Smarter Software Search @ G2
Leveraging AI to deliver personalized search results
I joined G2's Buyer Experience Team as a Product Design Intern, where I owned the site search redesign, working with with engineers to ship the first MVP of G2’s AI-supported search experience for software buyers.
This work performed well in usability testing and with real users after the MVP launch, establishing contextual AI design patterns across the site and advancing G2’s shift to an AI-first company strategy.
Timeline:
Summer 2025
Tools & Skills:
Figma & Make, Usertesting.com
With:
Mentored by: Allison Horrell,
Aryn Silverberg
Site search metrics in June 2025
Why Search
When software buyers on G2.com struggle to find relevant products, they drop off without engaging.
This leads to lower engagement and lost buyer-intent data, directly impacting the value G2 provides to software sellers on the platform.
Discovery and Research Insights
An outdated search results page optimized for SEO,
not for real buyer needs or a good site experience
I ran a usability test with 6 buyer persona testers and tracked G2 site analytics (Usertesting & Amplitude)
In addition to secondary research, I wanted to hear directly from real users to get the software-buying process from a buyer's perspective. Site analytics helped me see the bigger picture and how G2's current users behave.
Software buyers start their search with clear criteria and constraints: budget, integrations, team size, use cases…
Software buyers start their search with clear criteria and constraints: budget, integrations, team size, use cases…
and they want to quickly narrow down search results to fit these needs,
Software buyers start their search with clear criteria and constraints: budget, integrations, team size, use cases…
and they want to narrow down product results by these needs,
but current G2 search filters can't keep up.
On the other hand,
Search pages have the highest % conversion on G2 AI Chat, signaling that people are willing to turn to AI for better answers.
Early Brainstorming & Design Explorations
Brainstorming with engineers surfaced new possibilities (and limitations) of integrating G2 AI for a smarter search.
I led a HMW workshop, presenting my research insights and having the team reframe the problem statement.
The team voted on 3 directions to explore further:
Tailored filtering
Personalizing results
Exploring interventions for different parts of the search experience & iterating with design critique
I created low fidelity explorations for filtering within the search bar, freeform/table views of the results page, and prompt-to-filter interactions.
Design
G2 has rich data on software features and buyer insights that traditional search and static filters fail to leverage, but open ended prompting could leverage.
Context-aware product card information
G2 has rich data on software features and buyer insights that traditional search and static filters fail to leverage.
Contextual AI Entry Points
Takes the current search context and applied filters allowing buyers to seamlessly transition their search into G2 AI.
Prototype & Testing
Robust prototype with Figma Make for user testing
To truly validate user behavior with this novel search & filter, I needed a prototype that closely simulated the prompt to filter functionality with real text input and dynamic results.
Testing with 5 recent software buyers revealed:
✅
Buyers easily understood the AI powered prompt to filter interaction.
"… this met my expectations and worked exactly the way I would expect it to.…"
✅
All testers loved the "Perfect Match" & "Partial Match" system.
"I love the perfect matches, right? This tells me everything that fits the attributes they're in…"
✅
Testers found the personalized product information very helpful.
"… this has the exact amount of information that makes [search results] extremely useful"
⚠️
2 testers thought the prompt box would lead to a conversational interaction. **
"… I kind of expected to have a back and forth, like a conversation…"
**product domain ownership structures restricted conversational interactions for my project, but ask me about this direction!
Design changes after testing
Shipping 3 MVP feature designs with engineers
and seeing real impact in the first 2-weeks of launch.
I created 3 MVP handoff files with specs for engineers on my team to implement during my internship (huge shoutout to my team for getting these out so fast), and conducted my first design QAs!
Final Design
Reflections & Thanks :)
Hearing from end users, engineers, and taking ownership at a tech unicorn
Value of User Insight:
Hearing from end users was essential, especially for a product I’m not the target audience for. Their input shaped my most relevant and impactful design decisions.
Keeping the team involved:
Involving engineers early in the process ensured that design ideas were technically feasible and gave me a clearer understanding of what was possible early on.
Learning in a High-Growth Environment:
At G2, I enjoyed the greater ownership, exposure to real customer data, and the energy of startup culture. I'm especially grateful to my team for trusting and engaging with my work as if I was a full time designer!
Thank you to the PXD team!
I’m so grateful to the G2 product designers for their mentorship and thoughtful feedback throughout my summer. They created a space where I not only grew as a designer and also as a person and a creative beyond PXD work.
Thank you to my fellow G2 interns!
I learned so much from the other G2 interns' drive, energy, and kindness, and I'm excited to see everything they accomplish in the future :)