FERPA-Safe AI IEP Goals: What School BCBAs Can Use

AI-assisted draft; reviewed and edited by Rob Spain, BCBA.

School BCBAs are under pressure to write better documentation faster.

AI can help with that. It can draft IEP behavior goals, rewrite vague language, generate replacement behavior options, and help teams think through measurement.

But IEPs involve student records. That means AI use has to be handled carefully.

This article gives school-based BCBAs a practical way to think about FERPA-safe AI use when drafting IEP behavior goals.

For the broader strategy, see AI for behavior analysts. For goal writing support, see the IEP Goal Writer.

The Core Rule

If an AI tool is not approved for student data, do not enter identifiable student information.

That means no:

• Student names
• Student ID numbers
• Dates of birth
• School names when paired with unique details
• Full IEP documents
• Full evaluation reports
• Discipline records with identifying information
• Highly unique descriptions that could identify the student

Use de-identified summaries instead.

What Counts as De-Identified for AI Goal Drafting?

A de-identified AI prompt removes the student's identity while preserving the behavior analytic information needed to draft a useful goal.

Instead of:

Write an IEP goal for Marcus at Lincoln Elementary. He hits peers during recess and has an autism eligibility.

Use:

Draft three measurable IEP behavior goal options for an elementary student who hits or pushes peers during recess when adult attention is low and peer conflict occurs. The FBA hypothesis is that behavior is maintained by adult attention and escape from peer interaction. Baseline: the student uses a taught help request in 1 out of 5 peer conflict opportunities with adult prompts.

The second version gives AI enough information to help without giving it the student's identity.

What Information Is Usually Safe to Include?

When de-identified, these details are often useful for drafting:

• General grade band, such as elementary, middle school, or high school
• Target behavior in observable terms
• Setting or routine, such as recess, transitions, or independent work
• FBA hypothesis
• Replacement behavior
• Baseline level
• Prompt level
• Measurement method
• Desired criterion

Keep it general enough that the student cannot be identified, but specific enough that the goal is not generic.

What Information Should Stay Out?

Avoid putting sensitive records into unsecured AI systems.

That includes:

• Full psychoeducational evaluations
• Full IEP documents
• Medical information
• Names of staff or classmates
• Incident reports with dates and locations
• Parent names or contact information
• Unique family circumstances
• Discipline histories copied directly from student records

If the information is not needed to draft the goal, leave it out.

A FERPA-Conscious AI Workflow for IEP Goals

Use this workflow when drafting IEP behavior goals with AI.

Step 1: Do the behavior analytic work first

AI should not replace the FBA, team review, or present-level analysis.

Before drafting, identify:

• Target behavior
• Baseline
• Contexts where behavior occurs
• Hypothesized function
• Replacement behavior
• Supports already in place
• Data collection method

Step 2: Create a de-identified prompt

Strip the prompt down to behavior analytic variables.

Good prompt structure:

Draft [number] measurable IEP behavior goal options for a [grade band] student. Target behavior is [observable behavior]. The behavior occurs during [routine]. The FBA hypothesis is [function]. Current baseline is [de-identified baseline]. Replacement behavior should be [skill]. Make the goals observable, measurable, positively stated, and feasible for school data collection.

Step 3: Review the output like a BCBA

Do not accept the first draft because it sounds polished.

Review for:

• Observability
• Positive replacement behavior
• Function alignment
• Baseline fit
• Measurement feasibility
• School implementation
• Team understanding

For a full checklist, see How BCBAs Should Review AI-Generated IEP Goals.

Step 4: Revise before team use

The AI output is a draft. The final goal should reflect the IEP team's decisions and the student's actual needs.

Step 5: Document responsibly

If your district has AI guidance, follow it. If your district does not, be conservative:

• Keep AI prompts de-identified
• Do not store student data in general AI tools
• Do not paste full records into AI tools
• Keep final recommendations tied to actual data
• Make human review obvious in your workflow

Examples of Unsafe and Safer Prompts

Unsafe prompt

Write an IEP goal for a 4th grade student named [Name] at [School]. He has 17 office referrals this year, takes medication for ADHD, and his parents are divorced. He runs out of Ms. [Name]'s classroom during math.

This includes too much identifying and sensitive information.

Safer prompt

Draft three measurable IEP behavior goal options for an upper elementary student who leaves the assigned classroom area during math tasks. The FBA hypothesis is escape from difficult independent work. Baseline: during math, the student leaves the assigned area in 4 out of 5 observed work periods and currently requests help only with adult prompting. Replacement behavior should involve requesting help, requesting a short break, or beginning the first step of the task with support.

The safer prompt keeps the information needed for goal drafting while removing identifying details.

AI Does Not Remove Professional Responsibility

Even with privacy protections, AI use does not transfer responsibility away from the BCBA or IEP team.

The final goal still needs to be:

• Based on student data
• Reviewed by qualified professionals
• Understood by the team
• Appropriate for the student's educational program
• Implementable in the school setting

AI can assist with language. It cannot take responsibility for the recommendation.

Where Behavior School Fits

Behavior School focuses on AI-supported tools for behavior analysts, not generic education outputs.

That means we care about:

• Function-based thinking
• School implementation
• Measurable goals
• Privacy-conscious workflows
• Human review
• Practical behavior analytic language

You can explore the main AI hub at AI for behavior analysts, draft behavior goals with the IEP Goal Writer, or review examples in our IEP behavior goals guide.

Bottom Line

AI can be useful for IEP behavior goals, but school BCBAs need a privacy-first workflow.

Do the analysis first. De-identify the prompt. Review the output. Revise for the student. Keep the human decision-making visible.

That is the difference between using AI as a professional tool and using it as a shortcut.

Frequently Asked Questions

Can BCBAs use AI to write IEP goals under FERPA?

AI can be used to support drafting when student privacy is protected and district policies are followed. Do not enter identifiable student information into unsecured AI tools.

What should I remove before using AI for IEP goal drafting?

Remove names, dates of birth, student ID numbers, school names, full documents, medical details, unique family details, and any information that could identify the student.

Is a de-identified prompt still useful?

Yes. A good de-identified prompt can include target behavior, setting, function, baseline, replacement behavior, and measurement needs without naming or identifying the student.