The Problem With AI-Generated Teaching Materials Isn't the AI
On October 8, Teachers Pay Teachers (TPT), the largest online marketplace for teacher-created educational materials, announced plans to ban AI-generated content. According to More Lessons Less Planning’s detailed account of the seller town hall, the planned restrictions cover not only lesson plans and worksheets but also product descriptions and promotional images. They would also apply to products already being sold on the platform. A TPT spokeswoman subsequently told Chalkbeat that sellers’ work should be “unique, expertly crafted, and AI-free.”
I understand why TPT is concerned. Generative AI makes it remarkably easy to produce large quantities of educational content, some of it inaccurate or poorly constructed. Teachers shouldn’t have to discover that a worksheet contains an incorrect answer key or that a history lesson includes fabricated information after their students have already used it.
TPT has emphasized that teachers come to its marketplace for resources created by experienced educators, rather than materials they could generate themselves with AI.
Teachers do bring subject knowledge, professional judgment, and an understanding of their students that AI alone doesn’t provide. What I don’t understand is why TPT believes those teachers should have to create what they sell without using AI.
What AI Makes Possible
I’ve spent decades developing software, and I’ve watched AI fundamentally change how we work. Among the developers I work with, writing code by hand has become the exception. Their experience and judgment remain essential, but the cost of experimenting has fallen dramatically. We can explore different approaches to an idea, see what works, and build on the most promising version.
With AI, trying something new takes less time and effort, making it easier to discover possibilities you hadn’t considered.
Teachers have ideas for lessons, activities, and assessments they may never have had the time or technical skills to develop. AI can make those ideas practical and help them discover new ways to engage students.
Teachers still need to know what students should learn, anticipate where they’ll struggle, and judge whether an activity is useful.
The Trouble With Detecting AI
TPT has said it is developing tools to identify and remove AI-generated materials. I’m skeptical that this approach can be enforced reliably, particularly when AI is used as part of a larger creative process.
Consider a program that generates hundreds of vocabulary worksheets from a list of words. I could use AI to help write the program, then run it without involving AI again. Would the resulting worksheets violate TPT’s policy? Does the answer tell us anything useful about whether those worksheets belong in the marketplace?
AI makes it easier to flood marketplaces with repetitive, low-effort products. Conventional software can do the same. TPT needs to address that problem regardless of how the products were created.
Enforcing the ban presents another problem. Research on AI-polished writing has shown that detectors can flag human writing lightly edited with AI as AI-generated and struggle to distinguish different levels of AI involvement. These limitations matter when enforcement can affect someone’s livelihood.
Finding Problems Before They Reach the Classroom
In my work as a software engineer, I encourage developers to use AI to review their own code before submitting it. We still have automated tests and independent code reviews, but none of these catches every problem. The earlier we catch a mistake, the easier and less expensive it is to fix.
I think educational marketplaces could benefit from the same approach.
Imagine a teacher experimenting with several versions of a worksheet. Before publishing one, they run a check. The system notices that an answer key doesn’t match the questions or that a letter is missing from an alphabet exercise. It flags a questionable factual claim in a history lesson, with references for the teacher to investigate.
The teacher can correct the problems and try again. Validation becomes part of creating the resource.
Some checks could use conventional software. Others would benefit from AI’s ability to analyze documents and investigate factual claims.
I’ve helped build AI-powered software that links generated answers to supporting sources. That work has given me an appreciation for the difficulties of verifying factual claims. Researchers have already developed methods for checking individual claims against supporting evidence. We wouldn’t be starting from scratch.
AI can make mistakes when evaluating content, and these tools would need testing for accuracy and false positives. But we can start with problems we know how to detect and improve the tools over time. They could help anyone creating educational content.
TPT would still need to address repetitive listings, misleading descriptions, and resources with little educational value, regardless of how they were created. Disclosure could help buyers make informed choices about AI use. But I’d prioritize helping teachers identify and correct problems before they publish.
TPT already has automated moderation and a process for reviewing resources reported by buyers. Seller tools would build on those efforts.
I’d much rather see a teacher correct an error while developing a lesson than have another teacher discover it after thirty students have already completed the assignment.
The Marketplace TPT Wants to Build
As AI makes it easier for teachers to create their own resources, TPT will need to give them compelling reasons to keep buying and selling on its platform.
In its 2026 back-to-school survey, TPT identified resource creation as the most common use of AI among educators. The survey also found that educators strongly preferred teacher-created materials over purely AI-generated content, but most were comfortable with AI-assisted resources reviewed or refined by a human.
An AI-free marketplace may appeal to some buyers. But it could also create an opportunity for competitors who welcome teachers using AI and help them produce reliable materials.
IXL Learning, TPT’s parent company, could instead help establish how AI is used responsibly in educational publishing. Working with educators and AI companies such as Anthropic and OpenAI, it could develop shared evaluation frameworks and tools useful well beyond its marketplace. Educators should play a central role in directing that work.
I hope TPT reconsiders its planned ban. There are genuine problems with AI-generated educational content, but helping teachers create, experiment, and correct mistakes seems a more worthwhile investment than trying to determine which tools they used to create a lesson.
We don’t know what teachers will create as AI makes it easier to experiment. I’d like to see TPT help them find out.
I developed this article’s argument from my experience in software development, using AI to assist with drafting and editing. I reviewed and revised the text and take responsibility for the final piece.