policypatterns is an experimental tool to help researchers, analysts, and advocates collect public policy data, one question, one jurisdiction at a time.
Across the internet, public policy is published in thousands of formats, websites, PDFs, and local statutes. We believe access to this information shouldn’t be a barrier to research, compliance, or public understanding.
With policypatterns, you can:
- Ask a structured question (e.g. "Is there a plastic bag ban in each U.S. state?")
- Select an answer type (measure, category, list)
- Automatically generate a structured dataset from web-based sources using large language models (LLMs)
- Explore previous outputs in a searchable catalog
💡 Why We Built This
We’ve worked in government, policy, and research and kept running into the same problem:
Public policy is a patchwork, but nobody wants to stitch it together.
So we started with a simple goal: make it faster and easier to ask policy questions across jurisdictions and return structured, inspectable answers.
🔍 What This Is (and Isn’t)
This is a prototype. We’re testing whether AI-assisted search can produce trustworthy, useful policy data and how humans might review or build on it.
This is not a replacement for expert legal research yet. But it could become a valuable tool for those seeking to understand policy landscapes, track changes, or analyze trends.
🤝 How We Built This
This MVP is powered by:
- LLM-based search (via Perplexity API's structured outputs feature)
- Lightweight frontend tools (Made with Streamlit)
- Feedback from a small group of domain experts and researchers
We’re experimenting with open data and responsible AI — and your feedback shapes what comes next.
🧪 Built by Humans
This tool is built by humans, for humans. We’re not perfect, and neither is this tool. Want to collaborate, share feedback, or request a dataset not achievable via the UI?
📬 [Submit a note to our request form](https://docs.google.com/forms/d/e/1FAIpQLSfViFhKM6OmmuWfHmU6yfd4_F4uNC-Ni1WUPoyDuqOxe1bFAg/viewform?usp=header .
Built With
- perplexity
- python
- streamlit

Log in or sign up for Devpost to join the conversation.