Staying Human While Using AI for IEPs

After a little coaching and a handful of exploratory prompts, special education teacher Casey Celeste was convinced: generative AI, if used carefully, would lighten her workload.

On top of spending a full school day with her students, she faces a mountain of paperwork, from drafting IEPs and updating student data spreadsheets to writing progress reports and documenting accommodations. After being introduced to AI, however, she’s been able to chip away at much of the work that consumed her time outside of school. 

“If I didn’t use AI, I would not have weekends,” says Celeste.

For years, that work has been an unavoidable part of special education. But Celeste now joins a growing number of teachers experimenting with AI to ease administrative tasks, giving her more time to interact directly with her students. After feeding ChatGPT detailed observations about a student like their reading level, recent assessment data, and personal notes like their interests or frustrations—being careful not to reveal their name or other identifying information—she asks the powerful software to help organize her notes, draft IEP goals, and brainstorm ways to measure progress. 

“Without the AI tools, it could take anywhere from four to six hours if you want to write a solid IEP, which you don’t have time for,” Celeste says, noting that using AI has cut that time by more than half. “I did not have planning or lunch [time] in the last three years, that’s on your own time.”

New research suggests that teachers may have a good reason to keep experimenting with AI. In a 2025 study, experienced K–12 special education teachers wrote IEP goals based on a student’s profile and then used a chatbot to draft goals from the same information. After comparing both sets, researchers concluded that “generative AI tools like OpenAI’s ChatGPT can reduce workload burden for special education teachers without compromising the quality of IEP goals.”

It’s a potential game-changer, says Danielle Waterfield, a University of Virginia researcher, the lead author of the study, and a former special education teacher. If AI can reliably handle some of the drafting and documentation, teachers can spend less time behind a computer and more time working directly with students, she argues.

Unprepared

“We don’t do a great amount of training in teacher preparation programs around writing and implementing IEPs,” Waterfield says. As a result, teachers typically learn on the job, piecing together what a useful IEP looks like from colleagues, past IEPs, and online resources. Even experienced teachers can inherit plans that vary wildly for each student, Waterfield notes.

Writing a high-quality goal isn’t easy. In order to be legally compliant, they must be measurable, clearly identify a student’s current level of performance, and specific enough to guide instruction. “It can be a really convoluted thing sometimes, and folks need support around that,” Waterfield says. Compounding the problem of inadequate training is poor support once teachers are in the classroom. While there’s federal guidance on what an IEP must accomplish, the details are largely left to teachers.

In a 2024 study, researchers found that during pre-service training, “over half of special educators indicated not feeling prepared to independently write an IEP after completing their program.” Celeste’s experiences confirm that finding: Although she took a college-level course on writing IEPs, it covered the basic requirements without taking a “deep dive” into how to write an effective plan. It wasn’t until her third year of teaching that she found her footing. 

It’s experiences like Celeste’s that drove Waterfield to pursue a doctorate at the University of Virginia, where she began to think about the ways technology could shoulder some of the work for special education teachers. Shortly after the pandemic, she led a team of researchers to see how AI would perform, asking experienced K–12 special education teachers to write IEP goals based on a student’s profile, taking into account their disability, past performance, and areas of need. The same teachers also used ChatGPT to draft parallel IEP goals using the same information.

After analyzing both sets of goals on six dimensions including clarity, measurability, and timeliness, Waterfield and her team found “no statistically significant difference in their quality ratings.” When surveyed, the special education teachers who participated had an even more favorable impression of AI, with most agreeing that the AI-generated goals were “the same or better quality” than a special education teacher would have written.

Time Saved for What Matters Most

Perhaps the biggest impact of AI, however, isn’t the paperwork it can complete, but the time it gives back to teachers and students.

“Because I know that AI can help me with the paperwork, I can focus on the most important part of my job, which is building those relationships with students and forming connections with them,” says Celeste, who was paired with Waterfield during her first year of teaching as part of a coaching program. After spending a full day with her students—all of whom have special needs—she then spends hours after school planning lessons, updating IEPs, and communicating with families and other teachers. 

