Insights & Research

There's an ice cream shop near where I live in Durango that's been around quite a long time. On a summer Saturday, the line runs out the door and halfway down the block.

When you get to the front, you pay on a register that looks like it was salvaged from a 1940s soda fountain. Cash only. No receipt unless you ask. The owner has been running it this way for decades and sees no reason to change.

I spent most of my career in Silicon Valley. I helped create open source ecosystems, launched a couple of startups and watched the data economy get constructed from the inside. I understand the theory that AI systems reflect their training data. Moving to rural Colorado has really brought that concept home.

That ice cream transaction doesn't exist in any structured, machine-readable way that modern AI systems are trained on. No POS record, no API call, no inventory system, no digital customer history. The sale happens, money changes hands, but it leaves almost no usable signal for the systems increasingly tasked with understanding the economy.

And this is not an outlier.

It's the feed store running on handshake credit. The hunting guide who books by phone and collects cash at the trailhead. The farm stand that moves six figures in produce each season and issues zero digital receipts. The contractor whose customer list is a spiral notebook in his truck.

Such transactions are how a significant portion of the rural economy operates. The activity is very real, yet the data is sparse, delayed, fragmented, or non-attributable.

Some of that is due to infrastructure. The FCC estimates roughly 28% of rural Americans still lack access to modern broadband speeds, and Pew puts rural home broadband subscription well below suburban levels. But connectivity isn't the whole story. Even where the infrastructure exists, adoption is uneven. Cost, trust, habit, and perceived value all shape whether a business fully digitizes its operations.

The result is not that rural economies are invisible. It's that they are poorly instrumented. AI systems don't learn from reality. They learn from what's measured. And large parts of rural America aren't measured in ways those systems can easily ingest.

We already see the effects. Model performance degrades as you move away from dense, data-rich environments. Geographic bias shows up in everything from business recommendations to economic assumptions. Systems trained primarily on urban and suburban data generalize poorly to contexts they rarely encounter in structured form.

At the same time, the AI economy is concentrating in the opposite direction. Brookings Metro found that the San Francisco Bay Area alone accounts for roughly 13% of U.S. AI-related job postings, and about 30 metro areas generate close to two-thirds of the national total. The same locations building the models also produce the majority of the data informing those models — and the feedback loops that improve them.

That creates a compounding effect.

Rural economies generate less structured data. Models trained on existing data perform worse in rural contexts. Businesses then see less value in adopting those tools and adoption remains low. As a result, less data gets generated. The next generation of systems inherits the same bias, often amplified.

And the consequences go beyond inconvenience.

Credit models trained on incomplete signals misprice risk. Insurance models miss context. Supply chains optimize around what they can see. Policy decisions rely on datasets that systematically underrepresent entire categories of economic activity.

The issue isn't that rural America doesn't participate in the economy. It's that much of its participation isn't captured in the systems now shaping how the economy is understood.

I didn't move here to become an AI policy person. I moved here because I wanted to live differently. But when you participate in an economy, you begin to see how it is systematically mismeasured by technology. You understand the significant implications and how these will become greater problems if left unaddressed.

That ice cream shop is doing just fine without being part of the data economy. The question is: What happens as more decisions — financial, operational, and political — are made by systems that have no meaningful record of their existence?

And more importantly: Does rural America find a way to become instrumented on its own terms, or does it remain an afterthought in systems that increasingly define the mainstream economy?

Andrew Aitken is the Founder and Executive Director of the Center for Rural AI (ruralai.org), a 501(c)(3) nonprofit based in Durango, Colorado.

Rural Americans represent roughly 17% of the U.S. population — over 60 million citizens. According to Brookings Metro's July 2025 analysis of national AI activity, rural counties account for 0.3% of U.S. AI job postings, 0.3% of AI patents, and 1.5% of AI-related bachelor's degrees. AI startup and venture capital activity in rural counties is characterized as “virtually nonexistent.” However you measure it, rural America is capturing a fraction of AI economic activity proportional to its population — the gap is somewhere between stark and severe depending on how broadly you draw the boundaries.

I started the Center for Rural AI because I believe this is one of the most consequential economic challenges in the country right now, and because I think the conventional framing around it is wrong. The dominant narrative treats rural communities as disadvantaged recipients who need urban tech companies to charitably extend their solutions outward. That framing gets the problem backward. Rural communities have real advantages: 2 to 4 times lower operating costs than tech hubs, stronger talent retention once people can build meaningful careers locally, and environments that generate the kind of real-world edge cases that AI systems need. The issue isn't that rural America lacks potential. It's that the AI industry has been structurally set up to unconsciously ignore it.

The training data problem is more fundamental than most people realize

When an AI system underperforms in a rural context, broadband usually gets blamed. Connectivity matters, but the deeper problem is that most foundation models were built on data that treats rural America as nearly invisible.

Current foundation models draw 60 to 70% of their training data from web crawls that prioritize high-traffic, well-linked sites. Rural businesses, local governments, and community organizations have lower PageRank scores. They generate less digitized content. The curated text collections that make up another 20 to 30% of training data skew heavily toward R1 universities in major metros. The New York Times publishes roughly ten times more content about New York City than about all of Iowa.

