Can DeepSeek AI Turn User Complaints into Actionable Design Solutions?

User negative feedback often sparks experience optimization; the author wonders whether DeepSeek can extract actionable suggestions from complaints and whether those suggestions are implementable.

The case study focuses on the posting page list used by B‑side merchants on 58.com, a core component for publishing, editing, and deleting posts, which has undergone several redesigns with poor results and many merchant complaints about difficulty and discoverability.

Before feeding the problem to DeepSeek, a universal prompting formula is used: background + constraints + goal, optionally adding a role (e.g., "as a designer") and a solution framework (e.g., "using the STAR method").

The designer’s analysis addressed core functionality, basic architecture, and user fatigue. DeepSeek’s analysis aligned closely with the designer’s, covering similar details and even some additional points.

Despite not having full page knowledge, DeepSeek identified user pain points and proposed solutions that largely matched the designer’s revision strategies, effectively encompassing all the designer’s recommendations.

DeepSeek’s suggestions were surprisingly consistent with the designer’s final, data‑validated redesign, demonstrating that AI can provide constructive UX advice; however, occasional absurd or seemingly reasonable yet flawed proposals were also generated.

Some AI‑suggested interactions, such as long‑press actions, are unsuitable for B‑side efficiency‑focused scenarios with older users unfamiliar with swipe‑to‑close gestures.

Therefore, while DeepSeek is powerful, all AI‑generated recommendations still require designer review and validation before implementation.