A teacher catches you in the hallway between classes. She’s been using a free AI tool to draft lesson plans. Her students love its quiz feature, and she wants to know whether she can keep using it.
It sounds like a simple yes-or-no question.
But what would you need to know before answering?
You could check the vendor’s privacy policy. You could find out what happens to student data. You could make sure the tool works with your school accounts.
Those are important steps, but evaluating AI tools requires schools to go further.
What exactly is the AI being asked to do? What happens when it produces a bad answer? Who reviews its recommendations? Is there evidence that the tool actually accomplishes what the vendor says it does? And what happens if the technology behind it changes after you approve it?
As we discussed in The Pros and Cons of Artificial Intelligence in K-12 Schools, schools need to vet AI tools before rolling them out. Here’s a practical framework for doing it.
Start With Purpose: What Are You Asking AI to Do?
Before watching a product demo, define the problem.
There is a big difference between using AI to brainstorm ideas for a school event and using it to evaluate student work or recommend an academic intervention.
Schools might use AI to:
- Draft lesson plans or classroom communications
- Create quizzes and instructional materials
- Tutor students
- Provide feedback on student work
- Analyze student performance
- Recommend interventions or resources
- Automate administrative work
Each use comes with different risks.
An AI tool that suggests a bad theme for spirit week isn’t much of a concern. An inaccurate recommendation that influences a decision about a student is.
Before evaluating the product, ask:
What problem are we trying to solve? Be specific. “We want teachers to use AI” isn’t a goal.
Why are we using AI for it? The technology should solve a real problem or improve an existing process.
What happens if the AI is wrong? The higher the stakes, the more oversight you need.
Is the AI assisting a person or making a decision? Be especially careful when an AI-generated recommendation could affect an individual student.
It’s not possible to eliminate every possible risk, but the goal is to make sure the safeguards match the way the tool will actually be used.
Check the Evidence: Does the Tool Actually Work?
An impressive demo or a vendor saying the product is “research-based” doesn’t tell you whether a product will improve teaching or learning.
Ask what that actually means.
What evidence supports this specific product or feature?
Research showing that AI has potential in education isn’t evidence that one particular AI tool works.
What did the research measure?
Did students learn more? Did teachers save time? Did engagement increase? Did users simply report that they liked the product?
All of those outcomes can have value, but they aren’t interchangeable.
Who conducted the research?
Look at whether evidence comes from the vendor, an independent researcher, or schools that have implemented the product.
Were schools like yours included?
A study involving college students or a very different school environment may not tell you much about how the product will perform with your students.
Then ask one more question internally:
How will we know if it’s working here?
If you’re adopting a tool to save teachers time, decide how you’ll measure time saved. If the goal is better student feedback, define what better feedback looks like.
Otherwise, renewal season arrives and “people seem to like it” becomes your evaluation strategy.
Follow the Data: What Goes In, and Where Does It Go?
AI doesn’t replace the need for traditional technology and security vetting. If anything, it makes that work more important.
Schools still need to understand how a vendor collects, stores, shares, protects, and deletes information.
Start with these questions:
What student and staff information does the tool collect?
Look for specifics rather than broad statements about collecting information to “improve services.”
Does the company use your information to train or improve AI models?
Ask specifically about prompts, uploaded files, responses, and other content. Get the answer in writing, and determine whether the policy changes depending on the type of account or subscription.
Who else has access to the data?
Ask about hosting providers, subprocessors, and other third parties involved in delivering the service.
How long is the information kept, and can your school delete it?
Understand both the process and the timeline.
What happens if there is a security incident?
Review the vendor’s commitments around security and breach notification.
Are your students permitted to use the product under its terms?
Check age restrictions, account requirements, and any privacy requirements that apply to your school and students.
Depending on the school, the students involved, and how the tool will be used, FERPA, COPPA, state privacy laws, contracts, and other requirements may apply. Schools should work with their legal counsel when determining their specific obligations.
There is also an AI-specific issue worth considering: what users are likely to type into the tool.
A teacher may begin with:
Give me three ways to help a student who is struggling with fractions.
Then the teacher realizes a little more context would produce a better answer and adds the student’s name, assessment results, accommodations, or other sensitive information.
An open prompt box makes it easy to share more than intended.
When evaluating an AI tool, don’t look only at the information the vendor requires. Think about the information teachers, staff, and students could realistically enter while using it.
Keep People in the Process: Who Checks the AI?
Everyone has heard that AI can make mistakes.
The more useful question for school leaders is: What happens when it does?
The National Institute of Standards and Technology’s AI Risk Management Framework identifies factors such as reliability, transparency, privacy, and managing harmful bias as important considerations when evaluating AI systems.
Before approving a tool, determine where human judgment belongs in the process.
Who reviews the output?
For brainstorming or first drafts, the person using the tool may be enough. Student-facing content and recommendations affecting individual students may require greater oversight.
Will the person reviewing it recognize a bad answer?
