AI Discover: Custom Topics
Enterprise
Role
Type
Background
Peek Discover tracks how often properties appear in AI search across all major AI models.
Problem
The search topics that determined a property’s entire visibility score were AI-generated and unchangeable. Wrong topics meant wrong scores. Users had no way to fix it.
The Goal
Give Premium users control over their search topics without compromising data integrity, backend costs, or time-series continuity.
The Gaps
AI-generated topics could be a product of hallucinations, it can't be trusted with 100% accuracy just yet.
Property-specific differentiators were invisible. Unique amenities, commute destinations, and brand language had no path into the report.
Users couldn't connect their marketing strategy to their visibility data.
Execution
Timeline
The Messy Middle
Decisions & Pivots Along The Way
Every topic has a price
Each topic runs queries across 4 AI models every report cycle, so giving users unlimited additions was never just a UX call. It was a cost problem.
We landed on a shared pool of 35 topics. Users get a focused, expert-curated set and the cost constraint stays invisible.
We’re the expert. But they want control.
The topics users most wanted to add were the same ones hurting their visibility. Giving full control would have undermined the thing that makes Peek valuable.
Users can add up to 3 custom topics and overwrite defaults, but never remove them. They get enough agency to feel ownership without compromising expert authority.
What happens to old data?
Engineering flagged late that reports are immutable once generated. Any topic change means regenerating the entire report from scratch.
We turned immutability into the value prop. Your report always reflects exactly what you were tracking. Changing topics starts a new report and preserves your history.
Execution
The Flow
Topic changes branch on one question: has the first assessment run? If not, changes apply freely. If yes, users update the current report or start a new Report B with a fresh timeseries. Report A stays intact.
Final Designs
The Feature
Assured Flow
Change Means Commitment
Status: In Development
What Are The Next Steps?
Measure topic impact
Track whether custom topics actually improve visibility scores compared to defaults. This data closes the loop on whether user control adds real value or just perceived value.
Validate the cap
Monitor how often users hit the 3-addition limit post-launch. If most users only change 1-2 topics, the constraint holds. If they consistently push the cap, revisit.
Tie into credit pricing
Topic additions have a direct cost implication. As the credit-based pricing model takes shape, custom topics could become a natural unit of value. The groundwork is there, the model is still early.