AI tutoring tools policy for schools
AI tutors are the AI tools schools most want to say yes to. This guide shows how to allow them with controls: which kinds to approve, what the data agreement must say, how age limits work and how the filter enforces the decision.
What counts as an AI tutoring tool
"AI tutor" covers very different products. A policy that treats them all the same either blocks useful tools or lets risky ones through.
District-licensed tutors
Products sold to schools, set up with school accounts and usually a data agreement. Teachers can often see student sessions.
Tutoring inside learning platforms
AI help built into a platform the school already uses, such as a reading or maths program. Often switched on by default in updates.
General chatbots used as tutors
Consumer assistants that students ask for help. Their terms are written for the public, not for schools.
Photo-to-answer apps
Apps that solve a photographed problem and show the answer. Some show steps, but most are built for speed, not learning.
Language and reading tutors
Conversation practice, pronunciation feedback and reading coaches. Many use voice, which raises extra data questions.
Writing feedback tools
Tools that comment on drafts. The line between feedback and rewriting the work varies a lot between products.
The district-wide approach to every kind of AI tool, from chatbots to image generators, is set out in our AI tools policy for schools. This page goes deeper on tutoring only.
Allow the tutors you approved and hold back the rest
The hosted CIPA filter lets you decide at the level of a category and then make exceptions for named domains. Approved tutoring sites go on an allow list while other AI tools stay blocked.
- AI / LLM Tools can be blocked as a category, with each approved tutor allowed for the schools that need it.
- A new AI domain lands in the category data without you adding it, so a fresh answer app does not slip through.
- Edits to the allow list go live in seconds across all schools.
Three tiers: allow, allow with controls, block
Sorting tutoring tools into three tiers gives teachers a clear answer and gives the filter a rule it can enforce.
Allow for set grades
- Signed data agreement in place
- Students sign in with school accounts
- Guides students through steps rather than giving final answers
- Teacher can review sessions
- Age terms match the grades allowed
Allow with controls
- Useful but missing one safeguard
- Teacher-led use in class only
- Older grades only, or staff only
- Time-limited pilot with a review date
- Feature limits, such as voice turned off
Block for students
- No acceptable data terms
- Built to hand out answers to assignments
- Age terms exclude your students
- Companion or role-play features
- Advertising or data sale in the terms
Blocking every tutoring tool pushes students to unvetted apps on personal phones. A short approved list usually gives the district more control, not less.
An evaluation rubric for AI tutors
Score each tool before approval. Any "red flag" answer moves the tool to the block tier until the vendor fixes it.
| Area | Look for | Red flag |
|---|---|---|
| Teaching approach | Asks questions, hints and checks understanding. Can be set to withhold final answers. | Gives full solutions on request with no setting to change it. |
| Content safeguards | Refuses harmful, sexual and violent requests. Stays on academic topics. | Open conversation on any topic, or persona and role-play modes. |
| Wellbeing responses | Points students to a trusted adult or school support when distress comes up, as set by the school. | No defined response, or advice that replaces professional help. |
| Teacher visibility | Teachers see assignments, progress and, where the policy allows, session history. | No educator view at all. |
| Accounts | School single sign-on, roster sync and admin controls. | Personal email sign-up only. |
| Accuracy | Shows working, cites curriculum sources, lets teachers report errors. | No way to flag wrong answers. |
| Accessibility | Works with screen readers, text size and translation needs. | Excludes students with disabilities from a core resource. |
| Data terms | Signed student data agreement, see below. | Click-through consumer terms only. |
Involve at least one teacher who will use the tool and one person who reads contracts. Technology staff alone tend to miss classroom problems, and teachers alone tend to miss data terms.
Write down who scored the tool and when. A dated record answers parent questions and helps at the yearly review.
The data agreement: what it must cover
Tutoring tools collect more than most apps: questions, mistakes, writing, sometimes voice. The agreement decides what happens to all of it.
Many states have a standard student data privacy agreement, often shared through a state or regional alliance. Start from that template rather than the vendor's own terms.
Under FERPA, a vendor can act as a school official when it is under the district's direct control for the data and uses it only for the authorised purpose. Our guide to student data privacy, FERPA and web filtering explains how unvetted AI tools put that at risk.
Age limits and parental consent
Age rules come from two places: the law and each tool's own terms. Your approved list should record both for every tool.
Adult-led only
Students do not use tutoring chat themselves. Teachers may use tools to prepare activities.
Built for children
Only tools designed for this age, under a school agreement. Short sessions, teacher-assigned tasks.
Approved tutors
Under-13 rules apply to many students. Use school-agreement tools; check each tool's minimum age.
Wider choice
More tools fit by age, but many consumer terms still require parent permission under 18.
