01

CarGurus AI – Car Compare

CarGurus AI – Car Compare

UX Design · AI Integration · Strategy & Research

UX Design · AI Integration · Strategy & Research
UX Design · AI Integration · Strategy & Research

Before this project, shoppers on CarGurus often found it easy to search for cars , but hard to actually compare them. Once people narrowed down to a few good options, they’d open multiple tabs, take screenshots, or even jot down notes just to see which one was the better deal.

We heard the same frustration over and over: “I just want to know which car is better for me.”

This wasn’t just a user issue, it was a business opportunity. When buyers lose confidence or momentum at this stage, they drop off the funnel. Helping them compare options clearly could mean higher engagement, stronger trust, and better conversions for both consumers and dealers.

Project image
Project image
Project image

02

Process

My Role and Objective

Role: Lead UX Designer & Researcher
Scope: End-to-end experience design — from research and discovery to prototypes and testing

Success looked like:

  • Helping users easily compare vehicles side by side

  • Reducing bounce/drop-off at the comparison stage

  • Increasing engagement with dealer listings

Constraints:

  • Compressed 8-week timeline

  • Limited content team

  • Needed to work within existing CarGurus design system

Understanding the problem

We began by looking closely at how users were making decisions on the site. Our research included:

  • 10 remote user interviewss with active car shoppers

  • Behavioral analytics across search and listing pages

  • Competitive audit of other comparison tools

  • Search intent analysis

Key insights:

  • User reviews where the number 1 trusted source. They wanted a clear signal that they could rely on, something objective.

  • Lifestyle fit mattered more than specs. People weren’t comparing torque or fuel economy charts, they were asking, “Will this car actually work for my daily life?” Things like trunk space for kids’ gear or confidence in winter driving mattered most.

  • Comparison felt like work. Shoppers said it took too long to piece together everything they cared about. They wanted a faster, more guided way to decide, without opening five tabs or keeping notes in another app.

“I don’t care about horsepower, I just want to know if it’s a good fit for my family.”

These insights reframed the problem: users didn’t need more data. They needed clarity, confidence, and an easier way to see which car was right for them.

Project image
Project image

Defining the Opportunity

We synthesized everything into one guiding question:

How might we make car comparisons simpler, more trustworthy, and more personal, even with a small team and limited content support?

Through workshops with Product, AI, and Content, we aligned on three design principles:

  1. Build confidence, not complexity. Show users what’s different and why it matters.

  2. Focus on lifestyle fit. Translate technical specs into human language (“more cargo room for road trips”).

  3. Scale content with AI. Use natural language generation (NLG) to automate descriptive summaries that explain what makes each car unique.

We also mapped user and business needs:


User Needs

Business Needs

Clear, trustworthy comparisons

Higher engagement and dealer leads

Understand which car fits their lifestyle

Retain users through decision stage

Less manual effort to compare options

Scalable solution without new headcount

Project image
Project image

Ideation and Design Exploration

We explored a range of directions, from visual comparison tables to narrative-style summaries powered by AI.

Early sketches looked at:

  • A side-by-side comparison view anchored to the Search Results Page (SRP)

  • A dedicated “Compare” page that used AI to generate lifestyle-oriented summaries

  • A trust layer with user reviews and verified expert reviews

Our biggest challenge wasn’t design, it was content scalability. With only one content strategist and no writers available for daily updates, we turned to AI.

We built a lightweight pipeline using OpenAI’s text generation to automatically produce short, human-style summaries from structured data. Each summary pulled details like price, trim, mileage, and deal ratin, then translated them into plain, confidence-building language.

This became the foundation for our MVP: a clean, comparison-first experience enriched by AI-generated editorial insights.

Project image
Project image

Testing and Itiratiaon

We ran two rounds of user testing , one moderated, one unmoderated (via Maze).

Round 1 — Concept Validation:

  • Users instantly understood the side-by-side comparison model

  • The AI-generated summaries felt “personal” and “helpful,” though some wanted to know the source of the data

  • Trust increased when we displayed user reviews and aggrigate rating.

