Since rolling out an artificial intelligence tool last school year, one Texas school district has cut the time it takes to craft a special education student’s individualized education program (IEP) from hours to an average of just 45 minutes.
Mesquite independent school district, in a suburb just east of Dallas with about 38,000 students, developed an AI tool that consolidates data from across district platforms and helps special educators draft more nuanced, personalized IEPs.
AI has the potential to save special educators a tremendous amount of time on paperwork that siphons time away from classroom instruction, experts say. But without proper guardrails and training, districts risk running afoul of federal student privacy laws or removing the critical human element from IEP development.
Mesquite leaders believe the custom tool has bucked concerns about student privacy while keeping teachers as the main drivers of IEP decisions.
“Our adoption of AI wasn’t because it was some new thing,” said Matthew Morris, executive director of special education for the district. The goal is to “move the needle and support our students with disabilities so that they can have as much success as possible.”
Nationally, a growing number of special education teachers say they use AI platforms to draft all or part of students’ IEPs. Fifty-seven percent of special education teachers who responded to a 2025 survey by the Center for Democracy and Technology said they used AI to help with IEPs or 504 plans during the 2024-25 school year, up from 39% in 2023-24.
Additionally, 31% use AI to identify trends in student progress for goal setting, 30% summarize IEPs and 504 plans, and 28% help choose accommodations.
The number of students who qualify for special education services is on the rise.
About 8 million students ages 3-21 qualified for special education services in 2024—up nearly 13% from 2019, according to U.S. Department of Education data analyzed by The Advocacy Institute, a nonprofit advocacy organization for people with disabilities.
Used well, AI can help educators refine generic IEP goals and personalize them, said James Basham, a professor of special education at the University of Kansas. Teachers can achieve that personalization without inputting personally identifiable information. Generic descriptions of the student’s skills, abilities, and challenges are more valuable than entering their name or race, factors that have been shown to increase the likelihood of bias in AI output.
“There’s much more useful information to use that can result in more meaningful results,” Basham said.
Mesquite educators use an AI tool that pulls data from various district platforms to help identify trends and areas of strength and weakness, rather than making specific IEP decisions, said Morris.
The AI can flag attendance trends or classroom subjects where the child is struggling, which can help teachers create specific IEP goals. However, it will not dictate what interventions a student might need, Morris said.
The tool was trained on an IEP template, so it can highlight areas a teacher may have forgotten to fill out, rather than making up information and filling it out itself, Morris said.
Over the past 11 months, teachers reported that using the AI tool has cut the time required to develop IEPs from up to three hours to about 45 minutes per student.
The time saved has helped ease burnout and given educators more time to spend with students, Morris said.
Mesquite district leaders worked with education software company Panorama Education to develop the tool. Several other education companies offer similar AI tools that serve the same purpose.
Amelia Vance, President of the Public Interest Privacy Center, a nonprofit focused on protecting students’ privacy in the digital age, said she has concerns about the Mesquite district’s tool’s ability to pull so much student information from different platforms into an AI system.
“Essentially, the more places that information sits, the more places it can be breached from,” Vance said.
Vance also raised concerns about AI bias and the technology’s ability to fixate on a piece of data or information and make assumptions about a student that may not be true, and that the AI might not disclose what it’s using as part of its rationale behind recommendations.
For example, she said, Vance once used AI to prioritize which bills to pay and on what timeline. Later, while reading through AI code—where she could essentially see how the technology was “thinking” and making suggestions—she saw the AI was assuming she was financially irresponsible or didn’t pay her bills due to the past interaction. The AI had never disclosed in their chats that it held those beliefs—the only reason she caught the bias was that she knew how to access and read the behind-the-scenes code, Vance said.
“There are just so many ways like that that things can go wrong that districts probably don’t even know to think about or look for, and that’s where I get concerned about these sorts of activities with kids’ sensitive information,” she said.
In an emailed statement to Education Week, Morris said that the district did not “send a separate notification to families when the platform was introduced” because the tool “supports existing administrative responsibilities and does not change the district’s special education services or IEP development requirements.”
