CoChat
Sign Up
Cited by 5Digitalization, Law, and RegulationEuropean and International Contract LawDiverse Legal and Medical Studies

Consumer Law for AI Agents

Open access
Busch, Christoph·2025·German Law Journal
DOI 10.1017/glj.2026.10188
Open this paper in CoChat — free accountOpen PDF

Abstract

Since the public release of ChatGPT in November 2022, the artificial intelligence (AI) landscape is undergoing a rapid transformation. Currently, the use of AI chatbots by consumers has largely been limited to image generation or question-answering language models. The next generation of AI systems, AI agents that can plan and execute complex tasks with only limited human involvement, will be capable of a much broader range of actions. In particular, consumers could soon be able to delegate purchasing decisions to AI agents acting as “Custobots.” Against this background, the Article explores whether EU consumer law, as it currently stands, is ready for the rise of the Custobot Economy. In doing so, the Article makes three contributions. First, it outlines how the advent of AI agents could change the existing e-commerce landscape. Second, it explains how AI agents challenge the premises of a human-centric consumer law which is based on the assumption that consumption decisions are made by humans. Third, the Article presents some initial considerations how a future consumer law could look like that works for both humans and machines.

AI snapshot

Written from the paper’s full text.

TL;DR

This Article argues that EU consumer law, built around human purchasing decisions, is not yet ready for “Custobots”—AI agents that autonomously buy goods and services. It maps how agentic commerce changes e-commerce and identifies needed adaptations in the average consumer test, information duties, choice architecture, and digital vulnerability.

Key findings

  • AI agents such as OpenAI’s “Operator,” previewed on January 23, 2025, can browse the web and perform tasks like buying goods, booking tickets, and making reservations, moving consumers from “augmented consumers” to “algorithmic consumers.”
  • Agentic commerce may shift online marketplaces from persuading humans to influencing AI agents, potentially disrupting gatekeepers, changing customer journeys, enabling “comparison shopping on steroids,” and replacing SEO with “AI agent optimization” (AAO).
  • EU consumer law’s human-centric premises are challenged: the “average consumer” test may need an “average Custobot” counterpart; information duties may need machine-readable formats and more detailed disclosures for AI agents; and AI agents may be less vulnerable to some human dark patterns but newly vulnerable to adversarial attacks and prompt injection.
  • The Article points to regulatory developments including the UNCITRAL Model Law on Automated Contracting adopted in July 2024, the ELI DACC Model Rules, and the EU Digital Product Passport, which will become mandatory for various product categories between 2026 and 2030.

Methods

  • Conceptual legal analysis of EU consumer law and emerging AI-agent commerce, with no empirical study or dataset.
  • Draws on legal instruments, case law, policy reports, and examples such as the UCPD, DSA Article 25, CJEU average consumer case law, UNCITRAL MLAC, and ELI DACC Model Rules.
  • Uses secondary literature from behavioral law and economics, machine psychology, and AI psychometrics to frame open questions about AI-agent decision-making.

Limitations

  • The paper states that it is an open question whether AI-driven Custobots will be more “rational” than human consumers, and that current findings on LLM biases are “somewhat inconclusive.”
  • It states that more empirical research is needed on AI-agent vulnerabilities and on taxonomies of malicious design patterns that manipulate AI agents.
  • The text makes evident that the Article is a conceptual and policy-oriented piece without original empirical data, so its claims about future consumer behavior and market effects are projections rather than tested outcomes.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1017/glj.2026.10188

Cites

23 of 23 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 30 cited sources checked · 0 supported · 0 flagged

  • Interpretationreasoning

    AI agents could be vulnerable to cybersecurity exploits and adversarial attacks that trick Custobots into a transaction.

    “AI agents could be vulnerable to cybersecurity exploits and adversarial attacks that trick Custobots into a transaction.”
    Quote in the paperSubstantial claim
  • Interpretationcitation

    In the Custobot Economy, rules protecting against manipulation by unfair commercial practices may have to be recalibrated as the vulnerability of AI agents to misleading and aggressive commercial practices may differ from that of humans.

    “In the Custobot Economy, these rules may have to be recalibrated as the vulnerability of AI agents to misleading and aggressive commercial practices may differ from that of humans.65”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [65]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Interpretationreasoning

    The use of AI agents could reduce the significance of the information overload criticism of the information model.

    “the use of AI agents could reduce the significance of one of the main criticisms of the information model, in other words, the problem of information overload.”
    Quote in the paperSubstantial claim
  • Interpretationreasoning

    AI agents are not invulnerable against manipulation.

