Key Findings
- In a nationally representative survey, roughly half of districts reported that they have provided training to their teachers about generative artificial intelligence (AI)–powered tools, such as ChatGPT, as of fall 2024. This is double the proportion of districts who reported they had done so the previous fall.
- Another one-quarter of districts reported plans to provide such training for the first time during the 2024–2025 school year.
- According to district reports, low-poverty districts continue to outpace their higher-poverty counterparts in training teachers on AI use.
- In interviews, district leaders said that they focused initial teacher trainings on addressing teachers’ fear and discomfort with AI rather than jumping right into instructional tools.
- Many district leaders with whom we spoke had adopted a do-it-yourself approach to teacher training after struggling to find external partners to provide appropriate training.
Recent national surveys of teachers have found that about one-quarter are already using artificial intelligence (AI)–powered tools in their teaching, and there are more teachers who have at least tried using these tools or see themselves doing so more in the future (Diliberti et al., 2024; Kaufman et al., 2025; Peetz, 2024). Yet, as more teachers begin to use AI-powered tools in their instructional planning or teaching, little is known about how and to what extent their school systems are supporting them in navigating the rollout of AI in schools. As of fall 2024, 47 percent of teachers said that they had received at least some training on AI—a substantial increase over earlier that spring (Langreo, 2024). The teachers that have received some training typically said that this training has been a one-off, as opposed to ongoing professional development on AI (Peetz, 2024).
In this brief report, we update our understanding of how many districts have provided teacher training on AI use since we last investigated this topic in fall 2023 (Diliberti et al., 2024). That is, we provide an updated snapshot of the prevalence of district support of teachers’ training on the use of AI in schools. We examine which types of districts are leading the way on teacher training on AI, what this training looks like, and why districts decided to provide such training.
To do so, we use data from two sources. The first is survey data that we collected from national samples of K–12 public school districts in fall 2023 and in fall 2024. About 300 districts completed our fall 2024 survey, and about 200 districts completed our fall 2023 survey. For both surveys, we weighted districts’ responses to make them representative of all K–12 public school districts across the county at the time the survey was administered. In our surveys, we asked districts, “Has your district provided training to your teachers about use of generative artificial intelligence (like ChatGPT)?”[1] We instructed respondents to include trainings “about any of the following: understanding generative AI, considerations in the use of AI in the classroom, and students’ use of AI” when considering their responses. We interpret districts’ responses as an indication of whether they have ever trained their teachers on AI use. We did not ask our nationally representative samples of districts for more details about, for example, the duration of these trainings, the number of teachers who participated, whether the trainings were voluntary, or the specific topics they covered.
We did, however, conduct interviews with 14 district leaders who either started training their educators on AI or were planning to do so during the 2024–2025 school year. These leaders completed the fall 2024 survey on behalf of their districts and volunteered to speak with us about teacher trainings on AI in their district. They described what their districts’ training programs have focused on to date, how the programs were developed, and what their primary motivations were for running these programs. Although these interviews are not necessarily representative of the broader survey sample, they offer insight about how districts are rolling out teacher training around AI and what they are learning. For more details about how we conducted these surveys and interviews, please see the “Methodology” section at the end of this report.
This report is part of the American School District Panel (ASDP). The ASDP is a research partnership between RAND and the Center on Reinventing Public Education. The panel also collaborates with several other education organizations, including the Council of the Great City Schools and MGT. This report is intended to inform those who are tracking the rollout of AI into K–12 public education, including education policymakers, practitioners, and researchers.
Survey Results
By Fall 2024, Half of Districts Reported Having Provided Teacher Training on AI—Twice as Many Districts as the Previous Fall
In fall 2023, about one-quarter of districts (23 percent) reported that they had already trained their teachers on using AI (see Figure 1). However, 37 percent of districts told us at that time that they planned to provide such training at some point during the 2023–2024 school year. All told, 60 percent of districts that we surveyed back in fall 2023 reported plans to provide teacher trainings on AI by the end of the 2023–2024 school year (results not shown).