“Mesquite ISD uses several technology tools to support teachers and staff in their daily responsibilities,” the statement said. “Like other district-approved digital platforms, Solara was subject to Mesquite ISD’s established review process, which evaluates compliance with applicable state and federal requirements, as well as district standards for student data privacy and security.”
Vance said that there aren’t necessarily best practices or standards for when or how districts should disclose their use of AI tools when developing students’ IEP goals. But she said she would err on the side of proactively disclosing its use, explaining why, and allowing families the opportunity to opt out.
She cautioned that not doing so could lead to families being concerned and pushing back more than they otherwise would have had they felt more involved in the adoption.
AI decreases time to develop IEPs, but doesn’t always increase quality
One 2025 study found no significant difference in the quality between AI-produced IEPs and those created fully by teachers without AI assistance. However, the AI model took significantly less time to create the IEP, even when factoring in teacher review and modification time.
That tradeoff is worth considering, said Olivia Coleman, an assistant professor of exceptional student education at the University of Central Florida and a co-principal investigator at the Center for Innovation, Design, and Digital Learning, who conducted the research.
“If, in the worst case scenario, they’re the same quality, but you’re saving all that time, that’s time it puts back toward the teacher being able to be face-to-face with students,” Coleman said.
Making the initial drafting less daunting and labor-intensive could change how special education teachers view the process, Coleman said. Instead of feeling like an IEP is an administrative burden, more teachers may feel empowered to perfect the document to ensure students receive necessary services.
Teachers should refrain from inputting information like a child’s birthdate, address, or school name into a generic AI tool, like ChatGPT, to protect personal privacy, Coleman said. However, details like disability type, their strengths, motivations, and skills struggles are all helpful for producing individualized goals.
Educators can also train AI tools by having a “conversation” with the tool while revising its output, she said. For example, a teacher can ask AI to modify a proposed goal to better fit a student’s needs, and explain why they want that revision.
Districts should provide guidance, training about responsible and ethical AI use
The rise of AI in special education comes as districts struggle to draft policies and create professional development to train teachers in appropriate use of the technology.
The EdWeek Research Center has conducted national surveys in recent years in which it has asked K-12 teachers how much professional development they’ve had on AI.
Schools are making strides: In less than three years, the proportion of teachers reporting they’ve received at least some training on AI has grown from a minority to a clear majority.
In October 2024, almost 60% of respondents said they had received no such training.
By winter 2025, that figure dropped to 42%, with 22% reporting multiple training sessions and 9% reporting ongoing training.
Providing clear training and guidance to educators is important, said Coleman, particularly for special education teachers working with vulnerable students.
Efforts by some states and districts to restrict AI use amid safety concerns may be well-intentioned, Coleman said. But the “technology is here” to stay. Districts would be better served by teaching educators how to use AI responsibly and ethically, she said.
That’s the same logic Mesquite schools has employed in its AI rollout, Morris said.
“I knew staff were using AI, and … my worry was that they’re using ChatGPT or some other open AI platform that is potentially releasing personally identifiable information,” he said. “For us, it was important to select and provide a platform that could ensure our students’ safety.”
Schools ‘cannot take away the human-first approach’
As district leaders establish privacy agreements with AI companies, they should explore if the information the technology collects is encrypted and how that encryption works, how and when data is deleted, and if their privacy policy explicitly mentions the Family Educational Rights and Privacy Act or the Individuals with Disabilities Education Act, Basham said.
The most critical piece of the AI puzzle, Basham said, isn’t about the technology at all.
Teachers cannot treat AI as the primary developer of an IEP. Instead, AI must be viewed as a partner whose work needs to be reviewed, revised, and refined, Basham said.
“An AI system can help do some great problem-solving and come up with some great ideas that maybe the team hasn’t worked through in the past, but the teachers who work with and know the student themselves are the most important part of this whole thing,” he said. “You cannot take away the human-first approach.”