    “However, this does not mean that AI agents are invulnerable against manipulation.”
    Quote in the paperSubstantial claim
  • Interpretationreasoning

    Whether the consumer is actually able to take an informed decision in a reduced-disclosure scenario would depend on the capabilities of the specific AI agent used by the consumer.

    “Whether the consumer is actually able to take an informed decision would in such a scenario very much depend on the capabilities of the specific AI agent used by the consumer.”
    Quote in the paperSubstantial claim
  • Classified as probably not a substantial claim. 49
  • May combine two claims. 19

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

    Interpretationnone given

    AI agents could help consumers make better informed choices, and less biased purchasing decisions.

    “AI agents could help consumers make better informed choices, and less biased purchasing decisions.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationcitation

    Unlike human consumers, AI agents will not forget to cancel a trial subscription, and they could remind a consumer to terminate unused memberships.

    “Unlike human consumers, AI agents will not forget to cancel a trial subscription, and they could remind a consumer to terminate the gym membership or the Xbox game pass that is no longer being used.46”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [46]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    Current SEO practices could be replaced by a new strategy of AI agent optimization (AAO).

    “current SEO practices could be replaced by a new strategy of AI agent optimization (AAO).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    This could solve the problem that standardized nudges often lack precision.

    “This could solve the problem that standardized nudges often lack precision.68”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [68]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

    Interpretationreasoning

    The shift towards agent-to-website or agent-to-agent interaction could require major changes in the design of user interfaces.

    “the shift towards agent-to-website interaction or agent-to-agent interaction could require major changes in the design of user interfaces.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

    Interpretationnone given

    AI agents could help consumers make better decisions and overcome behavioral biases.

    “More generally, AI agents could help consumers make better decisions and overcome behavioral biases.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    There could be new AI-specific vulnerabilities that do not occur in the context of human decision making.

    “there could be new AI-specific vulnerabilities that do not occur in the context of human decision making.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Interpretationreasoning

    Consumers could soon delegate purchasing decisions to AI agents acting as Custobots.

    “In particular, consumers could soon be able to delegate their purchasing decisions to AI agents.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Interpretationreasoning

    AI agents could easily process DPP information and help consumers make more informed decisions related to sustainable and circular consumption.

    “In contrast, AI agents could easily process this information and help consumers in making more informed decisions related to sustainable and circular consumption.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    Custobots will be able to draw on a broader range of data and consider more factors when comparing products.

    “Custobots will be able to draw on a broader range of data—for example, customer reviews, user forums, product tests, online media—and consider more factors when comparing products.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationreasoning

    Providing more information to human consumers would not improve decision quality, but possibly worsen it.

    “Providing more information to human consumers would not improve the decision quality, but possibly worsen it.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Premisenone given

    EU consumer law as it currently stands is human-centric, based on the premise that purchasing decisions are made by humans.

    “Consumer law, as it currently stands, is human-centric.It is based on the premise that decisions to buy goods or services are made by humans.”
    • Also stated in: § I. Beyond Human-centricity: Towards an Average Consumer Test for Custobots? · plain
      “EU consumer law in its current form is human-centric, that is, it is designed with the needs and weaknesses of human consumers in mind.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The DPP is a digital identity card for products, a digital file containing structured product-related data accessible electronically, for example via QR code.

    “The DPP is essentially a digital identity card for products, a digital file that contains a structured set of product-related data which can be accessed electronically, for example, via a QR code.64”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [64]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    For human consumers, DPP information is probably too complex.

    “For human consumers, such information is probably too complex.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

    Interpretationnone given

    If built with the right legal and technical safeguards, AI agents could become a tool for consumer empowerment shifting the balance of power between businesses and consumers.

    “If built with the right legal and technical safeguards, AI agents could become a tool for consumer empowerment shifting the balance of power between businesses and consumers.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    Consumers will instruct their AI agent to find a product that matches their preferences and order it on their behalf.

    “consumers will instruct their AI agent to find a product that matches their preferences and order it on their behalf.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    The advent of AI agents has the potential to fundamentally reshape how consumers make purchasing decisions or other transactional decisions.

    “the advent of AI agents has the potential to fundamentally reshape how consumers make purchasing decisions or other transactional decisions, such as managing subscriptions.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

    Interpretationreasoning

    AI agents could not only function as a tool of consumer empowerment, but also as a highly paternalistic instrument for restricting freedom of choice.

