EUR-LEX AI Search
European Union online law database
PROJECT
Case study
SERVICES
User Experience Design
The EUR-Lex search before the redesign
The EUR-Lex search after the redesign
ABOUT THE PROJECT
In this project, we aimed to leverage AI to create a highly effective and user-friendly search experience for EUR-LEX, the primary EU law repository. Our challenge was to go beyond a typical chatbot, which can be slow and ineffective, and instead integrate AI meaningfully to enhance user interaction and efficiency.
CONTEXT
Many internal users of the European Parliament’s websites are expert users. Unsurprisingly, an important part of their work is exploring European Union laws, regulations, or tracking the status of legislative proposals and elements. A good and successful example of that is Legislative Train Schedule, which is well-understood and heavily used. This project focuses on enhancing EUR-LEX, the main EU Law repository, which provides access to EU legal documents in all 24 official EU languages and is updated daily.
Design challenge
We needed to design a solution that would:
Provide precise, contextually relevant search results
Improve navigation through intuitive filters
Utilize AI to enhance the search process beyond basic interactions
Research and Ideation
We needed to design a solution that would:
Provide precise, contextually relevant search results
Improve navigation through intuitive filters
Utilize AI to enhance the search process beyond basic interactions
EUR-Lex search redesign
Key Features
Additional Features
Once the user proceeds to the search results, several additional features enhance the user experience:
In-Document Search: Allows users to search within the document for specific terms or sections, improving document navigation.
Document Accuracy Indicator: Shows the accuracy of the document based on AI analysis, helping users assess the reliability of the information.
Human Verification: Highlights documents that have been verified by human experts, adding an extra layer of credibility.
Summary Function: Provides AI-generated summaries of documents, enabling users to quickly understand the key points.
Text Highlighting and Query Refinement: Users can highlight text within documents and refine their queries based on the highlighted sections for more precise results.
Search results redesign
Research summary
User Interviews - Key Issues
Complex Navigation & Suboptimal Search
Broad/Narrow Search Results (Low Relevance)
Mismatch Between Site Content & User Needs
Lack of Context for Search Engine Users
Scattered, Duplicated, and Contradictory Information
No Ongoing User Research & Testing
Analytics Insights
Traffic Sources: Mostly from search engines (Mobile 52% | Desktop 48%)
User Intent: Seeking information (news, facts, reference material)
Device Usage: Mobile Safari (Europe) | Windows 10 (Desktop)
Most Popular Pages: Fact Sheets | Liaison Offices | Homepage | MEPs
User Challenges
Project Constrains (New Design Must…)
Improve Task Completion & Error Recovery
Support 24 Languages & Meet Accessibility (AA) Standards
Be GDPR-Compliant & Inclusive (Diverse, LGBTQI-friendly, Neurodivergent-friendly)
Minimise Environmental Impact & Boost Engagement
Reduce Service Desk Requests & Speed Up Publishing
Ensure All Current Content Remains Accessible
OKRs – Success Metrics
Objective: Enhance Navigation & Search for All Users
KR1: Reduce navigation failure rate by 50%
KR2: Reduce search failure rate by 50%
KR3: Decrease service desk inquiries by 30%
KR4: Increase engagement in campaigns
Takeaway
The learnings from this project emphasize the importance of continuous user engagement and feedback to drive ongoing improvements and innovations in digital experiences. This project highlighted the power of AI in transforming the search experience on EUR-LEX, making it more efficient, accurate, and user-friendly.
AI search faces several significant challenges, including accuracy and relevance of results, and transparency and explainability. Difficulties in understanding context and handling multilingual queries further complicate effectiveness. Technical limitations, such as high computational resource needs, and the necessity for ongoing maintenance and continuous learning, add to the complexity.
Additionally, user trust and adoption can be hindered by past performance issues, and ethical and legal challenges must be navigated to ensure compliance and accountability. Effective searching, filtering, scoping, and sorting are crucial in this context, as they directly impact users' ability to find relevant information quickly and efficiently, enhancing the overall utility and user experience of AI search systems. Addressing these challenges is essential for developing reliable and user-friendly AI search technologies.