In fall 2024, we asked a new representative sample of districts whether they had trained their teachers on AI use. Their responses suggest that districts’ plans for the rest of the 2023–2024 school year largely—though not entirely—came to pass. In fall 2024, 48 percent of districts reported that they had trained teachers on the use of AI, which is a substantial, 25-percentage-point increase from the percentage of districts that reported implementing training on AI in fall 2023. This finding comports with results from a federal survey of school principals conducted in December 2024, which found that 59 percent of U.S. schools have trained some or all their teachers on use of AI (U.S. Department of Education, undated).
Figure 1. Percentage of Districts That Reported Having Provided Training (or Have Plans to Provide Training) to Teachers About AI Use, by Year
A line graph comparing the percentage of school districts that reported having provided AI training in the Fall of 2023, Fall of 2024 and those that plan to provide AI training in the Fall of 2025.Fall 2023Fall 2024Fall 2025(projected)01020304050607080234874
Projected (based on districts’ plans for the rest of the 2024–2025 school year)
Districts
- Fall 2023: 23
- Fall 2024: 48
- Fall 2025 (projected, based on districts’ plans for the rest of the 2024–2025 school year): 74
NOTE: This figure depicts response data from the following survey questions: “Has your district provided training to your teachers about use of generative artificial intelligence (like ChatGPT)?” administered in fall 2023 (n = 224), and “Has your district provided training to your teachers about generative artificial intelligence (AI) like ChatGPT? Include in your answer training about any of the following: understanding generative AI, considerations in the use of AI in the classroom, and students’ use of AI,” administered in fall 2024 (n = 290). Response options were “No,” “Not yet, but we plan to do so in [current school year],” and “Yes.” The solid line represents the percentage of districts that said “Yes” in fall 2023 and fall 2024, and the dashed line represents the percentage of districts that said “Not yet, but we plan to do so in 2024–2025.”
According to reports from district leaders, the share of districts that have trained teachers on AI is poised to grow even more. Another 26 percent of districts reported plans to provide teacher training on AI during the 2024–2025 school year. If districts’ plans for the 2024–2025 school year fully come to pass, about three-quarters of districts will have provided training on AI use by fall 2025. However, many—though not all—of these districts’ plans will come to pass, if what we observed between last school year and this one occurs again. The share of districts that provide teacher training on AI use is likely to continue to increase, although it might not grow to the level anticipated by districts at the time the survey was administered in fall 2024.
Low-Poverty Districts Continue to Outpace Their Counterparts in Training Teachers on AI, According to District Reports
Back in fall 2023, we observed that low-poverty districts had implemented more teacher training on AI use compared with their peers in higher-poverty districts, at least according to reports from districts. At that time, 43 percent of low-poverty districts reported that they had trained teachers on AI use, compared with 23 percent of middle-poverty districts and only 6 percent of high-poverty districts (see Figure 2). This represented a 37-percentage–point gap between low- and high-poverty districts.
Figure 2. Percentage of Districts That Reported Having Provided Training (or Have Plans to Provide Training) to Teachers About AI Use, by Year and Poverty Status
A line graph comparing the percentage of low-, middle-, and high-poverty school districts that reported having provided AI training in the Fall of 2023, Fall of 2024 and those that plan to provide AI training in the Fall of 2025.Fall 2023Fall 2024Fall 2025(projected)010203040506070809063923424367
Projected (based on districts’ plans for the rest of the 2024–2025 school year)
High-poverty districts
62
Middle-poverty districts
73
Low-poverty districts
87
| Districts | Fall 2023 | Fall 2024 | Fall 2025 (projected, based on districts’ plans for the rest of the 2024–2025 school year) |
|---|---|---|---|
| Low-poverty schools | 43 | 67 | 87 |
| Middle-poverty schools | 23 | 42 | 73 |
| High-poverty schools | 6 | 39 | 62 |
NOTE: This figure depicts response data from the following survey questions: “Has your district provided training to your teachers about use of generative artificial intelligence (like ChatGPT)?” administered in fall 2023 (n = 219) and “Has your district provided training to your teachers about generative artificial intelligence (AI) like ChatGPT? Include in your answer training about any of the following: understanding generative AI, considerations in the use of AI in the classroom, and students’ use of AI” administered in fall 2024 (n = 289). Response options were “No,” “Not yet, but we plan to do so in [current school year],” and “Yes.” The solid lines represent the percentage of districts that said “Yes” in fall 2023 and fall 2024, and the dashed lines represent the percentage of districts that said “Not yet, but we plan to do so in 2024–2025.”
