Inspiration

A teammate runs Reliable Roofing Solutions in Auckland. Homeowners now type roofing questions directly into ChatGPT — “How long will my roof last?” “Who can give me a quote?” — and the model recommends competitors, never his site. We realized thousands of local businesses are becoming invisible inside AI answers, so we created Prominence to help raise that visibility.

What it does

Prominence discovers every public asset tied to a brand, crawls it, and assigns a single AI-Engine Optimization (AEO) Score.

  • Breaks the score into structured-data quality, snippet clarity, and crawlability.
  • Generates an effort-versus-impact chart of optimization tasks, filterable by priority.
  • Benchmarks competitors, highlights content gaps, and suggests actionable insights.

How we built it

  • Frontend – React, TypeScript, Tailwind; animated charts via Recharts.
  • Backend – Python + FastAPI.
  • Web crawler – Node.js and Puppeteer workers, orchestrated through a poor man’s Kafka: a Postgres table acting as a queue.
  • AI layer – Gemini API calls for real-time ranking checks, LangChain for prompt templates.
  • Data store – Results in Supabase; assets in Bolt Blob Storage.

Challenges we ran into

  • Rate limits and CAPTCHAs – Solved with rotating proxies and exponential back-off.
  • URL deduplication at speed – Implemented a Bloom filter to keep the URL frontier lean.
  • LLM answer variability – Sampled three prompt variants and averaged rankings for stability.
  • Time – Shipping a SaaS in one month meant hard cuts; Jira export is still on the roadmap.
  • Crisis – Our video editor withdrew at the last minute, so we produced the product-tour and spotlight videos ourselves using a mix of tools.

Accomplishments that we’re proud of

  • Shipped a fully working SaaS in a single sprint — three developers and one PM.
  • Revamped the UI three times within the same period, all on Bolt.
  • Created an industry-scale web crawler.

What we learned

  • Traditional SEO metrics overlook how LLMs value contextual completeness and schema richness.
  • A single narrative score drives action better than a list of warnings.
  • Bolt is excellent for quick MVPs, but the web-container environment struggles as the codebase grows.

What’s next for Prominence

  1. One-click export of tasks into Jira, Asana, and Notion.
  2. Always-on monitoring that triggers Slack alerts when the AEO score drops.
  3. Open-sourcing our scoring schema and crawler framework for the community.
  4. A marketplace of “playbooks” so agencies can package their own optimization recipes.

Built with Bolt for the Bolt Hackathon, June 2025.

Built With

Share this project:

Updates