The Growth Number Looks Good Until You Check Who Showed Up
The EdWeek Research Center's survey of 852 educators, teachers, administrators, and district leaders together, fielded between August 26 and October 6, 2026, found roughly two-thirds reported at least one AI-related professional development session this past year, up 11 percentage points from 56 percent the year before. On a pure participation chart presented to a school board, that is a strong trend line for any district technology office trying to demonstrate visible adoption progress on an AI initiative it has funded.
Dallas high school math teacher Lance Barasch, who described himself to EdWeek as a self-identified AI champion on his campus, undercuts that headline number from the inside once you ask who actually walks through the door. He said many districts do not mandate or even strongly encourage AI training, so attendance skews heavily toward educators already interested while skeptics mostly stay away entirely. A participation rate built mostly from self-selecting enthusiasts tells a district far less about actual districtwide readiness than the topline number suggests, and it leaves the teachers least comfortable with AI the least likely to ever get trained on it properly.
Teachers Keep Asking for Practice, Not Demos
West Virginia high school science teacher Jaime Moss put the core complaint about current AI training plainly to EdWeek: "Teachers don't want another demonstration of what AI can do." That is a specific, highly actionable critique of how most district professional development is currently structured, built as a one-way presentation format rather than a hands-on practice format, and it echoes a pattern that procurement and learning-and-development teams in other industries have already relearned the hard way with earlier waves of enterprise software rollouts and training programs.
Rita Mortenson, an instructional coach based in Verona, Wisconsin, described the resulting gap in concrete terms: prompts demonstrated cleanly during a training session often "don't work the way you want them to work" once a teacher actually tries them against real classroom material back at their desk. Her recommendation is to teach the full iterative process of refining AI output over several attempts, supported by genuine practice materials such as sample spreadsheets and real student work, rather than relying on a single polished example that quietly collapses the moment conditions change even slightly in a live classroom.
Barasch Hits the Same Wall From the Other Side
Barasch, who runs informal training sessions for colleagues at his own campus on top of his regular teaching load, said the sessions he personally leads suffer from a structural scheduling problem: he struggles simply to find time to train other teachers at all during the normal school day. "When I get back to the campus, I have to scrape and claw to get that time in front of my colleagues," he told EdWeek, describing AI training as something perpetually squeezed into existing schedules rather than properly resourced as its own standalone priority with protected time.
He also said his own trainings lack the instant feedback loop he personally gets when attending outside professional sessions run by vendors or conferences, meaning even the educators most invested in teaching AI skills to their peers are working without the same iteration and support structures that made external training genuinely effective for them in the first place. That points to a resourcing failure at the district level, not a content failure, and it will not be solved simply by a better slide deck or a new vendor-supplied curriculum module dropped into an existing PD day.
The Question Nobody Is Training Teachers to Answer
Barasch said the question teachers ask most often, by a wide margin, is how to detect and prevent AI-enabled cheating in their own classrooms, and that many trainings simply do not cover this at all. He described the overall tenor of most AI professional development as "all-in on the new stuff" rather than addressing the real drawbacks, which leaves teachers unable to respond credibly when students raise the exact concerns the training itself skipped entirely. Questions about AI's effects on the environment, on future jobs, and on critical thinking get similarly thin coverage and are sometimes sidestepped outright by presenters.
That gap matters commercially just as much as it matters pedagogically for classroom outcomes. Any vendor selling AI-literacy or professional-development content into districts that cannot credibly answer the cheating-detection question within its own core curriculum is selling an incomplete product relative to what the EdWeek data shows teachers are actually asking for, repeatedly, in session after session. A training program built entirely around capability showcases, without a credible, practical answer to the integrity question, is solving only the easier half of the actual problem districts are facing.
What This Means for District PD Budgets
The actionable gap for a district technology or curriculum office right now runs three ways: make AI professional development something closer to mandatory rather than purely opt-in, rebuild session design around guided hands-on practice rather than one-way demonstration, and explicitly budget protected time for addressing drawbacks, cheating detection chief among them, rather than treating capability training and integrity training as separate line items that someone else will presumably cover later in the year.
None of those three fixes require purchasing any new AI tooling at all. They require redesigning how existing professional development time actually gets spent, and committing real scheduling authority to protect that redesigned time, rather than leaving it, as Barasch described plainly, something individual educators scrape together on their own initiative. For a district that has already invested meaningfully in AI licensing, this survey is a clear reminder that the real bottleneck has moved from access to instructional design capacity, and that particular gap does not close itself just because attendance numbers happen to be trending upward.


