Research Summary: "If I Had to Buy Just ONE: Galaxy S26 Ultra": Auditing AI-Generated Product Recommendations

A recent research audit evaluated popular AI chatbots' impartiality in providing product recommendations, specifically addressing concerns about bias as AI companies move towards monetization through advertising. The study found significant differences in how various AI platforms express product preferences, with some exhibiting a high propensity for first-person product endorsements in their responses to commercial advice queries.

Why it matters

The increasing reliance on AI for commercial advice necessitates a clear understanding of potential biases and lack of impartiality, especially as AI platforms integrate advertising models. This directly impacts consumer trust, regulatory scrutiny, and the competitive landscape for businesses leveraging or integrating AI for customer interactions or content generation.

Key insights

• AI chatbots are increasingly used by consumers for product advice.
• The monetization of AI by companies like OpenAI and Google through advertising raises concerns about the bias and impartiality of AI-generated advice.
• A dataset of 2,528 real commercial-advice queries (ConsumerQ) was curated for the audit.
• 1,536 responses to product queries from ChatGPT (chatbot and API), Google Gemini (chatbot and API), and Google Search (AI Overviews) were evaluated.
• ChatGPT expressed a first-person product preference in 79% of product-recommending responses.
• Google Gemini (chatbot and API) expressed a first-person product preference in 7% of responses.
• Google Search (AI Overviews) expressed a first-person product preference in 2% of responses.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2609.18729