Is AI safe for students? Should you use AI as a learning tool? Should you be using it in your classroom at all?
I get some version of that question at almost every educator training I run. And I get why. The headlines are scary, the tools are multiplying faster than anyone can vet them, and most teachers were never actually taught how to evaluate an AI tool. They were just handed one (or ten) and told to figure it out.
Here’s the honest answer: AI isn’t automatically good or bad for learning. It depends entirely on the tool and how you use it as a teacher.
That’s why I built the Guide, Don’t Give framework: a two-part method to help you:
- Choose the right AI tools to use with students
- Design the learning experience in a way that benefits learning
The Framework Part 1: Choose the Right Tool
Stanford’s AI Hub for Education released a March 2026 review of educational research around AI in education. They found that the majority of AI in education research focuses on students as users of AI tools and using AI as a learning tool versus how educators use AI in their work. A major finding in the research review was that tool design matters. Here are some quotes from the research summary:
- “Not all AI tools function the same way in learning environments.”
- “Evidence suggests that AI tools designed with pedagogical guardrails—such as tutoring systems that give hints or guide reasoning—show more promising outcomes than general-purpose chatbots that provide answers directly.”
- “Tools that scaffold reasoning may help support learning, while tools that simply generate answers may reduce the cognitive effort that supports durable skill development.”
Put simply: Not all AI tools function the same way in a learning environment. There is a huge difference between a tool that builds thinking and a tool that bypasses it. Tools that scaffold a student’s reasoning tend to support learning. Tools that just generate the answer tend to reduce the thinking the student actually has to do.
Before any AI tool earns a spot as an instructional tool, it needs to meet two requirements:
1. Guardrails
Think of the student’s experience as a target:
The Center: This is the learning goal that the teacher specifically sets for the AI.
The Guardrails: These are the supports that keep the student’s arrow aimed at the center—enabling exploration while maintaining focus.
Well-designed AI guardrails perform three essential functions that support learning:
- Ensure all outputs are school and developmentally-appropriate. Ensuring school appropriate outputs means students can access AI’s power without you worrying about inappropriate content showing up. The guardrails also flag any concerning user-AI interactions and make this very obvious to the teacher. Developmentally appropriate outputs from the AI is also really important. Think about it. You speak very differently to a 5th grader than a 12th grader. The teacher needs the ability to tell the AI what age student the activity is for.
- Constantly redirect students back to the learning goal. First of all, the tool has to be designed in a way where the teacher designs the interaction and identifies the learning goal on the backend. Then, if the student asks “When’s lunch?” when the goal is learning about ecosystem balance, the AI is designed to nudge the learner back to the learning goal the teacher identified. And just like in your classroom, the redirect isn’t always immediate. Often, the nudge involves finding something that sparks the learner’s interest in the topic at hand.
- Scaffold the answer rather than supplying the answer. Instead of giving the answer directly, these guardrailed tools made for educational environments guide learners to find the answer or end goal themselves. Instead of letting kids wander off into “write my essay for me” or “give me the answer” territory, these tools keep nudging students toward deeper thinking using what the research refers to as the “socratic method.” For example, if a student asks the AI to “write my essay for me,” a guardrailed tool would redirect that student to work through the steps to complete the task, such as outlining their argument first. The AI becomes a scaffold, not a shortcut.
When guardrails work well, they enable the good stuff, such as customization, translation, explanation, connection-making, while preventing the AI from doing the thinking for students. This is how you use AI as a learning tool!
2. Teacher Dashboard
The second non-negotiable in tool choice is teacher visibility. If you don’t get insights from the students’ interactions with the AI, you can’t teach with it, you can only guess.
Use AI tools in your classroom that include dashboards that allow you to:
- See how students are actually interacting with AI (not just guess)
- Monitor progress and misconceptions in real time (before they become bigger issues)
- Identify who needs support and who’s coasting (differentiation that’s actually informed)
- Keep AI use transparent, not hidden (so it becomes something you teach with, not police against)
- Provide key insights to guide the next instruction step (such as providing grouping recommendations)
- Clearly alert the teacher of any concerning user-AI interactions (alerting the teacher when human intervention is needed)
This changes the dynamic entirely. When AI use happens inside a teacher-controlled environment with visibility, it shifts from “catch them cheating” to “guide their learning.”
The Framework Part 2: It’s All in the Instructional Design
Even the right tool can be used the wrong way in an instructional context. Once you have the right tool, the rest depends on how you use that tool from an instructional perspective. I have found two key factors from an instructional design standpoint:
3. Short, Targeted Time Periods
AI isn’t meant to run your whole class period, and it’s not the answer to everything. It works best in short, targeted bursts, generally 5 to 10 minutes. The last thing we want is students talking to chatbots all day!
4. Key Instructional Moments
Use these AI tools at key instructional moments. I have field tested with learners of all ages and identified four key instructional moments where AI can really enhance the learning experience:
- The Hook: spark curiosity for a topic at the start of a lesson
- Brainstorm: fast, low-stakes idea generation
- Personalized Checkpoints: a quick check that a student is actually on track
- Immediate Feedback: push thinking further, or catch a misconception before it sticks
Notice what all four have in common: they’re moments where a quick, personalized nudge has outsized value. That’s the sweet spot.
The Bonus: Building AI Literacy Through Practice
Here’s something I didn’t expect when I started using guardrailed AI tools as an instructional tool: it doesn’t just support the learning, it builds AI literacy as a side effect. Students are learning how to use AI thoughtfully, a skill they’ll need for the rest of their lives.
One of the best ways to do this is to teach students to personalize their own learning by prompting the AI directly. A few prompts worth teaching explicitly:
- For simplification: “Explain this like a 3rd grader.”
- For reading level: “Change this to a 700 Lexile score.”
- For translation: “Translate this to Spanish.”
- For student interest: “How does this relate to volleyball?”
- For task initiation: “Break this task into steps.”
Notice that last one is also an executive functioning skill. Task initiation, cognitive flexibility, metacognition, these are things our students need regardless of whether AI is involved. When a student learns to ask a tool to break a task into steps, they’re practicing a skill that transfers far beyond that one assignment.
And when students use guardrailed AI tools in your classroom, they’re learning things that matter for the rest of their lives:
- When AI is actually helpful, and when it’s not
- How to prompt effectively for what they need
- How to evaluate and question an AI’s output instead of accepting it at face value
- How to use AI to support their thinking instead of replacing it
And maybe the most important one: How AI can feel like a friend. It listens, it responds instantly. But it isn’t a friend. It doesn’t know you, it has no stake in your life, and it’s built to keep the conversation going, tell you what you want to hear, not to act in your best interest.
That’s AI literacy. And in this day and age, it’s not an elective skill, it’s a core one. Integrating these tools into the learning process sparks critical AI literacy conversations that we must have with kids.
Where This Leaves You With Using AI as a Learning Tools
Not all AI is created equal, and not all AI use leads to learning. But when a tool has real guardrails, gives you real visibility, and you use it in short bursts at the right instructional moments, AI can become a genuine learning tool instead of a shortcut around learning.
That’s the framework: guide, don’t give.
Want the quick-reference version? I put the whole framework, both non-negotiables, the four key moments, and the prompting cheat sheet, into a visual infographic you can keep on your desk or share with your team.
Fill out the form below and it will land in your inbox:
