Exploring AI-Supported Messaging on eBay Mobile App

Picture from Unsplash.

Context 🛒

On platforms like eBay, it’s common for sellers to message buyers with questions about item listings. For sellers, receiving a timely response can strongly influence their decision to proceed with a purchase.

According to eBay’s internal metrics, faster buyer responses are associated with higher conversion rates and greater total sales volume.

With the growing potential of gen AI, there is a new opportunity to automate buyer responses at scale, which helps ensure timely communication and reducing buyer effort.

At the time of this research, our engineering team had also developed an early prototype of an AI-messaging feature, requiring foundational user research to inform its future direction.

Research Questions 🔍

  1. What benefits and concerns do eBay sellers perceive around using AI-generated responses when communicating with potential buyers?
  2. What design guidelines can be drawn to inform the development of future AI-supported messaging features for buyer–seller communication?
Picture from Pixabay.

Goal 🎯

This study served as a foundational research initiative to guide the design of AI agents for buyer–seller messaging in e-commerce.

The goal was to understand seller communication needs and evaluate how AI-generated responses could support smoother, more effective conversations on platforms like eBay.

Method 🧪

I conducted an online diary study to explore the longitudinal experiences of eBay sellers who were beta-testing our AI messaging feature.

Participants documented their interactions and reflections over time, providing real-world insights into how the feature influenced their communication with buyers.

Following the diary study, I conducted exit interviews to gather in-depth feedback on usability, perceived value, and areas for improvement.

Current project is an informative study using qualitative method.

Outcome and Impact 📝

  • Feature successfully launched to 60M+ mobile users.
  • Informed strategic planning for AI-powered features that aim to enhance buyer–seller communication and drive higher conversation and conversion rates on the platform.
  • Published a foundational internal report documenting key findings and design implications as one of the first research efforts focused on AI-supported messaging at eBay.
  • Shared actionable design considerations with over 40 cross-functional stakeholders across Product, Engineering, UX, and Design through dedicated insight-sharing sessions.
  • Presented the research at a company-wide showcase, increasing visibility and engagement around AI-messaging initiatives across teams.

Team 🤝

This research was conducted in close collaboration with multiple stakeholders across eBay:

  • Partnered with a Product Manager and the Seller Experience Engineering team to understand technical constraints and business priorities.
  • Collaborated with the Legal team to ensure the study met ethical standards and complied with company guidelines.
  • Coordinated with the UI Design team to align on design goals and co-develop study materials.
  • Supervised by a UX Research Manager, who provided strategic guidance throughout the project.

My Role 🧑‍💻

As the lead researcher on this project, I was responsible for driving the study from scoping to delivery:

  • Collaborated with PMs, engineers, and designers to contextualize the business problem and define the research scope.
  • Selected online diary studies and interviews as the primary methods to capture both real-time behaviors and in-depth perspectives.
  • Led all end-to-end research activities, including study design, recruitment, data collection, analysis, and stakeholder reporting.

Study Procedure 1

I recruited 18 eBay sellers enrolled in our beta testing program for a two-week online diary study, followed by an in-depth exit interview with each participant.

During the diary period, participants actively used the AI messaging feature and completed the following tasks:

  • Short survey questions assessing the quality and helpfulness of AI-generated messages
  • Reflection videos submitted twice a week, where participants shared their thoughts on the feature in the context of their real selling experiences

This mixed approach allowed us to capture both quantitative feedback and rich contextual insights, helping us understand not just how sellers used the feature, but why it worked, or didn’t, for different situations.

Key Findings and Implications2 🔍

The study uncovered several critical insights that shaped the direction of AI messaging on eBay. This feature is now successfully launched.

These included:

  • Sellers’ expectations for message tone and accuracy in automated responses
  • Tensions between efficiency and personalization in AI-generated communication
  • Opportunities to build trust and control into future AI messaging features

These insights directly informed design decisions and strategy for the next iteration of the feature.

  1. More details of study procedures are protected due to NDA. ↩︎
  2. Same as above. ↩︎
Return to the Portfolio Main Page