How to Tell if Your IEP Was Written by AI (or Just Badly Written)

If you don’t live in the online space like I do, you may not realize just how much buzz there is around AI right now. Depending on who you ask, it’s either going to cure every major disease and solve climate change, or it’s going to take over jobs and create a whole new set of problems we’re not ready for.

The truth is probably somewhere in the middle. But one thing is certain, AI is already being used to write content-a lot of it. And it’s only a matter of time before schools start using it more in special education, particularly for writing IEPs.

You might be hearing more about “AI IEP” tools or using AI for IEP writing, especially as schools look for ways to save time. But just because something can generate an IEP doesn’t mean it should be used without careful consideration.

On the surface, it makes sense. Schools are overworked, understaffed, and stretched thin. Writing IEPs is time-consuming and detailed, and anything that promises to speed up that process is going to be appealing.

But IEPs are not supposed to be quick. They are supposed to be individualized, thoughtful, and based on actual data about your child.

This is not an unbiased article. But, I am not anti AI. My own son had AI-assisted brain surgery. I am however, against most LLMs. LLM stands for Large Language Model. All LLMs are AI, not all AI is an LLM.

An LLM is the basic AI platform that the average person uses–you know, you put in a question like “how do I do this” and it replies. The things like chatgpt, claude, etc. And yes, some teachers are copying and pasting entire evaluation reports with a “write me a present levels for this child” and the copying and pasting that into your IEP.

It’s important to note: LLMs have limits. For several years, they scraped the internet for information and used it to train the LLM. So what an LLM basically does is quickly regurgitate the information that is already on the internet, based on your query or prompt. The LLM is limited by what information is available. The word “intelligence” isn’t really accurate–as LLMs cannot “think” beyond the information that is already in them.

Even if they are using one of the newer AI IEP platforms, the platform is limited by the information in it. And, more and more, true educators and publishers are protecting their content from being scraped by AI. (myself included!)

One thing I hear is “well get used to it, it’s coming whether your like it or not so you’d better just adapt.” Those arguments all originate from people who have millions or billions invested in LLMs. There’s a reason the term “AI slop” is now common, because much of what comes out of LLMs is slop. And that includes IEPs.

Why This Is Happening

IEPs are one of the most cumbersome tasks teachers are responsible for. Even with good software, they take time. They require input from multiple people, data analysis, and careful thought about goals and services.

AI offers a shortcut. It can take existing information and generate something that looks polished and complete in a fraction of the time. But that’s also the problem.

AI doesn’t “know” your child. It doesn’t attend meetings, observe behaviors, or understand nuance. It pulls from patterns in data that already exist elsewhere. And if those patterns are generic, biased, or incomplete, that’s exactly what you’re going to get back. LLMs are known to have bias–this has been tested repeatedly.

This Is Not Just a “Bad IEP” Problem

There are already tools online being marketed as an “AI IEP writer” or platforms designed specifically for AI IEP writing. Before anyone uses those tools with real student data, parents should understand what information is being entered and where it is going.

Some educators and parents are experimenting with tools like ChatGPT for IEPs, using them to draft goals or sections of the document. Again, the concern isn’t just the quality of what’s written, it’s what happens to the student data being entered.

It is tempting to make this only about whether the IEP is any good. And yes, of course that matters. But that is not the whole issue.

If school staff are entering your child’s evaluation results, disability category, present levels, behavior concerns, medical needs, therapy information, accommodations, goals, or other IEP data into an AI tool, that is a huge privacy concern.

IEPs contain protected student information. They often include deeply personal details about a child’s disability, behavior, communication, mental health, medical needs, academic struggles, and family concerns.

That information should not be used to “feed” a large language model. And parents should not find out after the fact that their child’s private information was entered into an AI system because someone wanted to save time writing a document.

The Real Question Parents Should Be Asking

The question is not only, “Was this IEP written by AI?” The better questions are:

• Was my child’s personally identifiable information entered into an AI tool?
• What tool was used?
• Is it approved by the district?
• Does the district have a written AI policy?
• Does the district have a data privacy agreement with that company?
• Was student information stored, retained, or used to train the model?
• Who has access to what was entered?
• Was parent consent required?

Because this is not the same as using spellcheck or cleaning up grammar. An IEP is full of sensitive student data. If that data was copied into an AI platform, parents deserve clear answers. And no, “we’re just trying to save time” is not an answer.

Signs Your IEP Might Be AI-Generated

You are not going to find a disclaimer that says “this document was written by AI.” What you will find are patterns.

If you’ve ever Googled how to tell if text is AI generated, you’ll see common signs like repetitive phrasing, generic language, and lack of detail. Those same signs can show up in IEPs too, which is why it’s important to read closely.

One of the biggest indicators is vague, generic language. Goals that say things like “the student will improve reading skills” without specifying how that improvement will be measured, over what period of time, or using what baseline data. If a goal could be copied and pasted into any other child’s IEP, it is not individualized.

If your overall IEP is dramatically different from the previous one, compare them. Unless the entire IEP was put into the LLM, the LLM has no context. So that means the IEP will be very different, without any reason.

Another issue I see often is the absence of meaningful baseline data. Without a clear starting point, there is no way to measure progress. This is already a common problem in IEPs, but AI tends to amplify it because it can generate language without grounding it in actual student data.