This training gap persisted in the 2024–2025 school year. By fall 2024, 67 percent of low-poverty districts reported having provided training for teachers on AI use, compared with 42 percent of middle-poverty districts and 39 percent of high-poverty districts. Although high-poverty districts gained some ground relative to their low-poverty counterparts in terms of providing teacher training on AI, a notable gap remained.
We used districts’ reports about their training plans for the rest of the 2024–2025 school year to project what might happen by fall 2025, as illustrated by the dashed line in Figure 2. Districts’ responses suggest that we can expect this training gap to persist in the 2025–2026 school year. District leaders estimate that by the beginning of the 2025–2026 school year, almost all low-poverty districts will have trained their teachers on AI use, while only six in ten high-poverty districts will have done so. A training gap is likely to persist even if districts’ plans for the 2025–2026 school year do not come fully to pass.
Interview Results
District Leaders Said They Focused Initial Teacher Trainings on Addressing Teacher Confusion and Fears; Some Districts Introduced AI Productivity Tools but Few Addressed Student AI Use
In interviews, nearly all school district leaders said that the primary purpose for their AI training was to address teachers’ concerns, confusion, and fears about the technology. Of the 14 interviewees who had trained teachers or were planning on doing so, 13 of them described encountering teachers with negative views of AI and reported that some teachers view it as a threat to traditional teaching methods or a tool for student cheating. By providing training around the fundamentals of AI’s capabilities and limitations, leaders aimed to lower anxiety, help teachers understand AI’s potential benefits, and shift away from an antagonistic, cheating-centered mentality. As one superintendent said, the initial goal of their training was “to lower the fear, lower the anxiety about students using it and encourage [teachers] to use it in whatever capacity . . . just play around with it, mess around with it.”
A majority of interviewees also had an interest in empowering teachers to leverage AI effectively in their work and prioritized demonstrating how it could assist with such tasks as lesson planning and content differentiation. For teachers that had moved from concern to curiosity regarding the technology, leaders introduced tools, such as ChatGPT, or education-specific software, such as MagicSchool. However, it is noteworthy that few leaders engaged in any training around student use of AI tools. Interviewee responses suggest that districts are taking a cautious approach, focusing first on educator proficiency before integrating AI into student learning experiences. One leader succinctly described their thinking behind this prioritization: “Why don’t we see what AI can do for teachers? And then once they get comfortable with it, then we can start having those conversations about . . . what we’re going to accept or not accept from students.”
Training Formats Varied Widely, Although Nearly All Training Was Optional
District leaders reported employing a wide variety of AI training formats with one common theme: Virtually all were opt-in trainings. Only one of our interviewees reported running a mandatory training session for all teachers in their district. Instead, leaders reported taking a more voluntary, buy-in approach, which one leader summarized, saying, “I am not trying to push AI with my people, but I am trying to give them the tools if they’re interested. I know that as [AI adoption] starts happening, it will grow organically and very quickly.”
Interviewees reported running training sessions that ranged from full days to providing more ongoing, targeted “bite-sized” learning opportunities that addressed the practical use of specific AI tools through newsletters and district listservs. In several districts, leaders also embedded AI topics in existing professional development sessions. Some districts took a more targeted approach, offering stand-alone training sessions at individual school sites. To both generate and disseminate knowledge about the ways that AI could support educators’ work, a few district leaders implemented peer leadership models. One district leader explained the value of teacher-to-teacher training, saying,
It’s one thing for me to say [to teachers], “Oh, AI is really cool,” but I need some teachers to go, “No, it really helped me formulate this test question or that test question.” And I think that gets more buy-in.
