Why AI Agents Ignore Your Website’s Design and Ads—And What Marketers Must Do Next

AI Agents See Websites and Ads Differently Than Humans

As AI-powered software agents become more common, marketers need to rethink how websites and ads are designed. These agents—AI models equipped with browsing and multimodal capabilities—don't respond to visual design or emotional appeals like humans do. Instead, they focus on structured data such as prices, availability, and product specifications.

Recent research from the Digital Media Lab at the University of Applied Sciences Upper Austria sheds light on how these AI agents interact with online ads in unexpected ways. The study explored how various AI agents perform tasks like hotel searches and bookings on travel websites, revealing that their interaction with ads and site elements is very different from human behavior.

Accessibility Matters for AI Agents, But Ads Need Rethinking

According to Andreas Stöckl, one of the researchers, websites optimized for accessibility tend to work better for AI agents. Accessibility features that help users with disabilities also benefit these agents, enabling smoother interaction with site content.

However, the design and placement of online ads need reconsideration. Traditional visual ads, especially image-based banners, often fail to capture agents’ attention. Unlike humans who respond to eye-catching visuals and emotional triggers, agents rely on clear, structured, and text-based information.

How AI Agents Handle Ads and Booking Tasks

The study tested three AI agents equipped with multimodal models capable of interpreting screenshots: OpenAI’s Operator with GPT-4o, Anthropic’s Claude 3.7 Sonnet, and Google’s Gemini 2.0 Flash. These agents were tasked with autonomously searching and booking hotels based on user preferences like destination and price range.

Findings showed that AI agents often missed image-only call-to-action ads. For example, Google’s Gemini 2.0 Flash struggled with image-based CTAs because it couldn't easily determine if those elements were clickable, leading to extra steps in the booking process.

Interestingly, Gemini 2.0 Flash was an exception: it interacted more with image-only banners and showed increased booking specificity, even though it engaged less with promotional language overall.

Text-Based Ads Outperform Image-Heavy Formats

The research compared three ad formats on a hotel booking site:

• Standard text-based banner ads
• Image-only banner ads
• Image banners with embedded promotional text

Text-based ads received more attention and clicks from the agents. For instance, Claude 3.7 Sonnet clicked on 59 text-based banners but ignored sponsored content entirely. GPT-4o clicked 59 banners and interacted 12 times with sponsored ads, while OpenAI's Operator clicked 47 banners and had 20 interactions with sponsored content.

The models' interaction with ads often depended on keyword relevance to the search query. Ads relying on embedded text within images performed worse, showing that clear, accessible text remains the most effective format for AI agents.

Bias in AI Agent Behavior Reflects Human Social Norms

The study uncovered subtle biases in how agents handled bookings for different types of couples. Claude 3.7 Sonnet, for example, recommended longer stays for married couples compared to dating couples, with Gemini 2.0 Flash showing a similar pattern. GPT-4o showed minimal difference.

This suggests that AI agents inherit social biases embedded in their training data, which can influence their recommendations and decisions.

Implications for Advertisers and Marketers

Ad formats that work well for humans may not be effective for AI agents. Stöckl emphasized that while new AI technology doesn’t necessarily increase invalid ad clicks, some ad designs are unsuitable for agent interaction.

Developing new ad formats that cater specifically to AI agents could open new opportunities. If agents acting on behalf of users can be influenced through targeted advertising, marketers could tap into a novel strategic channel.

Simon James, a data science and AI expert, highlighted that AI agents don’t browse—they execute commands. Unlike humans who might be drawn to images or catchy slogans, agents filter out what they consider noise to focus on completing tasks efficiently.

This means marketers must adapt their strategies to focus less on emotional appeal and more on clear, structured data that agents can process effectively.

Preparing for an AI-Agent-Driven Web

As AI agents become integral to online interactions, businesses should consider:

• Optimizing websites for accessibility to improve agent compatibility
• Using clear, text-based advertising formats rather than relying on images alone
• Reevaluating ad placement and design to engage AI agents effectively
• Monitoring AI-driven behaviors to identify and mitigate bias in recommendations

Understanding how AI agents operate can help marketers create more effective digital experiences that serve both humans and automated assistants.

For those interested in expanding their knowledge on AI technologies and their impact on marketing and development, Complete AI Training offers relevant courses tailored to various skill levels and roles.