Round 2 — Refinement & Flow Testing:

  • Too many specs cluttered the screen; simplifying to key lifestyle attributes improved comprehension

  • Tests showed users spent longer engaging with AI-enriched comparison views

We iterated quickly, simplifying the layout, clarifying attribution, and ensuring users could both trust and verify what they saw.

Final Designs

The final experience was a side-by-side comparison tool that blended human-centered design with AI efficiency.

Core features:

  • Clean, minimalist side-by-side comparison view

  • AI-generated insights focused on lifestyle fit and confidence

  • AI-generated CarGurus recommendations

  • Transparent trust cues: verified data sources and clear explanations of deal ratings

Users could instantly see how each car stacked up, not just on specs, but on how it would fit their life.

Project image

Defining the Opportunity

We synthesized everything into one guiding question:

How might we make car comparisons simpler, more trustworthy, and more personal, even with a small team and limited content support?

Through workshops with Product, AI, and Content, we aligned on three design principles:

  1. Build confidence, not complexity. Show users what’s different and why it matters.

  2. Focus on lifestyle fit. Translate technical specs into human language (“more cargo room for road trips”).

  3. Scale content with AI. Use natural language generation (NLG) to automate descriptive summaries that explain what makes each car unique.

We also mapped user and business needs:


User Needs

Business Needs

Clear, trustworthy comparisons

Higher engagement and dealer leads

Understand which car fits their lifestyle

Retain users through decision stage

Less manual effort to compare options

Scalable solution without new headcount

Project image

Ideation and Design Exploration

We explored a range of directions, from visual comparison tables to narrative-style summaries powered by AI.

Early sketches looked at:

  • A side-by-side comparison view anchored to the Search Results Page (SRP)

  • A dedicated “Compare” page that used AI to generate lifestyle-oriented summaries

  • A trust layer with user reviews and verified expert reviews

Our biggest challenge wasn’t design, it was content scalability. With only one content strategist and no writers available for daily updates, we turned to AI.

We built a lightweight pipeline using OpenAI’s text generation to automatically produce short, human-style summaries from structured data. Each summary pulled details like price, trim, mileage, and deal ratin, then translated them into plain, confidence-building language.

This became the foundation for our MVP: a clean, comparison-first experience enriched by AI-generated editorial insights.

Project image

Testing and Itiratiaon

We ran two rounds of user testing , one moderated, one unmoderated (via Maze).

Round 1 — Concept Validation:

  • Users instantly understood the side-by-side comparison model

  • The AI-generated summaries felt “personal” and “helpful,” though some wanted to know the source of the data

  • Trust increased when we displayed user reviews and aggrigate rating.

Round 2 — Refinement & Flow Testing:

  • Too many specs cluttered the screen; simplifying to key lifestyle attributes improved comprehension

  • Tests showed users spent longer engaging with AI-enriched comparison views

We iterated quickly, simplifying the layout, clarifying attribution, and ensuring users could both trust and verify what they saw.

Final Designs

The final experience was a side-by-side comparison tool that blended human-centered design with AI efficiency.

Core features:

  • Clean, minimalist side-by-side comparison view

  • AI-generated insights focused on lifestyle fit and confidence

  • AI-generated CarGurus recommendations

  • Transparent trust cues: verified data sources and clear explanations of deal ratings

Users could instantly see how each car stacked up, not just on specs, but on how it would fit their life.

03

Outcome

Quantitative results (first 90 days):

  • +22% increase in engagement

  • +15% lift in click-throughs to dealer SRP

  • +9% increase in average session duration

Qualitative impact:

  • Users described the feature as “clear,” “reliable,” and “way less work.”

  • Dealers saw an improvement in lead quality, buyers came more informed and confident.

What worked:

  • Translating specs into lifestyle language users could relate to

  • Clear trust signals throughout the experience

  • Leveraging AI to scale content creation without compromising tone or accuracy

What I’d improve next time:

  • Personalization: allow users to select what matters most (commute, family, performance)

  • Deeper transparency: show how AI-generated insights are validated

Reflection:
This project was a turning point in how we approached AI at CarGurus. Instead of replacing human writers, AI became a creative partner, helping us deliver clarity and confidence at scale. It reinforced that great UX isn’t just about what users see, but how sure they feel about the decisions they make.