SchoolAI, Flint and MagicSchool each list a behavior intervention plan (BIP) generator, and ChatGPT for Teachers and Gemini in Google Workspace for Education will draft one if you supply the structure. None of them can do the part that comes first. A BIP is built on a functional behavioral assessment (FBA), the observation data that shows why a behavior happens, and a tool only sees what you type. Use AI to turn de-identified notes into a first draft; the team decides the function and signs the plan.
This is written for special education teachers, behavior specialists and school psychologists who draft plans, and for the general education teachers who contribute the classroom data. Gallup's survey of 1,952 U.S. K-12 educators, published October 14, 2026, found that 59 percent used AI tools in the 2025-26 school year, and that required paperwork is the work educators least feel matters: only 9 percent said completing administrative or required paperwork makes them feel they matter "a great deal," while 38 percent said "not at all" (Gallup). Our view is that a behavior plan sits in the middle of that finding. It is required paperwork, and it also decides what happens to a child on Monday morning, so the drafting is worth handing over and the data and the decision are not.
What should AI draft in a behavior intervention plan, and what should it never touch?
The IRIS Center at Vanderbilt, the federally funded teacher-training project, lists six components of a BIP and describes the plan as one built on a thorough FBA (IRIS). The table below sorts those six by who has to supply them. The right-hand column is our editorial assessment, not a vendor claim.
| BIP component (per IRIS) | Who supplies it | Where AI can help |
|---|---|---|
| Description of the target behavior | You, from direct observation | Tighten it into observable, countable wording |
| Summary of the data collected during the FBA | You and the team: observation records, interviews, records review | Organize your de-identified notes into a summary; never collect the data |
| Hypothesized function of the behavior | The team | Suggest competing hypotheses to test; label every one a hypothesis |
| Replacement behavior to teach and reinforce | The team | List options at the right level for the student's age and skills |
| Specific interventions | The team | Draft prevention, teaching and response strategies for you to cut down |
| Progress monitoring plan | The team | Draft the data sheet and the review schedule |
Read down the middle column. Five of the six components start with a person who watched the student, and the sixth starts with a team that read what those people wrote. A generator that can open with a blank form and produce a confident plan is producing a plan from nothing, however well it is formatted.
How do the six tools compare?
Prices and plans were checked on October 21, 2026. Statements in the privacy column are what each provider says about itself, not independent findings.
| Tool | Best use | Strengths | Limitations | Price / free tier | Privacy considerations |
|---|---|---|---|---|---|
| SchoolAI (4.92 from 13 directory reviews) | A BIP draft you can edit and export | Tools > BIP > Generate; optional document upload; editable draft; exports as PDF or Google Docs | Help page names no input fields and says nothing about plans or storage | Free Trial includes Tools and Assistants; Pro and Scale have no public prices | Student-facing platform; the BIP help page is silent on how uploads are handled |
| Flint (not in the directory) | A plan organized around function | Outputs target behaviors, prevention, replacement skills and responses; treats function as a hypothesis | No input fields or BIP pricing listed | Free for teachers; school plans from $3,000 a year | Says student data is never used to train models; a first name or initials will do |
| MagicSchool (4.74 from 19) | Teachers already working in it | 80-plus teacher tools on the free plan | BIP tool details did not load: unverified | Free; Plus $8.33 a user a month billed annually; Enterprise on quote | Says student data is never sold or used to train models; custom data agreement is Enterprise only |
| ChatGPT for Teachers (4.35 from 23) | Fitting the plan to your district's own form | You set the structure; works on any template | Can invent a function if you leave a gap (our view) | Free through June 2028 for verified U.S. K-12 staff | FERPA-aligned protections and a student data privacy agreement (OpenAI's words); not used for training by default |
| Gemini in Workspace for Education (4.59 from 41) | Districts already on Google | Admin assigns the licenses and controls access | Needs an admin-assigned license; price not found: unverified | No public price found in the FAQ read | For that license, submissions are not used to train models and not reviewed by humans (Google's FAQ) |
| Glint (no directory reviews yet) | Teaching the replacement behavior | Social stories, schedules, communication boards | Does not write the plan | $18 a month or $150 a year | Says "FERPA & COPPA aligned," with no further detail on the page |
One finding from the comparison is worth stating on its own. The three tools that advertise a BIP generator do not say which pricing plan includes it. SchoolAI's pricing page does not mention its BIP tool on any plan (pricing), Flint's BIP page gives no price for the generator even though its pricing page says Flint is free for teachers (BIP page, pricing), and MagicSchool's tool page did not load in this run. Open each one with a free account and look before you plan a workflow around it.
