We have all sat through that vendor demo. Teachers in a cafeteria on an in-service day, watching someone from a detection tool company demonstrate features for 90 minutes. Half the room has never used ChatGPT. The other half has been using it daily for a year. The presentation helps neither group. This is the state of AI professional development in most districts, and it needs to change.
The AI Training Gap
What the Numbers Say
By fall 2025, roughly 50% of K-12 teachers had received some form of AI-related training, according to EdWeek Research Center surveys. That sounds like progress until you look at what "training" means. For most teachers, it was a single awareness session, not sustained skill-building.
61% using AI, few trained
61% of teachers reported using AI in some capacity in 2025. Most are self-training through trial and error rather than structured PD (EdWeek Research Center, 2025).
The equity gap in training
Teachers in high-poverty districts are significantly less likely to receive AI training than those in affluent districts, widening an already existing professional development equity gap (RAND Corporation, 2024).
One Workshop Is Not Training
The gap is not just about quantity of training but quality. A one-hour overview of ChatGPT features does not prepare a teacher to:
- Evaluate whether a student's essay was AI-generated
- Redesign assignments to be AI-resilient
- Have honest conversations with students about appropriate AI use
- Understand the limitations of detection tools
- Navigate academic integrity cases with fairness
These skills require practice, discussion, and iteration, not a single afternoon session.
Why Top-Down AI PD Fails
Two patterns dominate failed AI professional development:
The Vendor Demo Problem
A detection tool company runs a training session that doubles as a sales pitch. Teachers learn the tool's interface but not the pedagogical judgment required to use it well. When the tool flags an essay, teachers still have no framework for what to do next. Is this a false positive? How do I have this conversation? What does "75% AI probability" actually mean?
The vendor has no incentive to discuss limitations. The school has paid for a training that does not build real capacity.
The Compliance Checkbox Problem
A district mandates a 2-hour AI awareness module to satisfy a board directive. The content is generic, disconnected from grade levels or subject areas, and delivered by someone who has never taught. Teachers complete it, check the box, and go back to figuring things out on their own.
Both approaches share the same flaw: they treat teachers as recipients of training rather than professionals with expertise. The Caucus of Working Educators recognized this problem a decade ago in the context of testing mandates and curriculum adoption. The same pattern is repeating with AI.
What Teacher-Led AI PD Looks Like
Three models are working in real schools:
Small groups of 6-10 teachers meeting biweekly to share what they are trying, what is working, and what is failing. No outside facilitator needed. The expertise is in the room. Teachers bring real cases: a flagged essay, a redesigned assignment, a conversation that went well or poorly.
Instead of starting with a tool demo, start with a classroom problem: "Three students submitted suspiciously similar essays. What do I do?" Work backward from the problem to the tools, strategies, and policies that address it. This grounds training in real needs, not product features.
AI changes faster than any other technology teachers have encountered. A training from September is partially outdated by January. Effective PD builds in regular check-ins and updates, not a one-and-done session. Monthly 30-minute meetings beat annual 3-hour workshops.
What Research Says
Research from the Learning Policy Institute consistently shows that effective professional development is sustained (20+ hours over time), collaborative, content-specific, and embedded in teacher practice. Single-session workshops, the dominant model for AI PD, meet none of these criteria.
A Practical Framework for Schools
Here is a simple, repeatable structure schools can adopt:
Months 1-2: Awareness
What AI tools exist? What are students using? What do detection tools claim versus what they deliver? Build baseline understanding across the staff.
Months 3-4: Classroom Application
Each teacher redesigns one assignment to be AI-resilient. Test detection tools on known AI and human writing samples. Share results and iterate.
Months 5-6: Policy Integration
Develop department or grade-level AI use policies that teachers actually helped write. Create shared language and expectations for student communication.
Ongoing: Monthly Check-ins
30-minute monthly meetings to share updates, new tools, new student behaviors, and emerging challenges. Keep the conversation alive.
What Districts Should Fund (and Stop Funding)
- Release time for teacher-led cohort meetings
- Substitute coverage for peer observation
- Stipends for teachers developing AI curriculum
- Cross-school learning networks
- Teacher-chosen conference attendance
- Vendor-led training sessions disguised as PD
- Generic online compliance modules
- Outside consultants who have not taught recently
- One-shot awareness workshops
- "Train the trainer" models that overburden a few teachers
The Equity Dimension
The Training Equity Gap
Schools serving predominantly low-income students are less likely to have up-to-date technology, dedicated instructional coaches, or budget for external PD providers (NCES, 2023). This means the teachers who most need AI training are the least likely to receive it.
The AI training gap follows the same pattern as every other educational inequity: schools that can afford Turnitin subscriptions can also afford to send teachers to AI conferences. Schools that cannot afford the subscription also cannot afford the training.
Teacher-led PD is not just better pedagogy. It is a more equitable model. Peer learning does not require expensive outside facilitators. Problem-based cohorts do not require conference travel. The expertise is already in our buildings, waiting to be unlocked.
Working Educators built its reputation on teacher-led professional development long before AI was a concern. We know this model works because we have lived it. The challenge now is applying what we know to a rapidly changing technology landscape.