AI becomes more concerning when users don’t have the subject-matter knowledge needed to identify an error.
Can users question a recommendation?
If a tool recommends an intervention, placement, grade, resource, or other action, educators should be able to exercise their own professional judgment rather than simply accepting the output.
Has the tool been evaluated for bias?
Ask what the vendor has done to test for differences in accuracy, reliability, or outcomes across relevant groups of users.
Finally, ask:
What should this tool never be used to do?
That’s important because approving a product does not mean approving every possible way someone could use it.
Plan for What Happens After Approval
Approving an AI tool shouldn’t be a one-time decision.
AI products can change quickly: vendors add features, underlying AI models change, integrations are introduced. Terms and data practices can change, too.
Before rollout, establish some basic rules.
Who owns the decision?
Make it clear who can approve AI tools and when IT, academic leadership, legal counsel, or other stakeholders need to be involved.
What uses are actually approved?
Be specific. “Approved for staff lesson planning with no student information” is much more useful than “approved AI tool.”
What can administrators monitor and control?
Look for the ability to manage accounts, configure features, review usage, and make changes centrally.
When will you review the tool again?
The contract renewal date is an obvious checkpoint, but don’t make it the only one.
What should trigger an earlier review?
A major new feature, a change in terms, a security incident, or a proposed new use may all be reasons to look again.
What happens if you stop using it?
Know how to export school-owned materials, close accounts, and confirm that school data has been deleted.
Don’t Just Approve the Tool. Approve the Use.
Schools often want a simple list of approved and unapproved AI tools. That makes sense. Teachers need clear guidance.
But AI makes a simple yes-or-no list difficult.
The same tool might be perfectly reasonable for one task and a poor choice for another.
For example, your school might approve a generative AI tool for teachers brainstorming lesson ideas using nonsensitive information. That doesn’t necessarily mean teachers should paste student records into the same tool or use its recommendations to make decisions about individual students.
Instead of recording:
Tool X: Approved
try:
Tool X: Approved for staff brainstorming and first drafts. Do not enter confidential or identifiable student information. Review all outputs before using them with students. Review again in January 2027.
Now your teachers know what “approved” actually means.
Your AI Tool Evaluation Checklist
Before approving a new AI tool, work through these five areas.
Purpose
- What problem are we solving?
- What will the AI actually do?
- What happens if it is wrong?
- Is AI appropriate for this task?
Evidence
- What does the vendor claim the tool improves?
- What evidence supports that claim?
- Does the evidence apply to our students and use case?
- How will we measure success?
Data and Security
- What information will users enter?
- How is that information stored, used, shared, and deleted?
- Is school data used to train or improve AI models?
- Does the product meet our privacy and security requirements?
Human Oversight
- Who checks the output?
- Can that person recognize errors?
- Can users challenge or override recommendations?
- Has the vendor evaluated the tool for bias?
- Are there uses we should specifically prohibit?
Ongoing Management
- Who approves the tool?
- What specific uses are allowed?
- What can administrators monitor and control?
- When will we evaluate it again?
- What changes would trigger a new review?
- What happens to our information if we leave?
You don’t necessarily need an immediate yes-or-no answer to every AI tool that comes your way.
Sometimes the most useful result of the evaluation is discovering what you still need to know before saying yes.
Build an AI Strategy That Fits Your School
AI can save teachers time, support students, and help schools work more efficiently. But choosing the right tools takes more than reading a privacy policy or watching a product demo.
Schools need a repeatable process for determining where AI adds value, what safeguards are needed, and whether a tool continues to deliver on its promise.
Technology Lab works with K-12 schools to evaluate technology, strengthen security, and develop technology strategies aligned with each school’s goals.
If your school is deciding which AI tools to allow or needs a more consistent process for evaluating them, contact Technology Lab to start the conversation.
FAQs: Evaluating AI Tools for K-12 Schools
What should schools look for when evaluating an AI tool?
Schools should evaluate an AI tool’s purpose, evidence of effectiveness, data and security practices, need for human oversight, and ongoing management requirements. Schools should also evaluate the specific ways they plan to use the tool, since the same AI product may be appropriate for one use but not another.
How can schools determine whether an AI tool is safe for student use?
Schools should review what student information the AI tool collects, how that information is stored and shared, whether it is used to train or improve AI models, and what privacy and security protections the vendor provides. Schools should also consider age requirements, applicable privacy requirements, administrative controls, and how students will actually use the tool.
Should schools allow teachers to use free AI tools?
A free AI tool should go through an appropriate review before it is used with school or student information. Schools should pay particular attention to what teachers may enter into open-ended AI prompts, whether that information is retained or used to improve AI models, and whether the school’s approved use is consistent with the vendor’s terms and data practices.
How often should schools reevaluate approved AI tools?
Schools should establish regular review points for approved AI tools, such as contract renewals, while also identifying changes that could require an earlier review. A new AI model or feature, revised terms or data practices, a security incident, or a proposed new use may all be reasons to reevaluate a tool.