The extra hours aren’t trivial. According to a 2022 EdWeek survey, teachers only spend 46 percent of their time teaching, with the remainder dedicated to tasks like grading papers, doing administrative work, and planning lessons. That comes at a cost, with a 2024 OECD analysis finding that more than half—52 percent—report “excessive administrative work as a source of work-related stress.”

Special education teachers, in particular, have been hit with extreme shortages, deep cuts in federal funding, and increased caseloads, adding “an additional 15-20 unpaid hours each week to meet mandatory paperwork deadlines,” recent data shows. For Celeste, those extra hours bleed into nights and weekends, cutting into time she could be spending getting to know her students better and figuring out ways to accommodate their needs in large classrooms.

“The more a student with a disability is face-to-face with their teacher, the better the educational outcomes are for them,” Waterfield says. “That is really the driving force behind a lot of this.”

Bridging the Gap

As a former special education teacher, Waterfield also saw firsthand how time-consuming it was to align completed IEPs with classroom instruction. On paper, the plans were designed to give students the support they needed, but in a busy classroom, the requirements often clashed with reality.

“I have a class of 30 students, and 20 of them have the preferential seating accommodation. How do I even physically do that in my classroom?” a colleague once asked her—with similar problems forcing general and special education teachers to improvise solutions. The challenge wasn't simply following an IEP, Waterfield realized, but figuring out how to meet the complex needs of each student. A single lesson plan may need dozens of tweaks to account for various sensory needs, ability levels, and behavioral supports, turning an ordinary planning session into hours of additional work.

In a 2020 study, researchers confirmed that special education teachers were frustrated “with the lack of differentiated materials available to them,” scrambling to piece together resources and adapt them on the fly. There’s a persistent gap between “top-down regulations and requirements” and classroom realities, the researchers point out, leaving teachers to figure out how to translate IEPs into daily practice.

That’s an issue Celeste is quick to point out. “I had to find different ways to adapt because a lot of my kids were nonverbal and so I had to figure out appropriate goals for them,” she says. “How do you measure whether a child knows their letters, for example, when they’re nonverbal?” Classroom materials often assume that students can speak, write, or demonstrate their knowledge in conventional ways—assumptions that don’t always fit the students she teaches. 

On top of drafting IEP goals, Celeste now turns to AI for fresh ideas on implementing those goals in the classroom. “I’ll ask ChatGPT, ‘This student is nonverbal and can do A, B, and C, but I’m not sure how to measure what they know,’” she explains. “It’ll give me different ideas on what to do, different activities to try with them.”

A Healthy Dose of Caution

As useful as AI is, Waterfield cautions against using it to generate finished products. “It really does take a lot of time to think through what’s most appropriate for students,” she says. Observing students, tracking their progress, and drafting documents for families and other teachers are all crucial activities that can be assisted by, but not outsourced to, a chatbot. Instead, think of it as an “iterative process,” a way of translating a teacher’s notes about a student into a format that’s structured, measurable, and compliant with a school’s IEP policy. Teachers still need to interrogate the output, “going back and forth and making sure that it really is individualized, high quality, and something that they would be able to implement,” Waterfield explains. 

Celeste echoes that approach. Rather than asking AI to invent an IEP goal from scratch, she starts by feeding it everything she knows about a student. “What I’ll do is brain dump every single thing I know about this child without using their name,” she says, before prompting the chatbot to develop an IEP goal based on the information. More often than not, the end result will sound “professional and perfectly worded,” but the underlying knowledge—the student’s strengths, weaknesses, behaviors, and progress—still comes from Celeste. AI’s role is primarily to help her organize and articulate that information, turning a messy collection of observations into the formal language that an IEP requires.

“AI is a huge help with all of your paperwork and your lesson planning, but the human connection between you and the student is the most important part of your job,” Celeste says.