What Sellers Actually Need from eBay.ai Descriptions: Challenging Trends and Assumptions
Client
eBay
Timeline
Q1 2025
Summary
This case study examines how sellers interact with AI-generated item descriptions on eBay. Through an in-depth user study, we uncovered what sellers actually need, highlighted gaps in the listing workflow, and explored ways to give sellers control while still leveraging AI. The findings reveal opportunities to make AI-powered description tools more intuitive, practical, and aligned with how sellers work in real life.
Contribution
I led design for this work, driving key decisions and shaping the experience and strategy in close partnership with Reece Dano, our lead researcher.
AI-generated descriptions raised concerns around tone, effort, and buyer relevance.
Following the launch of eBay.ai item description, the feature introduced several challenges for sellers and buyers. Sellers often described the tone as overly “salesy,” generic, or too long, leading to extra editing and a lack of confidence in using the feature. Many sellers also believed that buyers tend to rely more on photos and ratings than on descriptions, which lowered their trust in the feature’s value. While other issues came up, such as missing category-specific details and inconsistent formatting, the team prioritized tone first since that was the most common and immediate feedback.
“It sounds so very phony....just like a manufactured ad. People want to know what they are buying or bidding on, not some baloney about creating precious memories. I rate it 100% nauseating.”
— eBay seller
GenAI Description MVP launch in 2023
The team ran A/B tests comparing the current eBay.ai item description with a version that used a more neutral, straightforward tone. As adoption grew, seller feedback shifted toward wanting more depth and accuracy. Complaints about the tone being too over the top dropped from 22% with the current version to 6% with the alternative.
We tested generative AI trends to see what actually supports eBay sellers.
I explored several concepts inspired by current industry trends on generative AI with Reece, our lead researcher to test with sellers.
We conducted 90-minute moderated interviews with 12 C2C sellers across various categories and experience levels. We observed how they write descriptions, interact with eBay.ai, and adapt content to fit their workflows.
Participants were asked to define what makes an ideal item description and evaluated a series of early concepts using the RITE (Rapid Iterative Testing and Evaluation) method. We focused on understanding perceived effort, alignment with their listing habits, and how well the designs addressed both seller needs and buyer expectations.
Sellers unanimously stated a preference for factual, bolded and bulleted highlights when asked to provide first impressions of item description concepts.
We suggested for the team to invest in prioritizing auto-generating descriptions with bolded headers and bullet points. Descriptions with this formatting would align closely with what our sellers want in their item description.
“I like that it’s not all just words, how it is bulleted, and there’s the bold... it makes it really easy to read.”
— eBay seller
Sellers expressed interest in lightweight, optional prompts that help them remember key buyer-relevant details, especially for used or high-value items.
Sellers often add details like the item’s condition, a personal story or insight that shows expertise, or a testimonial based on their own usage. Many also include a consistent sign-off across their listings, such as packaging methods, disclaimers, or notes like “from a smoke-free home,” especially when trying to build trust or reassure buyers.
As an improvement to eBay.ai, we suggested the team to build a text area that would allow sellers to add more details to the description in a seamless way using AI.
“It's hard to put into words or know what other people are looking for. So I love a good template that gives you an idea of the information you need to put out there.”
— eBay seller
Sellers saw tone toggling as a nice to have, but nonessential and potentially time-wasting feature.
Sellers preferred descriptions without persuasive adjectives, aligning with our findings prior to this study. Other found value in experimenting with tone if they are willing to invest more time into curating their listing, usually for their high-valued items.
This finding was a pivotal moment for our stakeholders, as many of our earlier conversations centered around introducing different tones into the experience, which is a common feature often paired with AI content generators in the market.
For MVP, we proposed the team to avoid building a feature that enables different tones for now. We may want to investigate further in a follow up study.
"I don't want it to sounds like a catalog. I want it to sounds like me. But I'm not going to try six different versions. That's kind of a trap, like, I might keep changing it forever. "
— eBay seller
Sellers appreciated the idea of adding photos inline with the description to highlight flaws, verify product dimensions, or illustrate key points.
While this concept resonated with sellers, I understood that this was not fit for an MVP proposal as this would be quite expensive and require massive framework changes.
During the share out with our partners, this concept was premised as a future state to consider and to revisit when the team is ready to investigate further.
"This would be good if I wanted to add a photo of the damage in the description. So then it's not 'See pictures at the top'. Buyers tend to miss the photos.
— eBay seller
Listing on desktop (Seller's view)
Viewing the listing on desktop (Buyer's view)
Lastly, not everyone trusts AI, so it’s important that ebay.ai descriptions remain optional.
Not all sellers are enthusiastic about AI, and many prefer to rely on their own expertise when creating listings. To address this, I worked with the team to explore designs that keep AI features optional while remaining easy to discover, ensuring sellers can engage with AI at their own pace without disrupting their workflow.
We encountered several challenges along the way.
Getting buy-in for in-depth user interviews was the first hurdle.
Early in the project, stakeholders shared strong ideas, but I encouraged the team to pause before jumping into solutions. We looked beyond the numbers to understand the why behind user behavior, saving time and effort that could have gone into building the wrong thing.
Partnering with Reece and our product manager, we outlined all our open questions around AI and item descriptions. These sessions confirmed the need for an in-depth user study and helped the team align on the value of qualitative insights for the next phase of eBay.ai descriptions.
Designing for AI came with ambiguity and it raised tough conversations on designing for the right interactions.
Our audit showed that few standards for generative AI content existed, which wasn’t surprising given how new consumer-facing AI still is in 2025. Internally, design teams had conflicting opinions on which patterns felt most intuitive. As the designer leading this work, I navigated these conversations while proposing solutions that balanced near-term feasibility with long-term scalability alongside our stakeholders.
Scoping for the buyer-side experience would have been beneficial.
Our user interviews provided valuable insights, but research was limited to the seller experience. While sellers sometimes shared buyer perspectives, including buyer insights would have given a more complete view. With teams focused on different priorities, I created mockups of hypothetical buyer-side updates to connect our ideas end-to-end and illustrate the full experience.
Snapshot of a brainstorming session with my product manager and research partner.
What's next?
With our engineering partners, we’re taking the next step to refine eBay.ai’s generated content and text, grounded in the insights from our study. These learnings will shape the next phases of the AI-powered eBay listing experience.
Final thoughts
I believe eBay.ai can better support sellers by creating an experience where they can add and save details in a continuous flow. Research highlighted the importance of minimizing disruptions and allowing sellers to stay in control of their content, so they can work confidently without feeling pushed by AI.
Building on these insights, eBay.ai could offer focused, buyer-driven prompts that guide sellers through short, targeted tasks without interrupting their workflow. I explored this through a concept showing how AI prompts could support intuitive editing while keeping sellers in control.
Thanks for all your concept explorations for our Gen AI Item Description study. You pushed past the status quo of design ideation, unlocking valuable concepts that helped us gather deeper insights. Your collaboration, engagement, and leadership in the RITE process generated understanding and direction that will guide this space in the near term and for years to come.
— Reece Dano (Lead UX Researcher)