Search Audit and AI Experience Improvements
Opportunity
When I joined The Washington Post's Central Design team in May 2025 following a team restructure, I inherited a search experience that had been neglected for years. Our users were frustrated—search results were often outdated and irrelevant, with no filtering or sorting options available.
The challenge was compounded by the 2024 launch of "Ask the Post," our AI-powered question-answering feature. While innovative, it had been placed directly on top of our search page without strategic integration. User surveys revealed a harsh reality: not only were our standard search results unhelpful, but the majority of users found our AI experience equally frustrating and intrusive.
As someone new to this team and coming from email tool design, I needed to quickly understand both the technical constraints and user needs while navigating stakeholder concerns about project prioritization.
Approach
I needed to audit and redesign our search experience to better integrate AI functionality while giving users more control and choice. When stakeholders initially considered deprioritizing the project, my AI enhancement proposals kept it moving forward.
Research and Discovery Coming from email tool design, I made building relationships with Design and Engineering leads my first priority, gathering existing data and user feedback. I then conducted a comprehensive audit by personally testing our search functionality across all environments, documenting when AI responses appeared, their helpfulness, and missed opportunities for improvement.
Key Problems Identified My audit revealed several critical friction points:
Duplicate content confusion: Articles used for AI responses also appeared in search results with completely different styling and no indication of how each was sourced
Poor space utilization: Ask the Post dominated page real estate, forcing users who just wanted quick search results to scroll past lengthy AI responses
Limited AI activation: Short queries or incomplete phrases didn't trigger helpful AI assistance, missing opportunities to guide users
Poor result organization: Outdated articles ranked highly with no user control over sorting or filtering
Confusing feedback system: Too many emoji options for sentiment feedback, particularly inappropriate when users were researching serious news topics
Design Solutions Working closely with other designers to ensure alignment across our AI product suite, I developed four key solutions:
Collapsible AI Responses: Shortened AI responses initially to give users immediate access to search results
Smart Question Suggestions: Surface relevant follow-up questions for incomplete queries to help users refine searches
Transparent Sourcing: Added inline citations linking to specific source passages to build trust and encourage exploration
Author-Focused Experiences: Created specialized author pages with bios and follow buttons instead of AI responses
New Features
Sourcing
One of the most jarring aspects of our original Ask the Post experience was the disconnect between AI responses and source articles. Users would see an AI answer followed by search results that included the same articles, but with different visual treatment and no indication of the connection between them.
Working with designers across product teams, I developed a sourcing system with inline citations that link directly to specific passages in source articles. This builds trust through transparency while creating natural pathways for deeper exploration, positioning AI responses as sophisticated starting points that showcase the quality of our journalism rather than replacements for it.
Follow up Questions
One of the most significant opportunities I identified was helping users who entered incomplete searches. When someone searched for just "climate" or "tariffs" our AI system wouldn't activate, leaving them with potentially overwhelming or irrelevant results and hundreds of thousands if not over a million results.
I designed an intelligent system that detects when users enter short or incomplete queries and responds by surfacing strategically curated follow-up questions. These aren't random suggestions—they're based on our most frequently asked questions related to that topic and current trending searches. This approach helps users narrow down their search from potentially thousands of results to hundreds of more relevant articles, while giving them agency to choose whether they want guided assistance with their search.
My research revealed that many users search for specific authors, but our AI system generated bloated, unhelpful responses while burying the actual author bio page in search results. This created a frustrating experience for users who simply wanted to learn about or follow a particular journalist.
I redesigned this by recognizing that author searches have different intent than topic searches. Instead of forcing AI responses, I created a specialized experience that prominently features the journalist's bio, recent articles, and a follow button for future content. This transforms a poor search experience into an engagement opportunity that supports our subscription and reader loyalty goals.
Result
Project Status and Goals This project is currently in development, with launch and user feedback collection planned for the near future.
Expected Outcomes:
A more streamlined AI experience that enhances rather than hinders the search process
Increased user agency and choice in how they interact with AI-generated content
Better content discovery through transparent sourcing and strategic question suggestions
Improved author engagement through dedicated author search experiences
A cohesive AI strategy that aligns with other product initiatives across The Washington Post
Personal Impact: This project demonstrated my ability to quickly adapt to a new domain, conduct thorough user experience audits, and design solutions that balance user needs with business objectives. My AI improvement proposals were compelling enough to keep a potentially canceled project moving forward, showing the value of thoughtful, user-centered design approaches.
Looking Forward: I'm excited to collect user feedback post-launch and iterate on these improvements. My goal is to create an AI experience that users genuinely find helpful—one that empowers them to explore our journalism more deeply rather than creating barriers to the content they're seeking.