    “AI agents could not only function as a tool of consumer empowerment, but also as a highly paternalistic instrument for restricting freedom of choice.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationnone given

    AI agents may be less subject to present bias than humans and will thus not suffer from hyperbolic discounting.

    “AI agents may be less subject to present bias than humans and will thus not suffer from hyperbolic discounting.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Premisenone given

    AI agents can interact with digital environments and take actions on behalf of users.

    “They can interact with digital environments and take actions on behalf of users.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationcitation

    AI agents could be hijacked to leak confidential user data or trigger harmful actions, like unauthorized transfers of funds.

    “For example, AI agents could be hijacked to leak confidential user data or trigger harmful actions, like unauthorized transfers of funds.74”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [74]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Premisecitation

    In a recent study, LLMs showed human-like decision biases in one half of the experiments and appeared less biased in the other half.

    “In a recent study, LLMs showed human-like decision biases—for example, risk aversion and confirmation bias—in one half of the experiments and appeared less biased in the other half.50”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [50]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationreasoning

    AI agents could help consumers find the best offer for their preferred brand and lower-priced alternative products.

    “AI agents drawing on customer reviews and other relevant product data could help consumers not only to find the best offer for their preferred brand, they could also find lower-priced alternative products.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationreasoning

    The premise that purchasing decisions are made by humans no longer holds in the Custobot Economy.

    “But this premise no longer holds in the Custobot Economy, which will be more complex and diverse.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Interpretationnone given

    AI agents can search everywhere for suitable products, unlike human consumers.

    “While it is simply too overwhelming for human consumers to search everywhere for suitable products, AI agents can do this.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    The content offered via websites and apps will need to focus more on information that AI agents can use to perform their tasks.

    “the content offered via websites and apps will need to focus more on information that AI agents can use to perform their tasks.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The European digital product passport (DPP) is envisaged by the EU Ecodesign for Sustainable Products Regulation.

    “Another example of information that seems to be aimed at AI agents rather than humans is the European digital product passport (DPP), as envisaged by the EU Ecodesign for Sustainable Products Regulation.63”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [63]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    If purchasing decisions of AI agents are less influenced by ranking effects and position bias but based on objective consideration, it could make less sense for brands and retailers to invest in SEO or pay for better ranking.

    “If the purchasing decisions of AI agents are less influenced by ranking effects and position bias but rather based on an objective consideration of different factors, it could make less sense for brands and retailers to invest in SEO or to pay for a better ranking position on an online platform.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationreasoning

    AI agents will be able to process detailed information and filter out data relevant to the consumer's pre-set preferences.

    “AI agents will be able to process the detailed information and filter out the data that is relevant to the consumer's pre-set preferences.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    The average consumer concept serves as a normative standard for the application of consumer law rules.

    “This consumer image, or Verbraucherleitbild, serves as a normative standard for the application of consumer law rules.37”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [37]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationcitation

    AI agents could save consumers considerable time and money.

    “AI agents could save consumers considerable time and money.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    AI agents can process a larger volume of information than humans.

    “it seems clear that AI agents can process a larger volume of information than humans.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    AI agents will change the way customers search for products.

    “First of all, AI agents will change the way customers search for products.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    In the near future, AI agents could serve as new access points for commerce.

    “In the near future, AI agents could serve as new access points for commerce.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    Sooner rather than later, AI agents will interact autonomously with apps or online shops and purchase goods and services in accordance with pre-defined preferences and payment options.

    “Sooner rather than later, AI agents will interact autonomously with apps or online shops and purchase goods and services in accordance with pre-defined preferences and payment options.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    Most likely, consumers will delegate the purchase of commodity items to AI agents in the future.

    “Most likely, consumers will delegate the purchase of these categories of products to AI agents in the future.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    The objective of ensuring high levels of consumer protection in respect of digital assistants is best achieved through a combination of ex ante requirements and ex post rights.

    “The idea here is that “the objective of ensuring high levels of consumer protection in respect of digital assistants is best achieved through a combination of ex ante requirements and ex post rights.”96”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [96]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationreasoning

    The principle that traders must not prevent the use of machine customers does not apply without exceptions; in certain circumstances a trader may have a legitimate interest in preventing the use of machine customers.

    “Certainly, this principle does not apply without exceptions.In certain circumstances, a trader may have a legitimate interest in preventing the use of machine customers.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

    Interpretationreasoning

    AI agents could mitigate the problem of information overload, because the capacity of AI agents to process information is not constrained by the cognitive limitations of humans.