Which tool should you open first? Four questions
Do you have the FBA data? If not, the next job is collecting it, and no tool here does that. The U.S. Office of Special Education Programs (OSEP) says parental consent is required when an FBA is conducted as part of an initial evaluation or a reevaluation (OSEP letter, April 2013), so confirm with your special education director which category your FBA falls in before you start recording.
Can you describe the student without identifying them? A description such as "a third grader who leaves her seat during independent math" needs no name, ID number or school. Flint's page says its tool does not need a full name or ID, and that a first name or initials will do (Flint). Start de-identified in every tool, whatever its settings say.
Which account are you signed into? Under FERPA, a vendor can receive education records as a "school official" only if it performs a service the district would otherwise use employees for, meets the district's criteria for a legitimate educational interest, is under the district's direct control over how records are used and maintained, and uses records only for authorized purposes (Department of Education Vendor FAQ). A personal account you opened with a work email does not meet the direct-control condition. A district-approved account might.
Does the draft state a function as fact? If it does, delete the sentence and write it back as a hypothesis for the team to test. The FBA decides the function, and the plan follows it.
Whatever your answers, the special education and accessibility tools in the directory list pricing and data handling side by side if you want to look beyond these six.
SchoolAI is the best fit for a first draft you can edit and export, because the BIP tool opens straight to editable text
SchoolAI's help center says the tool is reached through Tools > BIP > Generate, takes "the prompts required" plus optional supporting documents, and produces a draft in a text editor that you can revise and then share as a link, export as a PDF, or send to Google Docs or Google Classroom. The article says the draft "should be used as a first draft" for you to review and customize for the specific student (help center). The directory gives SchoolAI 4.92 from 13 reviews.
Watch out for: the article names no input fields and does not describe how uploaded documents are stored or used, so the safest input is a typed, de-identified summary rather than a file from a student's record.
Pricing: SchoolAI's pricing page lists Free Trial, Pro and Scale with no dollar amounts, and does not mention the BIP tool on any plan (pricing). Its help center says the free plan, as of July 10, 2026, includes Tools and Assistants "with no restrictions" and five Space sessions a year (free plan). Whether the BIP generator sits on the free plan is unverified. Pro and Scale: no public pricing. Checked October 21, 2026.
Safety: SchoolAI is mainly a student-facing platform, and the directory lists a vendor-declared FERPA status and the student data types it collects. The BIP article says nothing about data handling. Ask whether your district has an agreement with SchoolAI before any detail from a student's file goes near it.
Flint is the best fit for a plan organized around function, because it treats function as a hypothesis
Flint's BIP generator page says it "turns what you've seen into target behaviors, prevention strategies, replacement skills and responses," and that it treats the behavior's function as a hypothesis unless you have assessment data (Flint). That is the habit this whole post argues for, written into the tool. Of the BIP pages I could read in this run, it is the only one that says so.
Watch out for: the page lists no input fields and no price for the BIP generator specifically. Flint is not in the aieducator.tools directory, so there is no rating, review count or security assessment here.
Pricing: Flint's pricing page says "Free for teachers, forever," with a free plan for up to 80 users, then school plans from $3,000 a year for up to 150 users, $4,000 for up to 250, and $6,500 for up to 500 (pricing). Whether the free plan covers the BIP generator is unverified. Checked October 21, 2026.
Safety: Flint states that student data is never used to train AI models, is not shared with third parties, and is kept separate for each school (Flint). Those are the vendor's statements, and a district agreement is still the document that matters.