Several leaders reported using a combination of formats to both provide more entry points to learn about AI and promote sustained awareness of it. One superintendent said
We started [our in-service training] from a place of play. . . . Now, we create little [informational] videos and send them out on an ongoing basis. . . . What we find with professional development is when something comes to you in short bits, it’s one of the most effective ways to keep a topic going.
Innovative Approaches to AI Training
In our interviews with school district leaders, we uncovered several innovative approaches to AI training that stood out because of their creativity and compelling strategies. The following examples highlight how some districts are tackling the challenge of preparing educators for AI integration, often in the absence of established best practices or external expertise.
- In one urban mid-Atlantic district, a proactive high school student was alarmed at their teachers’ growing resistance to AI and advocated for district leadership to help educators understand and engage with the technology. The superintendent heeded this call and developed an organic, opt-in approach to training that encouraged building leaders to convene teacher-led AI Exploration teams. These teams curated knowledge and resources about AI tools that were relevant for their educators’ specific needs and contexts and then disseminated their learnings through monthly faculty meetings.
- A suburban West Coast district adopted a phased approach to AI training, starting with introductory “play” sessions designed to alleviate teacher anxiety and shifting toward more–instructionally focused learning. After hosting several stand-alone workshops on AI basics, the district developed a strategy for sustained learning, including the creation of an ongoing video series with “bite-sized” lessons on specific AI tools and embedding AI topics into existing professional development. Early results suggest that educators are feeling more comfortable with AI and are starting to experiment with ways that AI can decrease their workload and improve the quality and creativity of their lessons.
Personal Interest, Conflicting Priorities, and a Shortage of Outside Expertise Led Leaders to Take a Do-It-Yourself Approach to Developing AI Training Programs
Of the 14 district leaders interviewed, 11 took a do-it-yourself approach to developing their district’s AI training programs. This approach required the leaders to determine learning objectives, create or curate technical content, and design training formats either from scratch or with a patchwork of internal and external resources. More than one-third of the leaders we interviewed were personally engaged in implementing their districts’ training. These superintendents were often motivated to take such a hands-on approach because of their own learning and experimentation with AI tools. One interviewee’s negative experience at an AI training pushed them to create their own program in house: “The first AI [professional development] I got as a superintendent was super overwhelming and it scared me. . . . [If this is what teachers get], they won’t want to do anything with AI and just put blinders on.”
This do-it-yourself approach, though potentially employed to meet specific district needs, also reflects a scarcity of external experts who are capable of providing appropriate training. Half of the leaders interviewed reported struggling to find well-established experts knowledgeable about AI in educational contexts. One leader said candidly,
There are people that are claiming to have the best practices and are making money hand over fist. A lot of the [AI] workshops that I’ve attended, I’ve just come out of feeling like, “Okay, that was a waste of time and money.” I just want to tell everybody that if they claim to be telling you best practices, they don’t have them yet. They don’t exist yet.
Leaders turned to a variety of external sources to inform their professional development efforts, including state school board associations and educational technology organizations, such as Digital Promise and the International Society for Technology in Education. Some leaders also sought insights from tech companies, such as Google and OpenAI, to piece together a comprehensive training program.
Additionally, the need to balance AI training with other critical professional development areas created a complex landscape for district leaders to navigate, potentially limiting the depth and breadth of AI-focused training opportunities. As one superintendent explained, ‘The biggest barrier is time. We have so many competing priorities for professional development.”