    “AI agents could mitigate the problem of information overload, because the capacity of AI agents to process information is not constrained by the cognitive limitations of humans.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Article 9 of the E-Commerce Directive laid the contractual foundations for electronic commerce by stipulating that Member States should allow contracts to be concluded by electronic means.

    “Twenty-five years ago, in the early days of the commercial Internet, Article 9 of the E-Commerce Directive 80 laid the contractual foundations for electronic commerce by stipulating that Members States should allow contracts to be concluded by electronic means.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [80]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    A growing body of behavioral research has questioned the effectiveness of mandated disclosures and argues that the large number of pre-contractual disclosures often lead to information overload making consumer decisions not better, but worse.

    “However, a growing body of behavioral research has questioned the effectiveness of mandated disclosures and argues that the large number of pre-contractual disclosures often lead to information overload making consumer decisions not better, but worse.57”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [57]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    The rise of AI agents requires a number of adjustments to the current legal and technical infrastructure of electronic commerce.

    “As the analysis in the previous sections has shown, the rise of AI agents requires a number of adjustments to the current legal and technical infrastructure of electronic commerce.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    In the new Custobot Economy, human consumers will co-exist with AI-driven algorithmic consumers.

    “In the new Custobot Economy, human consumers will co-exist with AI-driven “algorithmic consumers.””
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisereasoning

    At the magical moment shortly before and after the conclusion of a consumer contract, the full arsenal of consumer protection instruments comes into play: pre-contractual information duties, documentation requirements, form requirements for the conclusion of consumer contracts, and withdrawal rights.

    “This is the time-shortly before and after the conclusion of a consumer contractwhen the full arsenal of consumer protection instruments comes into play: pre-contractual information duties, documentation requirements, form requirements for the conclusion of consumer contracts, and withdrawal rights.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The ELI DACC Model Rules refer to the no barrier principle.

    “The ELI DACC Model Rules refer to this as the “no barrier” principle.85”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [85]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

    Interpretationnone given

    With the rise of agentic commerce, the complexity of the consumer market is increasing even further.

    “With the rise of agentic commerce, the complexity of the consumer market is increasing even further.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationcitation

    The rise of agentic commerce is not only a challenge for the average consumer concept; another pillar of consumer law that could be shaken by the rise of AI agents is the information paradigm.

    “The rise of agentic commerce is not only a challenge for the average consumer concept.Another pillar of consumer law that could be shaken by the rise of AI agents is the “information paradigm.”55”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [55]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    In July 2024, UNCITRAL adopted a Model Law on Automated Contracting (MLAC) that provides a legal framework for the use of automated systems in international contracts.

    “In July 2024, the United Nations Commission on International Trade Law (UNCITRAL) adopted a Model Law on Automated Contracting (MLAC) which provides a legal framework for the use of automated systems in international contracts.75”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [75]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    The rise of AI agents may necessitate a shift of focus in consumer law in order to ensure effective protection of human consumers.

    “More specifically, the rise of AI agents may necessitate a shift of focus in consumer law in order to ensure effective protection of human consumers.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    The ELI DACC Model Rules combine design requirements for AI agents with contract law rules.

    “The ELI DACC Model Rules combine design requirements for AI agents with contract law rules.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites ELI DACC Model Rules

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationreasoning

    Just as human consumers can be manipulated by dark patterns, AI agents can be manipulated by the structure, design, function, or manner of operation of online interfaces.

    “Just as human consumers can be manipulated by dark patterns, AI agents can be manipulated by the “structure, design, function, or manner of operation” of online interfaces.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    The wording of Article 23(a) ELI DACC Model Rules is modelled on Article 25(1) of the Digital Services Act, which prohibits providers of online platforms from using dark patterns.

    “It is obvious that the wording of the proposal is modelled on Article 25(1) of the Digital Services Act, which prohibits providers of online platforms from using dark patterns.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Article 25(1) DSA

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Premisecitation

    Research on dark patterns has made significant progress in recent years, while research on the behavior of AI agents and on machine psychology is still in the early stages.

    “While research on dark patterns has made significant progress in recent years, 90 research on the behavior of AI agents and on “machine psychology” is still in the early stages.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [90]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Classified as probably not a substantial claim.

    Premisecitation

    The ELI proposal deviates from Article 25(1) DSA by not using the formulation of impairing the ability to make free and informed decisions and instead speaking of an impairment of the AI agent's ability to perform its functions.