MagicSchool is the best fit for teachers already working in it, because its free plan carries 80-plus teacher tools
MagicSchool's pricing page lists a Free plan with "80+ teacher tools," a Plus plan at $8.33 a user a month billed annually or $12.99 billed monthly, and an Enterprise plan on quote that adds a "Custom Data Privacy Agreement" (pricing). Searching the site surfaces tool pages titled "BIP suggestion generator" and "behavior intervention suggestion generator," but those pages did not load their tool details in this run, so what they ask for, what they return and which plan they sit on are unverified. The directory gives MagicSchool 4.74 from 19 reviews.
Watch out for: the custom data agreement is an Enterprise feature, so on a free or Plus account you are accepting standard terms on your own authority.
Pricing: Free $0; Plus $8.33 a user a month billed annually, $12.99 billed monthly; Enterprise on quote (pricing). Checked October 21, 2026.
Safety: MagicSchool's site says student data is "never sold or used to train AI models" (MagicSchool). Treat that as the provider's claim and check it against your district's agreement.
ChatGPT for Teachers is the best fit for fitting the plan to your district's own form, because you control the structure
A general assistant has no behavior-plan template, which is its strength when your district has a required form. Paste the form's headings and your de-identified notes and ask for a draft under each heading. OpenAI's help center says ChatGPT for Teachers is for staff at accredited U.S. K-12 schools and districts, is not for students, and is free through June 2028 (help center). The directory gives ChatGPT 4.35 from 23 reviews.
Watch out for: an assistant that has no data will invent a function to fill the gap. The prompt below tells it not to.
Pricing: free for eligible staff through June 2028, after which OpenAI says pricing may change (help center). Checked October 21, 2026.
Safety: OpenAI describes "FERPA-aligned protections" and "a student data privacy agreement," says shared information is "not used to train our models by default," and says content may still be classified by automated systems for safety and quality (help center). The article does not address entering student personal information, so ask your district whether it has accepted that agreement before you do.
Gemini in Workspace for Education is the best fit for Google districts, because the admin controls the license
Google's education FAQ says that for users with a Google AI Pro for Education license, "submissions aren't used to train models and are never reviewed by humans." It also says the add-on is for users 18 and older and that admins assign the licenses (Google FAQ). That makes it a staff tool, which suits drafting a plan. The directory gives Gemini 4.59 from 41 reviews.
Watch out for: you cannot sign yourself up. If your admin has not assigned the license, a personal Gemini account is a different product with different terms.
Pricing: the FAQ read for this post gives no price for the license: no public pricing found, unverified. Checked October 21, 2026.
Safety: the no-training and no-human-review statement applies to the licensed education product, as Google describes it. Confirm with your admin which version your account is.
Glint is the best fit for teaching the replacement behavior, because it makes the visual supports a plan calls for
A BIP names a replacement behavior to teach, such as asking for a break with a card. Glint is a visual-support tool, not a plan writer, and it covers that next step. A product directory describes it as helping "educators and therapists create personalized visual supports in seconds," including social stories, schedules, communication boards and task analyses (Closing The Gap). It has no directory reviews yet.
Watch out for: the plan decides what is taught, and the tool only produces the picture. A support built before the team agrees on the replacement behavior is guesswork in color.
Pricing: $18 a month or $150 a year, with Enterprise for schools and organizations on quote; the page offers a free start and free shared boards for parents but does not state trial length or limits (pricing). Checked October 21, 2026.
Safety: the pricing page says "FERPA & COPPA aligned" and gives no certifications or data-handling terms. The directory lists a minimum user age of 18, so the account is yours, not a student's. Build supports from the plan's description, not from the student's file.
What does a safe prompt look like? A worked example for a third grader
The example below is invented to show the shape of an input. It is not drawn from a real student, and the data in it is illustrative only.
