The State of K-12 AI Readiness, 2026.
How 117,704 US public school districts and schools publicly talk about artificial intelligence. A snapshot across all 50 states and the District of Columbia, drawn from board minutes, handbooks, policies, and press coverage published before May 2026.
By the spring of 2026, roughly one in five US public K-12 districts and schools had made their AI work visible — through a published policy, an approved-tools list, classroom guidance, vendor adoption, or governance language. The other four-fifths had not. The data behind this brief was drawn directly from those public artifacts, then synthesized into a comparable reading for each district and school. The pages that follow walk through what we found: where adoption is concentrated, which tools are doing the work, and where policy is running ahead of practice.
Five things to know.
Each finding is anchored by a section below, with the data behind it. Numbers are as of May 2026.
About 1 in 4 US school districts show some public AI activity, but most are still talking, not doing. Across all 18,301 districts, 22.6% have public AI signal of any kind; only 13.0% are at the "guided use" or "implementation" stages where AI is woven into daily work.
Maryland leads the country. 27.7% of its 1,439 districts and schools show public AI signal — +9.6 points above the national rate of 18.1%. The five highest-adoption states cluster in the Northeast and Upper Midwest; the lowest-adoption five sit in the Mountain West and the Deep South.
ChatGPT is the dominant tool by a wide margin. 3,121 districts and schools name it in their AI evidence — nearly 1.8× the next most-cited vendor, Google Gemini (1,775). The long tail thins fast: by vendor #10, the count is under 300.
Policy is running ahead of practice. 12.8% of US districts have published an AI policy, but only 5.8% show visible day-to-day implementation. The gap is the talk-vs-do distance — districts are writing rules faster than they are putting them to work.
The silent majority is still the majority. 81.9% of K-12 districts and schools — 96,408 of them — have published nothing public and measurable about their AI work. Some are deliberately quiet; many have simply not gotten there yet. Either way, the headroom for adoption is enormous.
Most US districts and schools are still silent on AI.
Where every US district and school sits on the public AI adoption arc, as of May 2026. The grey band is everyone who hasn't published anything measurable yet; the red bands are the visible tail moving from awareness to implementation.
All US districts and schools· 117,704 districts + schools
Most of the country is in the grey band — no published AI artifact of any kind.
Districts only· 18,301 US school districts
Most public AI artifacts are published at the district level, so the colored tail is materially larger here than in the combined cut above.
The visible tail is concentrated at the back end of the arc. 11,967 districts and schools — 10.2% of the national total — are at "guided use" or "implementation visible." Restricting to districts only, 13.0% of US districts are in that same tail (2,381 districts). The bigger gap nationally is at the front end: 9,329 districts and schools — roughly 7.9% — have talked about AI in public without committing to a formal posture yet.
Policy edges out practice. Both are rare.
Where every district and school sits on the AI policy authoring lifecycle. 9.9% of all districts and schools have reached "published" or "board-adopted." At the district level — where most AI policy gets authored — that share rises to 12.8%.
All US districts and schools· 117,704 districts + schools
Most schools don't author their own AI policy — they inherit their district's. This bar reflects that.
Districts only· 18,301 US school districts
Most AI policy gets authored at the district level, so the colored share here is the more honest read of policy adoption.
Both numbers are small, and the gap between them is partly a measurement story. 12.8% of the 18,301 US districts have published a formal AI policy; 5.8%have visible day-to-day AI work — approved tools in use, classroom evidence, staff training in place. Policy is, by design, a public artifact: board minutes, board-adopted documents, handbooks. Practice often isn’t — a classroom using ChatGPT every day produces no public record unless the district writes one. Read the 7.0-point gap as the gap between two paper trails, not between thinking and doing. Even so, districts that have policy and no visible practice are where partners, vendors, and professional-learning providers can most clearly add value.
Adoption is regional.
Share of each state's districts and schools with any public AI signal. The national rate is 18.1%; the top ten states average 24.7%; the bottom ten average 8.0%.
ChatGPT dominates. Everyone else is fighting for second place.
Vendors named in public AI readiness evidence by US districts and schools.
A vendor’s count is the number of districts and schools that have publicly named it — in a board minute, approved-tools list, policy, staff handbook, classroom guidance doc, or press release. It is a floor on real adoption, not a ceiling.
That gap matters. Vendors with strong direct-to-teacher distribution will routinely self-report customer numbers (e.g. “serving 10,000 schools”) that are an order of magnitude larger than what shows up in formal district documents, because a teacher signing up a class doesn’t produce the kind of public artifact this brief reads. Both numbers can be honest. They measure different things: provisioned access versus publicly documented adoption.
