Research

The knowledge infrastructure and evaluation work measuring how well AI systems serve and represent rural communities.

Why this research matters

AI models see, experience and learn about rural America like tourists, not locals.

Artificial intelligence systems learn from what has been written down, digitized, and published — and rural America is underrepresented at every stage of that record. Rural communities are less completely documented online, local news coverage is thinning, and much of what rural people know still lives in archives, organizations, and lived experience rather than in formats AI training pipelines can absorb. The result is measurable: the first systematic audit of place-based knowledge in large language models found that models answer questions about urban counties with significantly more accuracy and depth than rural ones, with the gap concentrated in exactly the knowledge that defines a community — its events, institutions, and social fabric. As AI moves into health care, public benefits, lending, and the software that runs everyday business, systems built on this incomplete record risk missing rural realities in decisions that carry real economic weight. Our research exists to make that gap visible, measurable, and fixable before rural knowledge becomes an afterthought in the systems shaping the next economy.

BRAIN

Internal Use · Aug 2026

Bedrock for Rural AI kNowledge

BRAIN is a curated, cited library of rural-AI intelligence — a centralized knowledge base built for our team and our AI tools to use. It plugs in as an MCP connector, so any tool we run can draw on the same shared, vetted, purpose-built memory rather than searching scattered files one query at a time.

It replaces scattered knowledge, repeated data-gathering, and dependence on any one person's storage with a single searchable library where every answer carries a citation. Centralizing that knowledge makes it far more efficient — and lets it work at a larger scale, and more accurately, than a language model reaching into a document folder.

It allows any entity to scale a database of authoritative, cited, referenced knowledge across their organization and then make it usable for others: a municipality, a tribal nation, and a trade association.

Built — and our team uses it every day

102
distinct sources
100%
of answers carry citations
24/7
automated ingestion

What Lives in BRAIN

ArticlesGrants & grant informationLegislationStoriesVideosPodcastsWhite papersResearch papersPress releases

What's Live & Working

  • AI tools query BRAIN directly
  • Auto-ingested data, approved by CRAI
  • Eval-tested and ready for handover

What's Next: From Internal Tool to Shared Rural Resource

1
More data, more partners

New datasets and external partnerships widen what BRAIN knows.

2
Smarter under the hood

Semantic search and wider ingestion sharpen how it answers.

3
Open it up

Other rural organizations plug in to the same shared, vetted memory.

The takeaway: a small nonprofit can build the curated, AI-ready knowledge layer rural communities need.

Benchmark

In Development

How well do today's AI models understand and represent rural information?

Benchmark tests how well today's leading AI models—Claude, Gemini, GPT, and others—actually perform for rural users and rural contexts. Currently, these models tend to default to urban, data-rich environments. This project turns that gap into a public, comparable score.

It won't fix the problem on its own but by making the metro–rural tech-equity gap visible, it starts the conversation and spreads awareness, rather than asking people to just take our word for it. The project is in active development.

How It Will Work

01
Ask

Pose a set of 50+ rural-specific questions to each model and its sub-models.

02
Score

Grade answers against a rubric: accurate terminology, avoids urban defaults, reflects rural reality.

03
Publish

Post the results publicly so anyone can see where the gaps are.

04
Repeat

Re-run on a regular cadence to track how models change over time.

Sub-project of Benchmark

BRAIN Bench

Pilot Evaluation

The first step in closing the Rural AI Gap.

CRAI invested in BRAIN on a strong hypothesis, but without proof. There was no quantitative measure of whether BRAIN's answers were more accurate than an AI model working on its own, and no reliable way to pin down where its knowledge gaps were or how to close them. BRAIN Bench is the evaluation that turns that hypothesis into evidence.

100
real prompts tested
9
industry categories
6
models compared
4
rubric dimensions scored
36%
of prompts from CRAI's own first users

It grades real prompts — many drawn straight from CRAI's own first users — across nine industry categories and six leading models, scoring each answer against a four-dimension rubric. The result moves BRAIN from guesswork to proof: something CRAI can measure, repeat, and tweak as the tool grows toward a public, rural-facing resource.

First-run results: blended score by model

CRAI's BRAIN configurationbaseline models, tested as-is

Claude + BRAIN MCP
85%
Sonnet 5
84%
Claude Haiku 4.5
79%
GPT-5.6 Sol
76%
GPT-5.6 Terra
71%
Claude Opus 5
63%
Scores by rubric dimension, with answer length and response time.
ModelAccuracyCompletenessRural nuanceHonestyAvg tokensAvg time
Claude + BRAIN MCP100%58%81%100%4008.0s
Sonnet 590%72%82%91%2,05627.7s
Claude Haiku 4.584%75%69%88%6096.4s
GPT-5.6 Sol82%65%71%84%1,65335.9s
GPT-5.6 Terra78%60%66%79%1,91423.8s
Claude Opus 568%55%62%66%5,12078.1s

Read this as a first measurement, not a victory lap. BRAIN leads on accuracy and honesty — it answers from cited sources or says it does not know — but its one-point margin over the strongest baseline is well inside the noise of a 100-prompt run, and its lower completeness score reflects notably shorter answers. Pinning down that tradeoff is exactly what the benchmark was built to do.

BRAIN Bench is already shaping CRAI's next initiative — a Rural Benchmark at a much bigger scale — applying the same method to the wider question of how well every major AI model represents rural America. Built by Emaliah Sawyer.

Grounded in Real Rural Expertise

Subject-matter experts help build the initial questions and grading rubric and validate the answers, so scoring reflects real rural experience. From there, we'll build an agentic grading system so the benchmark can run without relying on SMEs every cycle. Any rural or tribal dataset the Benchmark draws on or publishes routes through our Tribal & Community Data Sovereignty Review Process.

Why It Matters

When foundation models underrepresent rural contexts, they inherit blind spots that affect millions of Americans. Public, repeatable results give rural institutions evidence for which tools to trust and give model developers a concrete target for doing better by rural communities.