A third-grade teacher has ten 20-minute independent math periods of ABC notes (antecedent, behavior, consequence). In most of them, the behavior is leaving the assigned seat for more than 30 seconds after a worksheet of ten or more items is handed out, and the consequence is a peer talking back or a teacher redirecting. The team has not yet settled on a function. Here is a prompt that keeps the tool in its lane:
I'm a special education teacher drafting a behavior intervention plan for an IEP team. Student: a third grader, no name. Target behavior: leaves the assigned seat for more than 30 seconds during independent math practice. Setting: 20-minute practice, worksheets of 10 or more items. Summary of my observation notes: [paste de-identified ABC summary]. Possible functions to treat as hypotheses, not conclusions: escape from task difficulty; access to peer attention. Draft (1) an observable definition, (2) prevention strategies, (3) replacement behaviors to teach, (4) responses to the behavior, and (5) a progress-monitoring plan. Do not add facts I have not given you. Label anything you inferred as "inferred."
Then read the draft against four checks. Does the definition describe something a second observer could count? Does any sentence state the function as settled? Does a prevention strategy assume a resource the classroom lacks? Is anything marked "inferred" that the team should confirm or cut? Bring the edited draft to the meeting as a proposal. The team, which includes the parent, decides the function and the plan.
Which of these are free for teachers?
Three have a free entry point, and the details differ. Flint's pricing page says it is free for teachers, MagicSchool's Free plan carries 80-plus teacher tools, and OpenAI says ChatGPT for Teachers is free for eligible staff through June 2028. SchoolAI's free plan includes Tools and Assistants, but whether that covers the BIP generator is unverified. In each case, a free account means the standard terms; MagicSchool, for one, reserves its custom data privacy agreement for Enterprise.
Is IEP writing software the same as behavior plan software?
No. An IEP sets annual goals and services for a student with a disability, and a BIP is a separate document that describes one behavior and how the team will respond to it. Some platforms offer both, but they draw on different data. Our guides to AI tools for IEP writing and AI tools for writing 504 plans cover the neighboring documents, and AI tools for school psychologists covers the evaluation side.
What does the law actually require?
IDEA names specific situations that trigger an FBA and a BIP, and one of them is a manifestation finding. Under 34 CFR 300.530(f), when a student's team finds that conduct was a manifestation of the disability, the IEP team must either conduct an FBA and implement a BIP or, if a BIP already exists, review it and modify it as needed (34 CFR 300.530(f)). States and districts add their own rules, forms and timelines, and that is what to ask your special education director about. This post is not legal advice.
Does AI save time on behavior plans?
No one has measured it for BIPs. Searches for this post turned up no peer-reviewed study of AI-written behavior intervention plans, so any time-saving figure for plans specifically would be invented. The nearest evidence is a mixed-methods study in the Journal of Special Education Technology, which found no statistically significant difference (p = .67) in quality ratings between IEP goals written only by teachers and AI-generated goals, with teachers reporting positive views of AI for workload (Waterfield and colleagues). That study concerns IEP goals, not behavior plans, and its abstract does not report the sample size. Gallup's survey is broader still: among educators who used AI, 64 percent said it gave them more time for the parts of teaching that matter most (Gallup). The Gallup article reports no findings for special education teachers.
What about Claude, Copilot, Khanmigo and Gemini Notebook?
Claude (4.57 from 21 directory reviews) and Microsoft Copilot work like ChatGPT and Gemini here: a general assistant that follows your structure, to be used with the same four questions. Their education terms were not checked for this post. Khanmigo is built for tutoring and lesson support, and no behavior-plan tool for it turned up in this run's searches. Quizlet is a study tool. Gemini Notebook, the tool Google renamed from NotebookLM in July 2026, answers questions across documents you upload, and the documents a BIP draws on are the ones you are least free to upload.
Three jobs to keep off AI
Observing the behavior. The ABC record has to come from someone who saw it. A tool can format your notes afterward.
Deciding the function. The team weighs the data, the student's history and the parent's view. A generator weighs none of them.
Telling a family. A plan can be drafted. The conversation in which a parent hears what the school plans to do about their child's behavior cannot.
Where this leaves you
Pick one tool your district already allows and use it on a plan you have to write anyway, with a de-identified summary and a prompt that forbids invention. If your district has no written position on AI and special education documents, ask your special education director for one this month; AI tools for special education directors covers the compliance work no tool does. The observing, the deciding and the sitting down with a family are the job. The draft was only ever paperwork.
Dan Fitzpatrick is The AI Educator. Read more from him at aieducator.tools.