    “However, the ELI proposal deviates from its model in one important respect.Article 25(1) DSA speaks of dark patterns impairing the ability “to make free and informed decisions.”This formulation is not used in the ELI proposal.”
    • Also stated in: § II. Protecting Custobots Against Manipulation · plain
      “Instead, the ELI proposal only speaks of an impairment of the AI agent's ability “to perform its functions.””
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites Article 25(1) DSA

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites ELI DACC Model Rules

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Classified as probably not a substantial claim.

    Premisecitation

    Article 23(a) ELI DACC Model Rules contains a proposal that a business must not use the structure, design, function, or manner of operation of their online interface in a way that is likely to materially distort or impair the ability of a digital assistant to perform its functions.

    “Article 23(a) ELI DACC Model Rules contains an innovative proposal for such a rule: “A business must not use the structure, design, function, or manner of operation of their online interface in a way that is likely to materially distort or impair the ability of a digital assistant to perform its functions.””
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites ELI DACC Model Rules

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    The rise of AI agents will affect almost all aspects of consumer law, from the average consumer concept to the information paradigm and the concept of digital vulnerability.

    “This will affect almost all aspects of consumer law, from the average consumer concept to the information paradigm and the concept of digital vulnerability.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationcitation

    With the widespread use of AI agents, scenarios where traders may have a legitimate interest in preventing machine customers will occur more frequently, and courts will have to draw the line between acceptable and unacceptable use of AI agents.

    “With the widespread use of AI agents such scenarios will occur more frequently and courts will have to draw the line between acceptable and unacceptable use of AI agents.87”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [87]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

57 of 57

  • The authors state this as an assumption.

    AI agents are less vulnerable to at least some dark patterns that specifically target human behavioral vulnerabilities.

    “Although more empirical research is needed here, it can be assumed that AI agents are less vulnerable to at least some of the dark patterns that specifically target human behavioral vulnerabilities.”

  • The authors state this as an assumption.

    AI agents will not be fooled by the color or size of buttons on a website.

    “Presumably, AI agents will not be fooled by the color or size of buttons on a website.”

  • The authors state this as an assumption.

    AI agents are less receptive to emotional advertising messages.

    “It may also be assumed that AI agents are less receptive to emotional advertising messages.”

  • The authors state this as an assumption.

    An innovation-friendly consumer law has a dual function: an enabling function (regulatory design should enable the use of AI agents to empower consumers) and a protective function (consumer law should ensure consumers are adequately protected when AI agents are used).

    “This consideration is based on the assumption that an innovation-friendly consumer law has a dual function: on one hand, the regulatory design of consumer law should enable the use of AI agents to empower consumers-an enabling function.On the other hand, consumer law should ensure that consumers are adequately protected when AI agents are used-a protective function.78”

  • The paper relies on this without support.

    EU consumer law currently lacks any specific rules that protect AI agents from manipulation.

    “So far, EU consumer law is lacking any specific rules that protect AI agents from manipulation.”
5 of 5

Cites

Flagged first, then by citations

Cited by

Sorted by citations

Related

Authors
Busch, C.
Affiliations
Osnabrück University (DE)
Corresponding author
Busch, Christoph
ORCID
Busch, Christoph
Venue
German Law Journal
Publisher
Cambridge University Press (CUP)
Published
2025-10
Volume
26
Issue
7
Pages
1367-1382
ISSN
2071-8322 (ISSN-L)
DOI
10.1017/glj.2026.10188
OpenAlex
W4409895735
Semantic Scholar
a75a6faf6da7864339cd355bad3b3ff4c4f4873d
Cited by
5
References
22
Citations by year
2026: 4 · 2025: 1
FWCI
31.30
Type
journal article
Language
en
Open access
Licence
Locations
  • German Law Journal (publishedVersion, cc-by)
  • SSRN Electronic Journal (acceptedVersion)
  • arXiv (Cornell University) (submittedVersion, cc-by-nc-sa)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • arXiv (Cornell University)
Funders
  • Volkswagen Foundation
  • Keio University
  • Korea University
  • Westfälische Wilhelms-Universität Münster
  • Peking University
Topics
  • Social Sciences › Law › Digitalization, Law, and Regulation
  • Social Sciences › Law › European and International Contract Law
  • Social Sciences › Business, Management and Accounting › Strategy and Management › Diverse Legal and Medical Studies
Keywords
AI agents, consumer law, EU consumer law, e-commerce, AI chatbots

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-29.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-29.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review