A 2025 peer-reviewed study found that poverty-mapping AI performs significantly worse in rural areas than urban ones — not because of a technical flaw, but because those communities generate far less training data. Rural America is, in effect, statistically invisible to models built from, by, and for the web. Indigenous knowledge systems represent less than 0.1% of foundation model training data.

This creates real harm. An AI triage system might recommend immediate transport to a cardiac catheterization lab that is 120 miles from the rural critical access hospital where the patient is sitting. A weed identification tool for small farms — which the AgroBench evaluation published in July 2025 tested thoroughly — performs near random for most open-source vision-language models. Agricultural AI built for enterprise-scale precision farming, requiring $50,000 in sensors, does nothing for a diversified family operation in the San Juan Mountains. A healthcare AI trained on urban hospital data inherits assumptions about equipment, workflows, and demographics that don't transfer.

These aren't edge cases; they're the standard experience of rural AI users today.

What we're building

At the Center for Rural AI, we're working from a thesis that Fort Lewis College's AI Institute helped surface: the rural-urban AI gap isn't primarily a technology problem. It's a structural one. Rural communities lack the organized advocacy, coordinated infrastructure, training, and research programs to enable our citizens, small businesses, and institutional partnerships to claim a seat at the table where AI is being designed and deployed.

Our approach focuses on the 900-plus rural and tribal higher education institutions across the country, along with local and regional businesses and government. Colleges already sit inside their communities. They have trust. They have students who want to stay if there's something worth staying for. We're building the capacity for these institutions to become regional AI hubs, with AI Readiness Assessments designed specifically for rural contexts, open-source curricula that don't assume enterprise-scale infrastructure, and a hub-and-spoke model that lets us support hundreds of institutions and businesses without requiring each one to start from scratch.

The need is well-documented. More than 260 rural community colleges serve approximately 670,000 students annually (ACCT). According to TICAS research building on education geographer Nicholas Hillman's work, 3.1 million Americans live in education deserts — areas without a college within 25 miles — and 75% of those deserts are in rural communities. For a large share of students at these institutions, the nearest alternative is well beyond commuting distance.

Early evidence from AI adoption in higher education points to real opportunity for rural institutions specifically. Georgia State University's AI advising chatbot — studied via randomized controlled trial — reduced summer melt by roughly 4 percentage points among treated students, a meaningful outcome at an institution serving a large first-generation, lower-income population. Rural community colleges, which typically operate with minimal advising staff relative to enrollment, stand to benefit disproportionately from tools that extend institutional reach without adding headcount.

The economic opportunity argument is grounded in the same Brookings data that reveals the problem. Rural counties produce 1.5% of AI-related bachelor's degrees while representing roughly 17% of the population. They hold 0.3% of AI job postings. Applied to a U.S. AI economy that industry analysts project will exceed $1 trillion by 2030, that disproportion represents an opportunity gap well into the hundreds of billions of dollars. AI companies that understand rural markets, rural data, and rural deployment will have significant advantages as AI adoption spreads beyond major metros. Right now, almost no one is building for that.

What needs to change

Three things would move the needle.

First, the training data problem needs to be treated as an industry responsibility, not a charity project. Foundation model developers should actively partner with rural institutions, agricultural cooperatives, tribal nations, and rural health systems to incorporate representative data. This isn't altruism; it's how you build models that work for the full population. Datasets that exclude 17% of Americans will produce products that fail for 17% of Americans, and that failure has market consequences.

Second, federal agencies need implementation partners who understand rural contexts and can work at community scale. The EDA, USDA, NSF, and DOL have the funding authority. What they often lack is the rural-specific capacity to deploy it effectively. CRAI and organizations like us exist precisely to bridge that gap. The policy architecture is in place. The bottleneck is execution.

Third, rural communities need to be in the room where AI gets designed, not consulted after the fact. The RAISE AI Collaborative's approach of sitting down with rural teachers, parents, and students before building anything is a model worth replicating. The co-design principle sounds obvious when stated plainly: tools built without input from the people who will use them don't work as well for those people. But it remains the exception rather than the rule in how AI products get developed and deployed.

The choice before us

Here's what I tell communities, foundations, educational institutions, and tech companies I talk to: either rural communities help shape AI, or AI misses rural communities. The passive path leads somewhere predictable. The AI economy accelerates its geographic concentration. Rural talent pipelines drain faster as young people leave for places where the technology actually works. Communities that host AI data centers get construction jobs and property tax revenue but not economic participation. A gap that already looks like 0.3% versus 17% compounds further.

The active path requires treating this as the structural problem it is, not a connectivity challenge to be solved by the next broadband rollout. It requires AI companies to take rural data representation seriously. It requires federal programs to find implementation partners who can move at community speed. And it requires rural institutions to claim the capacity to lead, not just to receive.

That's what we're working on at the Center for Rural AI. The window to get this right is narrower than it looks.

Andrew Aitken is the Executive Director of the Center for Rural AI (ruralai.org), a 501(c)(3) nonprofit based in Durango, Colorado. CRAI is partnered with the AI Institute at Fort Lewis College.