Read this leaderboard as an ordering of which vendors are making it into the official record of US K-12, not as a market-share table.
Note · In K-12 documents, “ChatGPT” is often used as a generic synonym for generative AI — the way “Kleenex” stands in for tissue. This count almost certainly overstates first-party ChatGPT adoption and understates Google Gemini, Microsoft Copilot, and the long tail. Read this row as “districts naming a chatbot,” not “districts that chose OpenAI.”
What’s most striking about this top ten is the company it doesn’t keep. None of the traditional ed-tech incumbents — the LMS vendors, curriculum publishers, or assessment giants that have defined school technology for the last decade — show up here. Some of those incumbents are quietly adding AI features to existing products, but, as of this snapshot, districts and schools aren’t writing about them by name. The AI vendors that are getting written into board minutes and policies are a different set entirely.
The presence of two Google products in the top ten (Gemini and NotebookLM) reflects Google’s reach into the K-12 desktop more than it reflects explicit district-level adoption decisions. Conversely, the rise of purpose-built education AI tools — MagicSchool AI, Khanmigo, SchoolAI, Brisk Teaching — is the stronger signal: a district that names one of these in a policy or board doc is naming it because they chose it, not because it was already on the device.
Districts can declare AI faster than they can run it.
How many districts and schools publish each kind of AI signal. A district can publish more than one — these are not exclusive categories — so the rows do not sum to the universe.
Six of these eight signals cluster between roughly 9% and 14% of US districts and schools. The two that don’t — training (6.8%) and privacy or procurement review (3.8%) — are also the two signals that require the most operational lift. A board can name an AI committee or adopt a policy in a single meeting. Standing up a staff-wide training program, or running every new AI tool through a privacy and procurement gate, is real ongoing work. The shape of this mix says districts are getting the declarative pieces of AI readiness in place faster than the operational pieces, which is worth knowing if you sell training, governance services, or compliance tooling into K-12.
How this was measured.
Universe. The denominator for every rate in this brief is the full set of 18,301 US public school districts plus 99,403 schools (117,704in total), drawn from Bellwork's catalog across all 50 states and the District of Columbia. When the brief uses a phrase like "1 in 5 districts and schools," that ratio is taken against this combined denominator. US territories and the Bureau of Indian Education are not included; their public-web AI coverage is still too sparse to report honestly.
Evidence. A district or school is counted as having a public AI signal when our crawler has surfaced and our extraction pipeline has classified at least one of: a published AI policy or guidance document, an approved-tools list, classroom or instructional use evidence, a named AI vendor, a privacy or procurement review touching AI, or governance language (committee, owner, structure). The five-stage adoption arc ( no signal · aware · planning · guided use · implementation visible) is derived from those underlying signals.
What "no public AI signal" means. The grey band in the adoption charts above is 96,408 districts and schools for which we have not surfaced any public AI evidence as of May 2026. Some of these districts are deliberately doing AI work in private (board executive sessions, internal training, unpublished pilots). Some have simply not put anything in writing publicly yet. The brief does not distinguish between those two cases.
Vendor catalog.Bellwork maintains a curated catalog of AI vendors used in K-12, and the top vendors list is resolved against it. Bare mentions of generic ed-tech vendors (e.g. "Google" referring to Workspace or Chromebooks) are excluded from the AI vendor count when the evidence does not name a specific AI product.
Limitations.Every count in this brief is a floor on underlying behavior, not a ceiling. A district can use ChatGPT, MagicSchool, Khanmigo, or any other tool in classrooms every day without that usage ever appearing in a board minute, a handbook, an approved-tools list, or a press release — and if it doesn't appear there, this brief can't see it. As a result, the vendor counts in particular will look smaller than numbers vendors report themselves, especially for products with strong direct-to-teacher distribution. A vendor that publicly claims to serve tens of thousands of schools and that we count at a few hundred is not a contradiction; the vendor is counting provisioned accounts and we are counting district documents that name them. Both numbers are useful for different questions. This brief answers the documented-adoption question.
Refresh cycle. Bellwork subscribers see this picture updated continuously as new evidence is published. This page is the annual public snapshot — frozen so any number cited in the prose remains verifiable against the chart above. Future editions will appear at their own URLs (/reports/ai-readiness/2027 and so on).
Built for outreach and research.
Cross-filter all 117,704K-12 districts and schools by AI readiness alongside test scores, demographics, enrollment trends, vendors in place, and dozens of other signals — then pull the names and contact details of the administrators, IT directors, and curriculum leaders behind every district and school in the cut you care about. Whether you’re trying to reach those schools or doing the kind of research this brief is built on, the dashboard is where you actually work.