Marketing to AI Agents is the Future – Research Shows Why

Marketing to AI Agents is the Future – Research Shows Why

As per the new research paper, AI agents are interacting with online advertising, shaping their decision-making.

AI agents are transforming how marketing teams operate, shifting the focus from reactive data analysis to proactive, autonomous execution. They assess situations, reason through decisions, and take appropriate action on behalf of marketers, especially when managing time-consuming and complex tasks.

AI agents and Digital marketing

Traditional marketing approaches operate in cycles: plan, execute, measure, adjust, and repeat, which creates inevitable delays between identifying opportunities and acting on them. AI agents break this pattern by allowing continuous, proactive optimization.

Consider paid advertising management. In a traditional approach, a marketing team might review campaign performance weekly, making manual adjustments based on what happened last week. An AI agent, by contrast, monitors performance continuously in real time, spotting trends as they emerge and making adjustments immediately as needed.

As AI agents take on more execution-focused tasks, marketing teams are evolving toward a new division of labor that plays to the strengths of both humans and machines. AI agents thrive at processing large datasets, identifying patterns, executing precision tasks, and maintaining consistency across channels.

Compelling economics of AI agents

Economics case for AI agents is transforming marketing budget discussions from cost-centred to investment-centred. As per the research referenced by Ethan Mollick, 80% of tasks performed by AI agents cost less than 10% of what it would cost for human experts to do the same work.

With agent capabilities doubling approximately every seven months, the economic advantage continues to grow.

The following are the three scenarios where economics are compelling:-

1. Scale challenges
2. Speed requirements
3. Data complexity

Research titled as Are AI Agents Interacting With AI Ads? Conducted at the University of Applied Sciences Upper Austria, cited previous research on the interaction between AI Agents and online advertising for exploring the emerging relationships between agentic AI and machines driving display advertising.

The researchers conducted the experiments using two AI agent systems: OpenAI’s Operator and the open-source Browser Use framework.

Browser Use enables researchers to control the model that is used for the testing by connecting three different LLMs via API:

• GPT-4o
• Claude Sonnet 3.7
• Gemini 2.0 Flash

This setup enabled consistent testing across models by enabling them to use the page’s rendered HTML structure (DOM tree) and recording their decision-making behaviour.

Research Discovery

Ad Engagement

AI agents do not ignore online advertisements, but their engagement with ads and the extent to which those ads influence decision-making vary depending on the LLM.

• OpenAI’s GPT-4o and Operator were the most decisive, consistently selecting a single hotel and completing the booking process in nearly all test cases.
• Anthropic’s Claude Sonnet 3.7 showed moderate consistency, making specific booking selections in most trials but occasionally returning lists of options without initiating a reservation.
• Google’s Gemini 2.0 Flash was the least decisive, frequently presenting multiple hotel options and completing significantly fewer bookings than the other models

Banner Ads

Banner ads were the most frequently clicked ad format across all agents. But the impact of the presence of relevant keywords has more than visuals.

Text-Based Ads

Ads with keywords embedded in visible text effectively influence model behaviour than image-based text. GPT-4o and Claude are more responsive to keyword-based ad content, with Claude integrating more promotional language into its output.

Use of Filtering and Sorting Features

• Gemini extensively applied filters, often by combining multiple filter types across trials
• GPT-4o rarely uses filters
• Claude uses filters more than GPT-4o, but not as systematically as Gemini.

Consistency of AI Agents

Researchers are testing for consistency that how often AI agents, when given the same prompt multiple times, pick the same hotel or offer the same selection behaviour.

OpenAI GPT-4o: In terms of booking consistency, GPT-4o (with Browser Use) and Operator (OpenAI’s proprietary agent) consistently selected the same hotel when given the same prompt.

Anthropic’s Claude: Claude showed moderately high consistency in how often it selected the same hotel for the same prompt, though it chose from a slightly wider pool of hotels compared to GPT-4o or Operator.

Google Gemini: Gemini was the least consistent, producing a wider range of hotel choices and less predictable results across repeated queries.

Specificity of AI Agents

Specificity refers to how decisive the agent is in completing a booking task.

A higher specificity score more often the agent committed to a single choice. On the other hand, a lower score means it tended to return multiple options or respond less definitively.

• Gemini has the lowest specificity score at 60%, which means it frequently offers several hotels or vague selections rather than committing to one.
• GPT-4o has the highest specificity score at 95%, which means it almost always making a single, clear hotel recommendation.
• Claude scored 74%, meaning selection of a single hotel with more variation than GPT-4o.

Advertising strategies need to shift towards structured, keyword-rich formats aligning with how AI agents process and evaluate information, rather than relying on traditional visual design or emotional appeal.

Final Conclusion and Key Findings

• Investigated the AI agents for three language models (GPT-4o, Claude Sonnet 3.7, and Gemini 2.0 Flash) interacting with online advertisements during web-based hotel booking tasks, and each model received the same prompts and completed the same types of booking tasks.
• Banner ads received more clicks than sponsored or native ad formats. The most important factor in ad effectiveness was whether the ad contained relevant keywords in visible text.
• In terms of decision-making,
  • GPT-4o was the most decisive, usually selecting a single hotel and completing the booking
  • Claude was generally clear in its selections, but sometimes presented multiple options
  • Gemini tended to frequently offer several hotel options and completed fewer bookings overall
• Use of the booking site’s interactive filters
  • Gemini applied filters heavily
  • GPT-4o used filters occasionally
  • Claude’s behavior was between the two, using filters more than GPT-4o but not as consistently as Gemini.
• Consistency- how often the same hotel was selected when the same prompt was repeated
  • GPT-4o and Operator showed the most stable behaviour
  • Claude showed moderate consistency, drawing from a slightly broader pool of hotels
  • Gemini produced the most varied results

• Specificity- how often agents made a single, clear hotel recommendation
  • GPT-4o was the most specific, with a 95% rate of choosing one option
  • Claude scored 74%
  • Gemini was again the least decisive, with a specificity score of 60%

These findings suggested that digital advertising is anticipated to adapt AI agents, which means that keyword-rich formats are more effective than visual or emotional appeals, particularly as machines are the ones that interact with ad content.