision.

Sources:

Regulation 2022/2065 on the single market for digital services and amending Directive 2000/31/EC (Digital Services Act) (OJ EU L 277, 2022, No. 277, p. 1, as amended).

Directive (EU) 2019/2161 of the European Parliament and of the Council of 27 November 2019 amending Council Directive 93/13/EEC and Directives 98/6/EC, 2005/29/EC and 2011/83/EU of the European Parliament and of the Council as regards the better enforcement and modernisation of Union consumer protection rules (OJ L 328, 2019, p. 7, as amended).

Act of 30 May 2014 on consumer rights (consolidated text: Journal of Laws of 2024, item 1796, as amended).

Act of 23 August 2007 on counteracting unfair market practices (consolidated text: Journal of Laws of 2023, item 845).

Act of 16 February 2007 on competition and consumer protection (consolidated text: Journal of Laws of 2025, item 1714).

Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act) Text with EEA relevance (OJ L 1689, 2024).

Act of 14 June 1960, the Code of Administrative Procedure (consolidated text: Journal of Laws of 2025, item 1691).

Judgment of the Court of Appeal in Szczecin of 19 September 2016, I ACa 364/15, LEX no. 2147337.

Artykuł Multi-agent system in the service of the Polish Office of Competition and Consumer Protection – a new era of e-commerce control and the limits pochodzi z serwisu KIELTYKA GLADKOWSKI LEGAL | CROSS BORDER POLISH LAW FIRM RANKED IN THE LEGAL 500 EMEA SINCE 2019.