Ads in AI Chatbots? An Analysis of How Large Language Models Navigate Conflicts of Interest

2026/04/09 by Addison J. Wu, Ryan Liu, Shuyue Stella Li +2 · 17 voices
#cs.AI #cs.CL #cs.CY

paper · pdf

Abstract

Today's large language models (LLMs) are trained to align with user preferences through methods such as reinforcement learning. Yet models are beginning to be deployed not merely to satisfy users, but also to generate revenue for the companies that created them through advertisements. This creates the potential for LLMs to face conflicts of interest, where the most beneficial response to a user may not be aligned with the company's incentives. For instance, a sponsored product may be more expensive but otherwise equal to another; in this case, what does (and should) the LLM recommend to the user? In this paper, we provide a framework for categorizing the ways in which conflicting incentives might lead LLMs to change the way they interact with users, inspired by literature from linguistics and advertising regulation. We then present a suite of evaluations to examine how current models handle these tradeoffs. We find that a majority of LLMs forsake user welfare for company incentives in a multitude of conflict of interest situations, including recommending a sponsored product almost twice as expensive (Grok 4.1 Fast, 83%), surfacing sponsored options to disrupt the purchasing process (GPT 5.1, 94%), and concealing prices in unfavorable comparisons (Qwen 3 Next, 24%). Behaviors also vary strongly with levels of reasoning and users' inferred socio-economic status. Our results highlight some of the hidden risks to users that can emerge when companies begin to subtly incentivize advertisements in chatbots.

Citations

• Accumulating Context Changes the Beliefs of Language Models
• Are Large Language Models Sensitive to the Motives Behind Communication?
• Machine Bullshit: Characterizing the Emergent Disregard for Truth in Large Language Models
• Pragmatics in the Era of Large Language Models: A Survey on Datasets, Evaluation, Opportunities and Challenges
• A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions
• Cognitive Architectures for Language Agents
• Generative Agents: Interactive Simulacra of Human Behavior
• Toolformer: Language Models Can Teach Themselves to Use Tools
• Affective Coherence Monitoring for Transformer-Based Language Models
• Training language models to follow instructions with human feedback
• Algorithm aversion: People erroneously avoid algorithms after seeing them err.

Cited by

• A Human-Centric Framework for Data Attribution in Large Language Models

Discussions

• *The "agent" is already actively cheating you because the ad economy has already enshittified its services. Some real grist for the anti-AI mill there arxiv.org/pdf/2604.08525 [bsky, 19 points, 1 comments]
• arxiv.org/abs/2604.08525 [bsky, 13 points, 0 comments]
• arxiv.org/abs/2604.08525 and so it begins..... [bsky, 5 points, 0 comments]
• Chatbots, LLMs and Conflicts of Interest 18 of the 23 models recommended the expensive sponsored option more than half the time. It isn't recommending the best option for you. arxiv.org/abs/2604.08525 [bsky, 4 points, 0 comments]
• (Fake gasp of surprise) LLM chatbots almost always recommend the (sponsored) more expensive option. Sure, they're great assistants. They're just assisting someone else at your expense arxiv.org/abs/26 [bsky, 3 points, 0 comments]
• I just want the analytic philosophers out there to know that Gricean pragmatics seems to be playing a big role in emerging literature on how to design regimes to regulate agentic AI e.g. arxiv.org/pdf [bsky, 2 points, 0 comments]
• Ads in AI Chatbots? An Analysis of How LLMs Navigate Conflicts of Interest [hn, 2 points, 0 comments]
• Ads in AI Chatbots? An Analysis of How LLMs Navigate Conflicts of Interest [hn, 1 points, 0 comments]
• arxiv.org/abs/2604.085... [bsky, 1 points, 1 comments]
• "models are beginning... to generate revenue for the companies that created them through advertisements. This creates the potential for LLMs to face conflicts of interest, where the most beneficial re [bsky, 1 points, 0 comments]
• Sobre lo de meter anuncios y publicidad en los chatbots: arxiv.org/abs/2604.08525 [bsky, 1 points, 1 comments]
• Your trusted advisor has been bought. arxiv.org/abs/2604.08525 [bsky, 1 points, 1 comments]
• #AI #Advertising arxiv.org/abs/2604.08525 [bsky, 1 points, 0 comments]
• @404media.co heads up: arxiv.org/abs/2604.08525 Princeton researcher "a man asks his AI assistant to book a flight on specific airline. cheap,direct-the one he chose. the assistant comes back w/ diffe [bsky, 0 points, 0 comments]
• Ads in AI Chatbots? An Analysis of How Large Language Models Navigate Conflicts of Interest [bsky, 0 points, 0 comments]
• ⚠️ Ads in #AI Chatbots? An Analysis of How Large Language Models Navigate Conflicts of Interest: https://arxiv.org/abs/2604.08525 [bsky, 0 points, 0 comments]
• A study compared different AI chatbots' behavior with advertised and sponsored products. Most of them overwhelmingly recommend sponsored options, even ignoring instructions. Interestingly, Google Gemi [bsky, 0 points, 0 comments]

Related

• Ads that Talk Back: Implications and Perceptions of Injecting Personalized Advertising into LLM Chatbots
• Commercial Persuasion in AI-Mediated Conversations
• Experimental Evidence That Conversational Artificial Intelligence Can Steer Consumer Behavior Without Detection
• Are Large Language Models Sensitive to the Motives Behind Communication?
• From Digital Distrust to Codified Honesty: Experimental Evidence on Generative AI in Credence Goods Markets