You may also notice that the document feels repetitive or formulaic. The phrasing is consistent throughout, sometimes to the point where it feels unnatural. It may read smoothly, but when you slow down and really look at it, there is very little substance behind the words.

A mismatch between evaluations and goals is another red flag. If the evaluation identifies specific needs, but the goals do not address those needs directly, something is not lining up. That disconnect is a problem, regardless of how the document was written.

There is also what I call the “sniff test.” This is not scientific, but it matters. I wouldn’t go so far as to call it a “fake IEP,” but if a document is so generic that it could apply to any student, then it’s not truly individualized. And that’s the standard, IEPs are supposed to be written by humans who know your child, not something that feels like it was generated by a template.

Does the IEP sound like it was written by someone who knows your child? Does it reflect what you see at home? Does it feel noticeably different from previous IEPs in a way you cannot quite explain?

Many parents describe AI-generated text as having a slightly off tone, as if it is technically correct but lacking something human. They often sound a “little too perfect” to the point of sounding clinical.

If you are reading your child’s IEP and thinking, “this doesn’t quite sound right,” that is worth paying attention to.

The LLMs (large language models) have a habit of using the same words over and over, using filler words and they often are a little “off” with spacing.

Since I have an online business, I can tell you–once you “see” something written by AI, you can’t unsee it. All those long Facebook posts that are long stories that tug at your heartstrings, complete with an AI generated image? Yeah, all fake, never happened.

The cadence is all the same, and it sounds “cold.” I don’t know how else to explain it.

Is it legal to write an IEP with AI?

There is no law (right now) that specifically says, “schools cannot use AI to write an IEP.” But don’t stop there, because that’s not the real answer.

Here’s where it gets tricky–IEPs are governed by IDEA (federal law), and IDEA is very clear about a few things:

• The IEP must be developed by a team (not a machine)
• It must be individualized to the child
• It must be based on data and evaluations
• Parents must have meaningful participation in the process

So if AI is being used in a way that interferes with any of that? Then you potentially have a legal problem.

Where AI can cross the line

Using AI as a tool (like cleaning up wording) is one thing.

But if AI is:

• generating goals without real data
• creating a “finished” IEP before the meeting
• replacing team discussion and decision-making

Then you could argue that the IEP was not actually developed by the team or truly individualized. And, parent participation was not meaningful. And those are all IDEA issues.

The BIG legal concern (and this is the one I care about most)

Privacy. IEPs contain personally identifiable student information. That’s protected under FERPA.

If a teacher or district is putting your child’s data into:

• ChatGPT
• an “AI IEP writer” tool
• any platform without a proper data agreement

Then you have to ask:

• Where is that data going?
• Is it being stored?
• Is it being used to train models?
• Who has access to it?

Because FERPA requires schools to protect that data. And right now? The law has not caught up to AI use in schools. So we’re in that messy gray area where:

• it might not be explicitly illegal
• but it could absolutely become a violation depending on how it’s used

What to Do If You Suspect It

If you suspect that your child’s IEP may have been generated or heavily influenced by AI, the goal is not to prove that point in a dramatic confrontation. The goal is to make sure your child has an appropriate and effective IEP.

Start by asking for the data that was used to develop the goals. Ask how those goals were determined and what sources were used. These are reasonable questions, and they should be answered clearly.

Take the time to compare the evaluations, present levels, and goals. They should align. If they do not, that is where your focus should be.

And, as always, keep your communication in writing. Documentation matters. What is said in a meeting can be forgotten or reframed later. What is written creates a record.

What I would do as a parent: I wouldn’t walk into a meeting accusing anyone of “illegally using AI.” But I would absolutely ask, if I had concerns:

• What tools were used to draft this IEP?
• Was any student data entered into AI platforms?
• Does the district have an AI/data privacy policy?

The Privacy Issue No One Is Talking About Enough

There is another layer to this that deserves attention, and that is privacy.

AI systems rely on data. If a school is using an open AI platform to generate content, there is a possibility that the information entered into that system is being stored or used in ways you are not aware of.

Even if a district is using a closed system, the data is still being added to some form of database. That raises questions about who has access to that information and how it is being used. Is it being used to train the models for the entire platform–in other words, every school district who buys this same software could have access?

This is a concern I would raise with the school board, as it is your school board that writes policy for the district. If they do not have an AI policy, they should have one in the works and it should include IEP development.

Right now, the laws have not fully caught up to this issue. FERPA applies, but HIPAA does not in this context. If you have concerns about how your child’s information may have been handled, it may be worth consulting with someone who understands education law.

At the end of the day, I cannot tell you with certainty whether your child’s IEP was written by AI. Most will not admit it, even if pressed, is my guess.

What I can tell you is this: if the IEP is vague, lacks data, and does not reflect your child’s actual needs, then it is not doing its job. That is where your focus should be.

Because whether it was written by a person, copied and pasted from another document, or generated by AI, the standard does not change. The IEP still has to be individualized, data-driven, and appropriate.

If you are unsure what that should look like, start with understanding how goals are written, how data is used, and how to document your concerns. Those are the pieces that will make the biggest difference in the long run.

And, I would talk with other parents in your district and reach out to your school board to learn what your school board’s AI policies are, especially as pertains to IEPs.

Using AI to write IEPs might sound like a harmless shortcut, but IEPs contain protected student information. That makes this very different from using AI to write an email or lesson plan. Make sure you read yours carefully.