What We Are Hearing from the Council of the Great City Schools
Our ASDP partners at the Council of the Great City Schools (CGCS) also told us a little bit about how they are helping develop supports for the use of AI in their member districts. CGCS has made concerted efforts in recent years to improve the rates of AI training in their member districts. As part of these efforts, CGCS, in partnership with the Consortium for School Networking and supported by a grant from Amazon Web Services, has developed the K–12 Gen AI Maturity Tool (Consortium for School Networking, 2024) and the K–12 Gen AI Readiness Checklist (CGCS, 2023). These tools serve as vital resources for assessing the state of AI readiness within school districts across six major domains: Leadership, Operational, Data, Technical, Security, and Legal Risk. CGCS conducted a half-day, face-to-face workshop with leadership teams from four school districts. The workshop’s objectives were to validate self-assessment findings from the K–12 Gen AI Maturity Tool and establish future organizational goals for Gen AI. This initiative also aimed to identify cross-functional leadership roles to guide policy development, manage risks, develop procedures, and coordinate AI implementation. CGCS also initiated an AI cross-functional planning team to provide ongoing direction, resources, and support to member schools. Its goal is to address ongoing policy, processes, and guidance for AI implementation.
Implications
Implication 1. Training and support organizations can help better address educator fear and reluctance to use AI by taking these concerns seriously. In interviews, nearly all school district leaders said that their primary focus for their AI training was to address teachers’ concerns, confusion, and fears about the technology. This echoes findings from our prior survey of teachers; we found that teachers’ fears about the rollout of AI into public education—including concerns about AI as whole, data privacy issues, and potential biases built into tools—are common among both AI adopters and nonadopters (Diliberti et al., 2024). These concerns should not be minimized by district leaders and should instead be openly discussed. The fact that nearly all the interviewed leaders felt compelled to develop their own introductory AI training suggests that many of the existing external resources might not be helpful in addressing teachers’ initial misconceptions and discomfort. Teacher training organizations (e.g., schools of education) and technical assistance providers should design AI professional development that explicitly starts with trust-building and emphasizes an understanding of AI fundamentals before moving on to practical applications and student-centered integration. Such support should include hands-on learning opportunities to experiment with new technology in playful and low-stakes contexts. Teachers would also benefit from peer-based learning structures that encourage knowledge-sharing and collective sensemaking. These activities draw on established principles of effective professional learning, such as active engagement, sustained duration, and collaboration (Desimone and Stuckey, 2014), while addressing the unique affordances and challenges posed by generative AI.
Implication 2. Districts—especially those serving students in high-poverty schools—likely need additional support to prepare their teachers for AI. Although middle- and high-poverty districts are starting to offer more teacher training on AI, they remain far behind low-poverty districts. This training gap is likely to persist based on districts’ plans for the rest of the 2024–2025 school year. The faster take-up of AI in historically advantaged settings raises concerns about wide disparities in teachers’ and students’ opportunities to learn with these tools—with the notable caveat that it remains unknown to what extent adoption of these generative AI tools will improve teaching and learning. But, certainly, whatever best practices arise for teachers’ AI use in schools should be equitably shared. To close the AI training gap, state and federal education agencies, philanthropy, and technical assistance providers should prioritize targeted funding and support for high-poverty districts. This support could include regional, state, or national organizations trusted by districts that can provide model professional development units and connections to vetted professional development providers. Additionally, state and regional education networks should facilitate knowledge-sharing between districts to ensure equitable access to AI-related capacity-building opportunities.
