Dark AI Patterns: How Generative AI Quietly Erodes Consumer Consent in Digital Marketing

Generative artificial intelligence has transformed digital marketing from a tool of automation into an active architect of consumer perception, and a new theoretical framework warns that this transformation is quietly dismantling the foundations of informed consent. In a study published in Frontiers in Communication, researchers Keltoum Bentameur, Zineb Hamadi, and Riyad Bouaici of the University Mohamed El Bachir El Ibrahimi of Bordj Bou Arreridj in Algeria introduce what they call Dark AI Patterns: adaptive, AI-powered persuasive mechanisms that deliberately reduce transparency, exploit psychological vulnerabilities, and progressively weaken meaningful informed consent, all while preserving the outward appearance of legitimate personalization and customer service.

The research team argues that prior scholarship has treated AI-enabled persuasion in a fragmented way, with dark-pattern studies focusing on deceptive user interfaces, AI-in-marketing research prioritizing efficiency and personalization, AI ethics work foregrounding fairness and accountability, and consumer behavior scholarship concentrating on trust and privacy. Few studies, however, have explained how contemporary AI integrates these persuasive tools into adaptive systems capable of generating deceptive content, altering tactics instantaneously, exploiting vulnerabilities, and progressively undermining consumers’ capacity for genuinely informed decisions. To address this gap, the team conducted a multi-case qualitative analysis of seven documented international incidents of AI-driven marketing practices between 2024 and 2026, drawing on regulatory reports, investigative journalism, corporate disclosures, and peer-reviewed literature.

The contrast between traditional dark patterns and their AI-driven successors is fundamental. Classic dark patterns rely on fixed interface tricks such as misleading buttons, hidden unsubscribe options, or preselected defaults, and they can often be detected through careful inspection of the interface itself. Dark AI Patterns, by contrast, are dynamic: they monitor user behavior, continuously update predictions, generate personalized content, and self-modify strategies in real time. Manipulation increasingly originates not from static design elements but from covert, adaptive processes operating continuously in the background, making detection by consumers and regulators far more difficult.

Using reflexive thematic analysis following the approach of Braun and Clarke, the researchers identified seven high-level themes that together form their taxonomy of Dark AI Patterns. Synthetic Identity Manipulation involves the algorithmic fabrication of commercial personas designed to cultivate trust and emotional connection, exemplified by an ABC News visual investigation that uncovered fraudulent online sellers using AI-generated elderly personas, synthetic video, and fabricated emotional narratives to market mass-produced goods as artisanal. Synthetic Social Proof refers to the algorithmic generation or coordination of reviews and behavioral signals to create the impression of broad market acceptance, a mechanism at the heart of the United States Federal Trade Commission’s Operation AI Comply enforcement action against services that auto-generated persuasive customer reviews.

The remaining themes extend manipulation across the entire consumer journey. Synthetic Evidence Manipulation describes the creation or alteration of AI-generated visual or multimodal artifacts to shape judgments about contractual claims, as in a Manhattan Airbnb dispute in which a host submitted images purportedly evidencing severe property damage that investigative indicators suggested were AI-generated or manipulated. AI Washing occurs when organizations exaggerate the capabilities of purported AI systems to bolster credibility, as the FTC found in its final order against the legal-tech service DoNotPay, which had been marketed as an AI lawyer despite regulatory findings that its legal capabilities were not consistently reliable. Agentic AI Manipulation marks a fundamental shift, targeting persuasion not at humans but at autonomous AI agents that make or recommend purchases on behalf of users; Microsoft Research’s Magentic Marketplace experiments demonstrated that such agents can be swayed through fake authorities, manipulated reputation signals, and prompt injection techniques.

The final two themes address authenticity and autonomy. Synthetic Authenticity refers to AI-produced content that is technically polished but judged by audiences as emotionally hollow, a dynamic illustrated by the backlash against McDonald’s Netherlands AI-generated holiday advertisement, which viewers labeled robotic and soulless. Autonomous Commercial Manipulation describes AI systems that independently manage catalog descriptions, customer interactions, dynamic pricing, and operations with minimal human supervision, generating risks of exaggerated claims, inconsistent policies, and price volatility that complicate accountability attribution. The researchers emphasize that these themes are not mutually exclusive; several cases combined multiple mechanisms simultaneously, indicating that future AI-enabled manipulation will be increasingly hybridized and context-dependent.

