How K-12 Educators Use Ai: Llm-Assisted Qualitative Analysis At Scale

Alex Liu, Lief Esbenshade, Shawon Sarkar, Zewei Tian, Zachary Zhang, Kevin He, Min Sun

Date

12/2025

AI Overview

Why it matters

Understanding how practicing K-12 educators use generative AI at scale in authentic professional settings remains limited, creating a critical gap for designing responsive tools and systemic supports.

Study at a glance

Question: How do K-12 educators use generative AI tools to support their instructional and professional responsibilities?
Setting and data: Over 140,000 messages across 13,071 conversations from K-12 educators using an open-registration AI platform, May 2025.
Method: Human-AI collaborative qualitative coding pipeline combining grounded theory methods with LLM-assisted coding and descriptive analysis.
AI tools: Claude 3.5 Haiku, Claude 3 Sonnet, ChatGPT-4o

Key findings

Instructional practices appeared in 80.1% of educator conversations, led by planning (63.8%), explicit teaching (45.9%), assessment (42.4%), and differentiation and accessibility (36.9%).
Nearly 88% of conversations involved multiple instructional categories, such as combining planning with explicit teaching (41.0%), critical thinking (35.9%), and differentiation (30.1%).
Educators frequently refined AI feedback and scaffolds, while AI models often introduced unprompted instructional additions such as tiered supports and anticipated misconceptions.

Summary drafted with AI from the full paper and may contain errors. See the full paper for details.

This study investigates how K-12 educators use generative AI tools in real-world instructional contexts and how large language models (LLMs) can support scalable qualitative analysis of these interactions. Drawing on over 13,000 unscripted educator-AI conversations from an open-access platform, we examine educators' use of AI for lesson planning, differentiation, assessment, and pedagogical reflection. Methodologically, we introduce a replicable, LLM-assisted qualitative analysis pipeline that supports inductive theme discovery, codebook development, and large-scale annotation while preserving researcher control over conceptual synthesis. Empirically, the findings surface concrete patterns in how educators prompt, adapt, and evaluate AI-generated suggestions as part of their instructional reasoning. This work demonstrates the feasibility of combining LLM support with qualitative rigor to analyze complex educator behaviors at scale and inform the design of AI-powered educational tools.