Methodology
Survey Data
Our methodology for analyzing ASDP survey data remains relatively consistent across survey waves; therefore, the description of our methods here is text that we updated from a previous publication (Diliberti et al., 2024). The fall 2023 ASDP was administered to a national sample of K–12 public school districts who are members of the American School District Panel between October 12, 2023, and December 14, 2023. All enrolled member districts were invited to complete the survey. Of the 1,167 public school districts that we invited to take our survey, 231 districts completed our survey (a 19.8-percent completion rate). In fall 2024, we expanded to invite non-ASDP member districts to complete our surveys. The fall 2024 ASDP was administered to a national sample of K–12 public school districts between October 24, 2024, and December 23, 2024. Of the 8,017 public school districts that we invited to take our survey, 291 districts completed our survey (a 4-percent response rate). We design our ASDP surveys to allow multiple respondents from the same district central office to complete portions of the survey—for example, a superintendent, human resources director, or research director might answer questions about district staffing levels, while an academic director might complete questions about math instruction. We do not know which person(s) in each district completed the survey on behalf of their district. Estimates for each survey were separately produced using cross-sectional survey weights designed specifically to provide nationally representative estimates at the time when the survey was administered. To produce these weights, we obtained data on district demographics by linking survey data files to the Common Core of Data issued by the National Center for Education Statistics (NCES). Importantly, survey responses were weighted to be representative of the national population of public school districts, not the national population of public school students. For more information about the weighting procedures for ASDP surveys, see Diliberti et al., 2025. We obtained data on the district poverty level from the U.S. Census Bureau’s Small Area Income and Poverty Estimates Program School District Estimates. We divided public school districts into quartiles using the family poverty rate of their 5- to 17-year-old population in the district’s attendance boundary. Per NCES guidance, we chose our cut points for these quartiles such that each quarter contains roughly the same number of students. Low-poverty districts are those in the first quartile (that is, those with the fewest families with income below the federal poverty rate). Middle-poverty districts are those in the second and third quartiles. High-poverty districts are those in the fourth quartile (that is, those with the highest shares of families with incomes below the federal poverty rate). In each survey year, we conducted significance testing to assess whether subgroups were statistically different at the p < 0.05 level. However, we did not conduct formal significance testing of differences across survey waves (e.g., comparing districts’ responses on survey items in fall 2024 versus fall 2023) because of a lack of longitudinal survey weights that properly account for the partial overlap in respondents and changes in representativeness of survey respondents across years. Some differences that appear to be changes over time might be due to uncertainty associated with survey estimates in each time period. In the text, we describe only those differences among district subgroups that are statistically significant at the 5-percent level. Furthermore, because of the exploratory nature of this study, we did not apply multiple hypothesis test corrections.
Qualitative Interviews
Between November 2024 and January 2025, we conducted qualitative, semistructured interviews with 15 district leaders who responded to the 2024 ASPD survey. Participants represented a variety of school system sizes and geographic areas, including rural districts with fewer than 1,000 students and large urban ones with more than 75,000 students. The sample was largely composed of district leaders who have already trained teachers on AI or were planning to do so in the near future. Only one of the 15 interviewees had no plans for formal AI teacher training. The interviews covered a variety of topics related to AI in K–12 school districts, including leaders’ motivations for developing AI training, goals for and formats of professional development, challenges faced, and resources used. These interviews lasted between 20 and 40 minutes and were audio recorded and transcribed. Researchers coded the data using deductive themes based on the interview protocol and employed an analytic matrix to track patterns across respondents. Interview protocols are available on request.
Acknowledgments
We are extremely grateful to the educators who agreed to participate in the panels. Their time and willingness to share their experiences were invaluable to this effort and helping us understand how to better support their hard work in schools. We thank Daniel Ibarrola for helping manage the survey; Gerald Hunter for serving as the data manager for this survey; and Tim Colvin, Roberto Guevara, and Julie Newell for programming the survey. Thanks to Claude Messan Setodji for managing the sampling and weighting for these analyses. Thanks also to AK Keskin, who assisted with some of the survey analyses. We greatly appreciate the administrative support provided by Tina Petrossian and Erin Levendorf and ASDP management provided by Samantha DiNicola. We also appreciate feedback from Heather Schwartz, Lydia Rainey, Sarah McCann, and Ben Master in drafting this report. We thank Morgan Polikoff and Elizabeth Steiner for helpful feedback that greatly improved this short report. We also thank Melissa Parmelee for her editorial expertise and Monette Velasco for overseeing the publication process.
Note
- [1] We did not directly specify for districts whether they should include in their responses only trainings that occurred in the current school year or trainings that occurred in any preceding school year. Because no period was included in the survey item wording—and because the rollout of AI into public education is relatively new—we assume that districts interpreted the question as asking whether they have ever trained their teachers on AI use. If districts assumed that we were asking about trainings only in the current school year, it is possible that we undercounted districts that previously provided training but did not in the current school year.
Source: rand.org