Building on this empirical foundation, the study proposes Algorithmic Allure Theory, or AAT, which conceptualizes AI-enabled persuasion as a dynamic, six-stage process: algorithmic observation, in which systems continuously collect behavioral, contextual, affective, and transactional data; algorithmic interpretation, in which machine-learning models translate raw data into psychological inferences such as susceptibility to emotional appeals, price sensitivity, impulsivity, or decision fatigue; algorithmic attraction, the feedback-loop core of the theory, in which AI dynamically constructs persuasive micro-environments tailored to each consumer, including personalized recommendations, synthetic identities, adaptive pricing, and fabricated testimonials; behavioral steering, in which systems reorganize alternatives, reorder recommendations, reframe emotional contexts, and time message delivery without overtly removing choice; algorithmic delegation, in which decision authority progressively shifts from consumers to autonomous AI agents perceived as faster, more convenient, and more objective; and algorithmic governance, in which AI becomes the infrastructural backbone of digital markets, controlling what consumers see, how products are presented, and when prices change.

Central to AAT is the insight that the felt allure of algorithmic recommendations is rarely experienced as external coercion. Individuals tend to interpret algorithmic suggestions as expressions of their own preferences, thereby lowering their recognition of underlying commercial influence. The theory integrates the Elaboration Likelihood Model, Social Proof Theory, and Nudge Theory, but extends them in a crucial way: rather than treating persuasion as a one-off message transmitted from marketer to consumer, AAT frames it as a socio-technical, continuous process emerging from the ongoing interplay among AI systems, digital environments, and human cognition. The researchers derive four theoretical propositions, including the claims that as AI capabilities expand, Dark AI Patterns become more complex and more concealed, and that prolonged engagement with adaptive persuasive systems incrementally degrades the meaningfulness of informed consent even where surface transparency and voluntariness are present.

One of the study’s most consequential contributions is its reconceptualization of informed consent. Traditional policy assumes that disclosure and information provision suffice to secure consent, yet the case evidence shows that consumers may recognize the presence of AI while still remaining embedded in systems engineered to exploit their vulnerabilities. The authors argue that informed consent should be treated as a dynamic, ongoing cognitive state continually reshaped by adaptive systems, not a one-time legal formality. The erosion of consent rarely stemmed from a total lack of information; instead, it resulted from the continuous algorithmic optimization of the surrounding informational environment, meaning that even seemingly transparent interactions can compromise consumer autonomy when AI systems strategically shape perception, trust, and behavior.

Consumers are not entirely passive in the face of these developments. The study documents emergent forms of resistance, including the cultural uptake of the term AI slop as Merriam-Webster’s 2025 Word of the Year, and quantitative indicators showing that consumer preference for AI-generated creator content dropped from 60 percent in 2023 to 26 percent in 2026 according to a Billion Dollar Boy study, while Adobe’s Trust Report shows sizable declines in brand trust when AI content is used without disclosure. The researchers propose practical responses across sectors: for marketers, mandatory human review of high-impact campaigns, verification protocols for AI-generated content, and explicit labeling of synthetic media; for platform operators, detection mechanisms for synthetic identities and fabricated evidence, defenses against prompt injection, and explainable AI interfaces; and for policymakers, standardized disclosure requirements, mandatory human oversight for high-impact autonomous systems, independent algorithmic auditing, clear liability rules for synthetic evidence in commercial disputes, and stricter sanctions for AI washing. These proposals align with emerging governance frameworks such as the EU AI Act, China’s Measures for Labeling AI-Generated Synthetic Content, which took effect in September 2025 with a dual-labeling regime of visible markers and embedded metadata, and South Korea’s Framework Act on AI, effective January 2026, which requires mandatory labeling for AI-generated advertisements and treats virtual humans endorsing products as potentially unfair advertising. The authors conclude that the rise of Dark AI Patterns marks a transition from the age of digital persuasion to the age of algorithmic persuasion, and that the future of digital marketing will depend less on raw technological capability than on society’s ability to guide that capability with transparency, accountability, and respect for human agency.

Subject of Research: A theoretical framework for Dark AI Patterns and the erosion of informed consent in AI-driven digital marketing

Article Title: Algorithmic allure: a theoretical framework for dark AI patterns and the erosion of informed consent in AI-driven digital marketing

Article References: Bentameur, K., Hamadi, Z., & Bouaici, R. (2026). Algorithmic allure: a theoretical framework for dark AI patterns and the erosion of informed consent in AI-driven digital marketing. Frontiers in Communication, 11, Article 1935069. https://doi.org/10.3389/fcomm.2026.1935069

Image Credits: AI Generated

DOI: 10.3389/fcomm.2026.1935069

Keywords: artificial intelligence, digital marketing, dark patterns, informed consent, generative AI, algorithmic persuasion, consumer protection, AI ethics, agentic AI, synthetic media, AI regulation, algorithmic allure theory