AI-Generated IEP Goals: A BCBA Review Checklist

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

AI-generated IEP goals can look polished before they are actually usable.

That is the trap.

The language may sound professional, the criteria may look measurable, and the goal may fit neatly into an IEP system. But if the goal is not observable, function-aligned, realistic to measure, and connected to present levels, it will not help the student or the team.

This checklist is for BCBAs and school-based behavior analysts who are using AI to draft IEP behavior goals and need a fast, defensible way to review the output.

For the larger strategy, see AI IEP Goals for Behavior Analysts and the AI for behavior analysts hub.

The BCBA AI Goal Review Checklist

Before using an AI-generated IEP goal, review it against these questions.

1. Can the team observe the behavior?

A goal should describe what the student will do, not what the student will feel, understand, appreciate, or demonstrate internally.

Watch for phrases like:

• Improve self-regulation
• Demonstrate better coping
• Understand expectations
• Show appropriate behavior
• Manage frustration

Those may point toward important skills, but they are not enough by themselves.

Ask: What would a staff member actually see or hear?

Better language includes actions:

• Request a break
• Request help
• Use a taught script
• Move to a designated calm area
• Begin the first step of the task
• Raise a hand before speaking
• Return to the group within a specified time

If the behavior cannot be observed, the goal needs revision.

2. Is the goal positively stated?

AI often writes goals around reducing problem behavior:

The student will reduce aggression to fewer than two incidents per week.

Reduction goals may have a place in progress monitoring, but IEP goals should usually teach the skill the student needs instead.

Ask: What replacement behavior is being taught?

If the goal only says what the student will stop doing, revise it to include what the student will do instead.

3. Does the replacement behavior match the function?

This is the most important behavior analytic review.

If the FBA suggests escape from difficult work, a replacement behavior might involve:

• Requesting help
• Requesting a break
• Requesting a modified task
• Starting with a smaller response requirement

If the FBA suggests adult attention, the replacement behavior might involve:

• Asking for attention appropriately
• Requesting feedback
• Using a check-in card
• Waiting for adult attention with a visual support

If the FBA suggests access to tangibles, the replacement behavior might involve:

• Requesting the item
• Waiting
• Accepting "not available"
• Choosing from available alternatives

AI can suggest reasonable-looking skills that do not match the student's function. The BCBA has to catch that.

4. Is the criterion realistic from baseline?

An AI-generated goal may jump too far.

Example:

The student will independently request a break in 90% of opportunities.

That may be fine if the student already does this in 70% of opportunities with occasional prompting. It may be unrealistic if the student has never used a break request without full physical prompting.

Check:

• What is the current baseline?
• What level of prompting is currently needed?
• How often does the replacement behavior occur now?
• Is the target criterion ambitious but achievable?
• Does the student need objectives or benchmarks?

A good goal should create stretch without pretending instruction already happened.

5. Can the team collect the data?

School data systems have to survive actual school days.

Ask:

• Who collects the data?
• How often?
• In which settings?
• During which routines?
• What counts as an opportunity?
• How will substitutes or new staff understand the system?

Avoid goals that require constant data collection unless the team truly has that capacity.

Often, a simpler system is stronger:

• Frequency during a defined routine
• Percent of opportunities during selected observation windows
• Duration during specific transitions
• Daily rating tied to clear anchors
• Work completion or latency during defined tasks

AI should not create a data system the team cannot maintain.

6. Is the goal written for the school context?

Generic AI tools may write clinical-sounding goals that do not fit the classroom.

A school-based goal should consider:

• Classroom routines
• Staffing patterns
• Peer presence
• Instructional demands
• Transition times
• General education settings
• Special education supports
• Feasible teacher data collection

If the goal requires a level of precision that only works in a clinic, revise it for school implementation.

7. Is the prompt de-identified?

If you used AI to draft the goal, review the input too.

Do not include:

• Student name
• Date of birth
• Student ID number
• School name
• Highly unique identifying details
• Full IEP text copied into an unsecured tool

Use de-identified summaries unless your district has approved the system for student data.

8. Does the goal connect to the BIP or intervention plan?

IEP goals should not float separately from the behavior plan.

If the BIP teaches break requests, tolerance responses, transition routines, or help-seeking, the IEP goal should reflect those same skills when appropriate.

This is where the FBA to BIP connection matters. Assessment, intervention, and IEP goals should tell the same story.

A Simple Review Table

Use this table when reviewing AI-generated goals.

Review Area	Pass Question	Red Flag
Observable	Can staff see or hear the behavior?	"Improve self-regulation" with no action
Positive	Does it teach what to do?	Reduction-only wording
Function	Does it match the FBA hypothesis?	Replacement behavior unrelated to function
Baseline	Is the criterion realistic?	90% independence from near-zero baseline
Measurement	Can the team collect the data?	Constant data collection no one can do
School fit	Does it work in classrooms?	Clinic-style goal with no school routine
Privacy	Was student information protected?	Identifiable data in unsecured AI

Example Review

AI draft:

The student will reduce disruptive behavior and demonstrate improved coping skills in 80% of opportunities.

BCBA review:

• Not observable enough
• "Disruptive behavior" is vague
• "Improved coping skills" is vague
• Replacement behavior is missing
• Function is missing
• Measurement is unclear

Revised:

During independent writing tasks, when presented with a task lasting at least 10 minutes, the student will request help or request a 3-minute break using a taught phrase or break card instead of leaving the assigned area, in 4 out of 5 opportunities across 3 consecutive weeks, as measured by teacher opportunity data.

This version is stronger because it names the routine, replacement behavior, problem behavior alternative, criterion, and measurement method.

Bottom Line

AI can save time, but the BCBA review is what makes the goal defensible.

Use AI for drafting. Use behavior analysis for decision-making.

When in doubt, ask five questions:

1. Can we observe it?
2. Can we measure it?
3. Does it match the function?
4. Does it fit the student's baseline?
5. Can the school team actually implement it?

If the answer is yes, the AI draft may be useful.

If the answer is no, it is just polished text.

Next Steps

Read the full guide to AI IEP goals for behavior analysts, explore AI for behavior analysts, or use the IEP Goal Writer to draft behavior goals with a more behavior-analytic frame.