Using AI for IEP writing: legal risks a case manager should know

It is 11:45 PM on a Tuesday in November 2026, and I am staring at three open tabs of psychoeducational evaluations, a cold cup of coffee, and a blinking cursor on an IEP goal page that has been empty for twenty minutes. Every special education case manager knows this exact brand of fatigue. When you are juggling twenty-eight active caseloads, four manifestation determinations, and a stack of re-evaluations, the temptation to paste a student's profile into an AI tool and let it draft your Present Levels of Academic Achievement and Functional Performance is nearly overwhelming. Do not do it. Using AI for IEP writing: legal risks a case manager should know is not a theoretical webinar topic; it is about keeping your license and keeping your district out of a due process hearing.

The FERPA Trap in General-Purpose Tools

The single biggest mistake special educators make is treating consumer-grade AI models like a secure vault. Tools like Writesonic and Claude are brilliant for administrative drafting, long-context document review, and general planning. Writesonic offers plenty of templates at roughly $16 per user per month, and Claude handles long-context planning exceptionally well for around $20 per user per month, but neither is built for student data. If you paste a student’s diagnostic scores, medical diagnoses, and full name into a consumer-tier AI prompt box, you have likely violated the Family Educational Rights and Privacy Act before your morning prep period even starts.

Even when paid tiers offer data privacy controls, consumer platforms do not automatically align with your district’s specific data processing agreements. When you need general assistance without student identifiers, you might look at tools that make the best free option lists like Quizizz, or browser-based helpers like Brisk Teaching at roughly $10 per user per month, which brings AI inline into Google Docs and Slides. Even so, Brisk reads document content directly, meaning any unredacted IEP draft sitting in that tab is within reach of the extension's processing scope. If your district has not cleared the tool through IT and signed a DPA, you are taking a massive compliance gamble.

What You Can Automate Without Breaking the Law

You do not need to feed student names or protected health information into an algorithm to get value out of these platforms. The trick to safe AI use in special education is strict anonymization. If you want to build instructional materials, adapt texts, or generate accommodation ideas, strip every single identifier out of the prompt. For instance, if you are looking for ways to present information at three different reading levels, take a public domain source text and use an assistant to generate supported, core, and extension versions. You must still check the vocabulary list and comprehension questions by hand, because invented reading comprehension prompts are the most common failure point and the easiest to catch.

Prompt:
"Act as a special education curriculum specialist. Based on the following anonymized fifth-grade passage about ecosystems, generate three reading levels: one at a third-grade accessibility level, one at grade level, and one extension version. Provide a five-word vocabulary list for each."

First-pass output:
The model generated three clear text tiers with appropriate vocabulary, but included comprehension questions that referenced fictional characters not present in the original source text.

Edits that made it usable:
1. Stripped out the hallucinated comprehension questions entirely, keeping only the leveled texts and vocabulary lists.
2. Cross-referenced the leveled vocabulary against the students' actual tier-two intervention lists to ensure alignment.

When building quizzes or review games, platforms like Kahoot! at roughly $10 per user per month and Quizizz make it easy to generate practice questions for mixed-ability classrooms. You can safely paste a state standard into Quizizz without exposing any child's IEP data, ensuring your review sessions target the exact skills your students need to practice. For a broader look at how classrooms are adapting these workflows, check out our guide on the best AI tools for teachers.

The Danger of AI-Generated IEP Goals and Baselines

Never let an AI write a Measurable Annual Goal from scratch using real student data. A valid IEP goal requires a precise baseline, specific conditions, a clear behavior, and measurable criteria. When general AI writing tools attempt this, they often produce vague language that sounds professional on the surface but fails the legal test of measurability. A goal that states "The student will improve reading comprehension with 80% accuracy" is useless if the baseline data is fabricated or generalized by an algorithm trying to fill in blanks.

If you use a tool like Claude to help structure a complex accommodation list, you must feed it entirely scrubbed scenarios. Better yet, look to purpose-built educational ecosystems like Khanmigo at roughly $4 per learner per month, which offers nonprofit trust signals and education-first design, though it is still primarily built for student tutoring and teacher scaffolding rather than admin paperwork. If you are exploring targeted interventions for specific plans, you can also consult specialized resources such as our breakdown on AI 504 plans for special education.

Maintaining Human Oversight on Every Document

Every word in a finalized IEP is your professional responsibility. If a due process hearing officer asks why a specific accommodation or modification was selected, you cannot point to a software log and say the machine suggested it. Use AI as a sounding board for phrasing, a grammar checker for rough parent communication emails, or a generator for generic graphic organizers. Never use it to synthesize evaluation summaries, determine placement, or write present levels. Keep your student records in your secure district SIS, and keep the AI strictly in the realm of anonymous brainstorming.

FAQ

Is AI allowed for lesson planning in my school?

It depends on your district's approved list and what the approval covers. Ask IT which tools are cleared and who owns the account before a document with a legal consequence is involved.

Who is responsible if something goes wrong?

The school, and in practice the named teacher who reviewed the output. That is why a dated note about what was generated and what a human changed matters.

What should never go into a prompt?

Names, rosters, scans of student work and gradebook screenshots. Topic-level prompts do most of the work without any of that.

Sources

• Writesonic Pricing and Features: https://writesonic.com/pricing
• Kahoot! Educator Plans: https://kahoot.com/schools-u/pricing/
• Quizizz Educator Tiers: https://quizizz.com/home/pricing
• Brisk Teaching Extension Scope: https://briskteaching.com/
• Khanmigo for Teachers: https://www.khanacademy.org/khanmigo
• Anthropic Claude Privacy and Terms: https://support.anthropic.com/