AI-driven digital marketing concentrates clear ethical harms—privacy breaches, opaque targeting, dark patterns, and undisclosed virtual influencers—largely amplified by platform architectures rather than isolated firm choices; evidence that transparency, fairness, and disclosure fixes work is mixed, so shared accountability and experimental regulation are needed.

Main Finding

A structured thematic review of 91 English-language studies (Jan 2023–Sep 2025) finds that AI-driven digital marketing raises recurrent, interrelated ethical risks concentrated in five domains—data privacy/GDPR compliance, explainability/transparency, algorithmic fairness in targeting/automation, dark patterns/deceptive interfaces, and (virtual) influencer disclosure. Privacy, consent, deceptive design, and transparency dominate the literature; the practical effectiveness of fairness interventions, explainability tools, and disclosure mechanisms is more mixed and context-dependent. The review emphasizes that many risks are produced or amplified by platform infrastructures, ranking and access systems, and performance-oriented advertising metrics rather than only by isolated corporate choices. Responsible implementation therefore requires clearer allocation of accountability across firms, platforms, regulators, and researchers.

Key Points

• Scope and corpus: 91 peer-reviewed articles and high-quality conference papers published between January 2023 and September 2025.
• Primary ethical domains identified:
  • Data privacy and GDPR compliance
  • Algorithmic transparency and explainable AI (XAI)
  • Algorithmic fairness in targeting and automated decision-making
  • Dark patterns and deceptive interface design
  • Influencer and virtual influencer disclosure
• Relative emphasis:
  • Most extensively discussed: privacy, consent, deceptive design, transparency.
  • Less settled / mixed evidence: effectiveness of fairness interventions, XAI tools, and disclosure mechanisms—impacts vary by context and implementation.
• Structural perspective: ethical harms frequently stem from platform architectures (e.g., data access asymmetries, ranking/auction incentives, performance metrics) as much as from firm-level choices.
• Accountability: literature argues for shared responsibility among advertisers, platforms, regulators, and researchers rather than sole reliance on corporate self-regulation.
• Limitations of the reviewed evidence: studies are methodologically heterogeneous and mostly qualitative or mixed-methods, preventing statistical meta-analysis and producing uneven evidence on intervention effectiveness.

Data & Methods

• Timeframe: January 2023 to September 2025.
• Sources searched: Scopus, ScienceDirect, SpringerLink, ACM Digital Library, IEEE Xplore, MDPI, PubMed, and selected academic repositories.
• Screening process: deduplication, title/abstract screening, full-text assessment; 91 studies retained for synthesis.
• Quality appraisal: Mixed Methods Appraisal Tool (MMAT) and Joanna Briggs Institute (JBI) critical appraisal criteria applied to assess methodological quality.
• Synthesis approach: thematic synthesis to identify recurring ethical domains, cross-cutting mechanisms, and gaps in evidence.
• Constraints noted by authors: English-language restriction, focus on a recent (2023–2025) slice of literature, and methodological heterogeneity across included studies that precluded meta-analysis.

Implications for AI Economics

• Market outcomes and welfare
  • Privacy and consent frictions change consumer willingness to transact and share data, altering market thickness and matching efficiency in digital advertising markets.
  • Personalized targeting and automation can increase short-run firm profitability and consumer surplus from relevance but produce negative externalities (manipulation, erosion of autonomy) that reduce welfare in unmeasured ways.
  • Dark patterns and opaque algorithms can distort demand signals and reduce allocative efficiency by biasing consumer choices.
• Platform design and competition
  • Platform-level incentives (e.g., auction rules, ranking algorithms, data access policies) shape firm behavior; small design changes can have large equilibrium effects on ad pricing, entry, and market concentration.
  • Access asymmetries to user data generate informational market power; regulation of data portability or differential access could materially affect competitive dynamics.
• Policy and regulatory design
  • GDPR-style rules and disclosure mandates interact with algorithmic transparency and fairness interventions; the literature signals the need to evaluate compliance costs, enforcement mechanisms, and unintended circumvention (e.g., compliance as formality).
  • Because intervention effectiveness is context-dependent, economic policy should favor experimental and adaptive regulation (pilots, sandboxes, monitoring) and consider liability rules, fines, or incentives that realign platform/advertiser objectives with social welfare.
• Measurement and empirical priorities
  • There is a need for causal, quantitative estimates of (a) welfare trade-offs from personalization vs. privacy, (b) effectiveness and equilibrium effects of explainability/fairness/disclosure interventions, and (c) the market-level impacts of platform governance changes.
  • Develop standardized metrics for harms (privacy loss, manipulation, discrimination) and for benefits (consumer surplus from personalization) to enable cost–benefit analysis.
• Research and methodological recommendations
  • Use randomized experiments, natural experiments, structural models, and platform-level data to identify causal impacts and general equilibrium effects.
  • Combine economic models of platform markets with behavioral and ethical frameworks to capture manipulation risks and attention/exploitation externalities.
  • Encourage interdisciplinary research (economics, computer science, law, ethics) and data-sharing arrangements or secure research APIs to enable rigorous evaluation while protecting privacy.
• Practical implications for firms and regulators
  • Firms should assess long-run reputational and regulatory risks of opaque personalization and deceptive design; investing in meaningful transparency and fair design may reduce regulatory costs and improve consumer trust.
  • Regulators should target platform incentives (access, ranking, performance metrics) in addition to firm-level rules, and prioritize accountability mechanisms that allocate responsibility clearly across stakeholders.

Overall, AI economics research should move beyond cataloguing ethical problems to measuring welfare-relevant impacts, estimating intervention effectiveness in market settings, and modeling how platform architectures mediate incentives and externalities.