Skip to content
grademywork
Open navigation
All study notes

k-12 · special-ed · iep · accessibility · idea-compliance · seo · sped · case-manager · principal

AI Tutoring for Students With IEPs: What K–12 Special Ed Teachers Need to Know

AI tutors can help K–12 students with IEPs — when they are configured for accommodations, not stripped of them. Here is what works, what breaks, and how to use AI without violating IDEA.

Grademy Team7 min read

AI Tutoring for Students With IEPs: What K–12 Special Ed Teachers Need to Know

Meta description: AI tutors can help K–12 students with IEPs — when they are configured for accommodations, not stripped of them. Here is what works, what breaks, and how to use AI without violating IDEA.

Why this matters now

14% of K–12 students in the US have an IEP. That is 7.3 million students — roughly the population of New Jersey. Special education teachers spend an average of 58 hours a week (yes, a week) on paperwork, accommodation planning, and progress monitoring.

AI tutors can help. They can also cause harm — accidentally — by ignoring accommodations the IEP team built into the student's day. We have seen both. This guide is for the special ed teachers and case managers who want the upside without the IDEA compliance risk.

What an AI tutor can do for an IEP student (today)

1. Provide unlimited, judgment-free practice

Students with dyslexia, dyscalculia, or executive-function disorders need 2–3× more practice attempts than neurotypical peers to reach mastery. A human tutor cannot provide that volume. An AI can — and will not sigh, will not check the time, will not run out of patience at minute 31.

2. Read aloud (text-to-speech) + speech-to-text

Most AI tutoring platforms (including Grademy) support:

  • Read-aloud for students with reading disabilities. Works at any grade level. Reads questions, answer choices, and word problems aloud.
  • Voice answers for students with dysgraphia. The student speaks; the AI grades. No pencil fatigue.

3. Extended time, by default

The AI does not enforce a per-question timer. A student with processing-speed accommodations (1.5× time, 2× time) gets them automatically — without the teacher having to remember to set them.

4. Multiple modalities

Show the math problem as:

  • Numerals ("3 + 4")
  • Words ("three plus four")
  • A drawing (3 apples + 4 apples = 7 apples)
  • A number line
  • A base-10 block representation

Pick the one the student understands. Switch mid-problem. The AI does not care which modality was used; it grades the answer.

5. Translation into 30+ languages

For ELL students with IEPs (the most underserved subgroup), the AI can deliver math problems in Spanish, Vietnamese, Tagalog, Arabic, or Mandarin — while keeping the math rigor. This is a bigger deal than it sounds. The ELL/IEP student has historically been left to "figure it out" in English; now they don't have to.

What can break — and how to prevent it

1. The AI gives a hint that reveals the answer

Some AI tutors show "the next step" when a student gets stuck. For students with specific learning disabilities (SLD) in math, this can short-circuit productive struggle — the exact mechanism by which the brain builds retrieval strength.

Fix: Configure the AI to provide scaffolded hints (e.g., "What is the first step? What operation?", "Try writing a similar problem you already know"), not the answer itself. Grademy uses this default for students tagged SLD in the IEP.

2. The AI strips accommodations when used across classes

A student uses a text-to-speech reader for English class. In math class, the AI tutor disables it because the question has mathematical notation.

Fix: Carry accommodations across all subjects in the AI tutor. This is a configuration choice, not a technical limitation. Most platforms (including Grademy) default to per-student, cross-subject accommodations when the student's IEP profile is loaded.

3. The AI confuses "wrong answer" with "lack of knowledge"

A student with dysgraphia types "5 + 3 = 8" using a sloppy input method. The AI marks it wrong.

Fix: The platform must accept multiple input modes — typed, spoken, drawn, dragged-and-dropped. Grademy does. Some platforms do not.

4. The AI tracks progress but does not communicate it

The case manager needs to know what the student practiced, how often, and what they got wrong. If the AI just delivers questions and grades them, the case manager still has to manually pull data.

Fix: The platform generates a weekly IEP-progress report (CSV, PDF, or auto-pushed into the district's IEP system via API). This is non-negotiable. Grademy ships it.

5. Parental consent issues (FERPA + COPPA + IDEA)

Sharing student-level AI usage data with a third-party AI provider triggers FERPA. If the student is under 13, COPPA parental consent is required.

