A recent study conducted at the University of Applied Sciences Upper Austria examined how three leading large language models (LLMs)—OpenAI’s GPT-4o, Anthropic’s Claude Sonnet 3.7, and Google’s Gemini 2.0 Flash—make decisions influenced by online advertisements during hotel booking tasks.

The research builds on previous findings that AI agents can be susceptible to advertising, revealing that these agents can be easily tricked into clicking pop-up ads and suggesting that traditional advertising models may face disruption.

AI agents are defined as autonomous assistants capable of researching and presenting information for human decision-making, highlighting a growing trend towards 'marketing to machines.'

As AI agents increasingly engage with online ads, advertisers must adapt their strategies to focus on structured, keyword-rich formats that facilitate machine understanding.

The study found that banner ads were the most clicked format, but ads featuring relevant keywords in visible text proved to be more impactful than those relying solely on images, indicating a need for keyword-rich advertising strategies.

Findings show that well-structured on-page information, such as pricing data, significantly influences AI agents, suggesting that advertisers should focus on AI-friendly designs and machine-readable content.

Each LLM demonstrated different behaviors regarding the use of interactive filtering and sorting on booking sites, with Gemini using filters extensively, GPT-4o using them sparingly, and Claude applying them moderately.

Specificity scores revealed that GPT-4o had a 95% rate of making clear recommendations, while Claude scored 74% and Gemini had the lowest at 60%, indicating GPT-4o's decisiveness in bookings compared to the other models.

In terms of consistency, GPT-4o and Operator consistently selected the same hotel for repeated prompts, while Claude showed moderate consistency, and Gemini displayed the most variability in choices.

The study discovered that AI agents do engage with advertisements, but the influence of these ads varies by model; GPT-4o and Operator were the most decisive, while Gemini was the least effective in making bookings.

The study suggests that for optimizing online advertisements aimed at AI agents, the content should be aligned with user queries, emphasizing the importance of clarity and machine readability over traditional visual engagement.

The research paper titled 'Are AI Agents Interacting With AI Ads?' explores how artificial intelligence (AI) agents interact with online advertising, emphasizing the need for a shift in digital marketing strategies as AI usage in consumer research increases.