When Ads Become Profiles: Large-Scale Audit of Algorithmic Biases and LLM Profiling Risks

2025-09-23

AI Summary

Social media advertising targeting mechanisms lack transparency, raising concerns about algorithmic bias and unintended inference of sensitive attributes. This paper introduces the first multi-stage socio-technical auditing framework for automated advertising systems, integrating large-scale ad exposure log analysis (435,000 ads across 891 Australian users), multimodal large language model (LLM) reasoning, and statistical modeling. We empirically demonstrate that ad feed composition constitutes a “digital footprint” exploitable by publicly available AI: LMs reconstruct user demographic attributes from ad sequences alone with accuracy surpassing census baselines and approaching human judgment. Furthermore, we identify disproportionate exposure of marginalized groups to gambling and political advertisements. These findings reveal dual systemic risks—algorithmic discrimination and privacy leakage—and establish a reproducible, interdisciplinary evaluation paradigm for auditing advertising algorithms.

Technology Category

• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy
• Machine Learning: Privacy
• Application Domains: Humanities & Computational Social Science

Application Category

• User Modeling, Personalization and Recommendation: User modeling for targeted and personalized online advertising
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Social Networks and Social Media: Fairness and bias in social network and social media analysis

Abstract

Automated ad targeting on social media is opaque, creating risks of exploitation and invisibility to external scrutiny. Users may be steered toward harmful content while independent auditing of these processes remains blocked. Large Language Models (LLMs) raise a new concern: the potential to reverse-engineer sensitive user attributes from exposure alone. We introduce a multi-stage auditing framework to investigate these risks. First, a large-scale audit of over 435,000 ad impressions delivered to 891 Australian Facebook users reveals algorithmic biases, including disproportionate Gambling and Politics ads shown to socioeconomically vulnerable and politically aligned groups. Second, a multimodal LLM can reconstruct users' demographic profiles from ad streams, outperforming census-based baselines and matching or exceeding human performance. Our results provide the first empirical evidence that ad streams constitute rich digital footprints for public AI inference, highlighting urgent privacy risks and the need for content-level auditing and governance.

Problem

• Auditing algorithmic biases in automated social media ad targeting systems
• Investigating LLM profiling risks from ad exposure data alone
• Addressing privacy risks from AI inference of sensitive user attributes

Innovation

• Large-scale audit of algorithmic biases in ad targeting
• Multimodal LLM reconstructs demographic profiles from ad streams
• Content-level auditing framework for privacy risk assessment