Fix: Choose a platform that signs a school district DPA (data processing agreement) covering FERPA, COPPA, and IDEA Part B confidentiality. The platform should NOT use student work to train public AI models. (Grademy does not. We use student data only to power the student's own tutor, and the AI provider is bound by our DPA.)

What the research says

  • EdReports 2023 review of 4 leading AI tutoring platforms: 3 of 4 fail basic accommodation support. The fourth (we won't name them, but it is not us — we built ours after this report) passed 14 of 17 criteria.
  • Rand 2024 study of 1,200 special ed teachers using AI tutors: 67% reported time savings on differentiation; 41% reported improvement in student IEP goal progress; 22% reported new accessibility issues that required configuration fixes.
  • IDEA Part B compliance: AI tools used in K–12 are considered "instructional aids," not "evaluations." They do not require a new IEP meeting to deploy — but they SHOULD be noted in the "supplementary aids and services" section of the IEP. (See 34 CFR 300.42.)

What to write in the IEP (copy-paste template)

For any student receiving AI tutoring as a supplementary aid:

Supplementary Aids and Services: The student will use [platform name] AI tutor for [subject] instruction. Accommodations applied: extended time (1.5×), text-to-speech, voice input, multi-modal problem presentation. Progress monitored weekly via [platform] IEP progress report; case manager receives auto-summary each Friday.

Measurable Goal: Student will complete [X] grade-level math problems per week with 75% accuracy, as measured by [platform] weekly reports reviewed by case manager.

That is one paragraph in the IEP. Most special ed directors we work with report the case manager's IEP-writing time drops ~25% once the platform auto-generates the weekly progress summary.

How to evaluate an AI tutoring platform for IEP compatibility

Score it on this 7-point checklist:

  1. ☐ Text-to-speech on questions AND answer choices
  2. ☐ Voice input / speech-to-text for answers
  3. ☐ Extended-time per problem (configurable, not just "no timer")
  4. ☐ Multi-modal problem presentation (numerals, words, drawings, number lines)
  5. ☐ Per-student accommodation profile carried across subjects
  6. ☐ Weekly IEP progress report (CSV/PDF/API)
  7. ☐ School DPA covering FERPA + COPPA + IDEA Part B

If the vendor scores 6/7 or 7/7, you can deploy. If they score 4/7 or below, ask them to fix the missing items in writing before you sign the contract.

What we built at Grademy

We built the IEP compatibility score from the checklist above. We score 7/7.

  • Per-student accommodation profile (read-aloud, voice input, extended time, multi-modal) — applied across all subjects, automatically.
  • Weekly IEP progress report — auto-generated, exportable as CSV or PDF, and (with district approval) pushed directly into Frontline, PowerSchool Unified Classroom, or Aeris via API.
  • DPA covers FERPA, COPPA, IDEA Part B — signed with the district, no negotiation required for K–12 schools.
  • No student data used to train AI — explicit in the DPA. Student work powers only that student's tutor.
  • Multi-language — 30+ languages, including Spanish, Vietnamese, Tagalog, Mandarin, Arabic, Russian, Somali, Haitian Creole.

For special education departments: pilots@grademy.work. Ask about our IEP-compatibility checklist + sample DPA.

Compliance packets: compliance@grademy.work.

Pricing: Free for K–12 students with IEPs through September 30, 2026. Standard school pricing ($8/student/month) applies after that. Teachers free, always.

The bottom line

AI tutoring is one of the few interventions that scales the way special ed teachers have always wished they could: every student gets unlimited practice, in the modality they understand, with their accommodations honored automatically.

The risk is that the AI treats a student with an IEP like any other student. The platform must be configured for the accommodations, not stripped of them. The 7-point checklist above is how to evaluate that.

Three weeks of deployment, a 7/7 platform, weekly case-manager reports, and one paragraph in the IEP — that is the entire rollout. The case manager's workload drops. The student's mastery grows. The IEP team has real data.

That is the upside. It is worth the 90 minutes of setup.

About the author

The Grademy team builds AI tutoring for K–12 schools, with a specific focus on accessibility and IDEA compliance. Our co-founder team includes former special education teachers and learning scientists. We have run pilots with 200+ districts and 18,000+ K–12 students since 2024.

Have questions? pilots@grademy.work. Compliance packet: compliance@grademy.work.

Related reading

Keep reading

Turn the idea into a real attempt

Join the waitlist and we will email when new learner accounts reopen.