AI REVIEW INSIGHTS

TRIPADVISOR

Overview

I led the design of an AI-powered feature that synthesizes large volumes of user reviews into clear, categorized insights to improve decision-making for travelers booking experiences on TripAdvisor.com.

Role: Lead Product Designer
Company: TripAdvisor
Timeline: 5 Months

Impact

  • Enabled faster, more confident booking decisions and reduced review fatigue

  • Increased tour bookings by 5.82%

  • Established a scalable foundation for future AI-driven review experiences across TripAdvisor for Restaurants & Hotels PDP

Problem

Internal research and user feedback showed users were overwhelmed by large volumes of reviews and struggled to quickly extract meaningful insights, slowing their decision making on tour PDPs.

Key issues:

  • Reviews were at the bottom of the page, reducing visibility

  • Too much scrolling

  • Repetitive content with no synthesis

  • Important themes buried in long-form text

INSIGHTS

Users weren’t looking for more reviews—they were looking for patterns they could trust and act on quickly.

  • Consistency across reviews is valuable, but only when surfaced clearly

  • Users scan for themes, not individual opinions

  • Trust depends on being able to verify insights against real reviews

Design Decisions

I evaluated tradeoffs and defined clear rules to translate raw AI outputs into insights that are clear, trustworthy, and valuable to users.

How I Structured the Output

  • Defined how machine-learning outputs are structured and surfaced as consistent, scannable insight statements

  • Applied character limits to maintain readability and ensure layout consistency

Defined What Qualifies as an Insight

  • Set minimum thresholds for frequency and sentiment confidence before surfacing a theme

  • Filtered out low-signal and one-off feedback to reduce noise

  • Prioritized statistically meaningful, experience-relevant insights (e.g., food quality on food tours over scenery)

Designed for Trust

  • Connected each insight to supporting reviews (“Mentioned in X reviews”)

  • Enabled validation through a modal with full review context

  • Included both positive and negative sentiment to ensure balanced, credible insights

Final EXPERIENCE

I designed a cross-platform insight layer that surfaces the most relevant traveler takeaways directly within the booking flow, enabling users to quickly evaluate experiences without relying on individual reviews.

Structured Insight Cards: Insights are organized into key themes and presented as concise, scannable statements, allowing users to quickly compare what stands out across experiences while reducing cognitive load.

Expandable Validation: Each insight connects back to supporting reviews, allowing users to explore context and validate the information.

Designed for Trust: The experience prioritizes transparency by clearly communicating how insights are derived and presents a balanced view of both positive and negative sentiment.

Presentation
View Slides
Back to portfolio