The law: COPPA for under-13s
- Operators need verifiable parental consent to collect personal data from children under 13
- Under FTC guidance, a school may consent in place of parents when data is used only for the educational purpose
- The FTC amended the COPPA Rule in 2025, with most compliance due by 22 April 2026
- The FTC did not adopt a new school authorization rule, so the existing guidance applies
The terms: each tool's own age rules
- Many consumer AI services set 13 as the minimum age
- Many also require parent or guardian permission for users under 18
- Some education products are built for younger students under a school contract
- Terms change, so recheck them at each review and after major updates
State student privacy laws can add duties on top of COPPA and FERPA. Check your state's page in the state school internet safety laws section.
Where well-known tutoring and AI tools sit today
Many tutors are filed under AI / LLM Tools, and some also carry an education label. This is why a policy needs both a category rule and a short list of named approvals.
| Tool | Our categories |
|---|---|
| khanmigo.ai | Education/Reference, AI / LLM Tools, Web-based Applications, Webinars/E-learning |
| magicschool.ai | AI / LLM Tools, Education/Reference, Web-based Applications, Webinars/E-learning |
| photomath.net | AI / LLM Tools, Science |
| gauthmath.com | AI / LLM Tools, Education/Reference |
| socratic.org | AI / LLM Tools |
| duolingo.com | Education/Reference |
How the filter enforces the tutoring policy
A written list of approved tutors does nothing on its own. The filter makes the list real on every school device and network.
Block AI by default
Set the AI tool categories to block for students, with tutoring reviewed separately.
Allow approved tutors
Add approved tutoring domains to an allow rule for the grades named in the list.
Hold new tools
New AI domains stay blocked until someone reviews them.
Watch embedded AI
Check settings in approved platforms when new AI features arrive in updates.
Review requests
Teacher requests for a new tutor go through the rubric, not a quick unblock.
Our AI tools set covers 11,800+ AI-tool domains in 18 categories and 165 subcategories, updated daily, so a tutoring app launched last week is still recognised. See the AI tools blocklist for schools.
Many answer apps are phone apps. A filter on school devices cannot reach personal phones, so pair it with your phone policy and with assessment design.
Approvals and holds that you can change quickly
Always-allow and always-block lists are set per school, so a tutor approved for one grade band does not open for everyone. Each change is recorded in the policy record that supports your CIPA file.
Keeping tutoring honest: assignments and integrity
A good tutor helps a student learn. The same tool used at the wrong moment completes the work for them.
Where tutors help most
- Practice sets and homework help with hints
- Reviewing a concept before a test
- Reading practice and vocabulary
- Language conversation practice
- Support for students who missed a lesson
Where to switch them off
- Tests and graded quizzes, using test-mode filter rules
- Take-home essays where the writing is assessed
- Coursework for external exams with their own AI rules
- Any task labelled "no AI" by the teacher
- Diagnostic assessments used to place students
For essay mills, answer sites and test-time controls, see academic integrity beyond AI. For the broader question of general chatbots in class, see whether general AI chatbots should be allowed in schools.
Funding, equity and evidence
AI tutoring is often sold as a way to reach more students for less. Ask for evidence and plan for students without devices or home internet.
Count the hidden costs too: staff time for reviews, training for teachers and the work of keeping the approved list current in the filter.
Confirm allowable use of any federal funds with your state program office before you commit money.
A one-page AI tutoring policy outline
This outline sits under your wider AI policy. Keep it short enough that teachers actually read it.
| Section | What it says |
|---|---|
| Purpose | Why the district uses AI tutoring and what it should improve for students. |
| Approved list | Where the current list of approved tutors lives, with grade bands for each. |
| Approval process | Who scores new tools, the rubric used, and how long a review takes in normal cases. |
| Data rules | No tool is used with students without a signed data agreement. |
| Age rules | Tools are only used within their age terms and the grade bands set by the district. |
| Classroom use | When teachers may assign tutors, and when AI help is not allowed. |
| Families | How parents are told which tutors are used and how to ask questions or opt out where offered. |
| Review | The list and the policy are reviewed at least once a year, and when a tool changes its terms. |
Publish the approved list where teachers and parents can find it, and date each update.
Fold the tutoring rules into your acceptable use policy so students and families see them in one place.
Running a tutoring pilot before you approve
A one-term pilot with a few classes shows how a tutor behaves with real students. It also gives you evidence to show the board and families.
Pick the classes
Choose two or three teachers who volunteer, across different grades or subjects.
Set the goal
Name one measure, such as practice completion or unit test results, and compare with similar classes.
Sign first
Put the data agreement in place before any student account is created.
Open the filter narrowly
Allow the tutor only for pilot classes, using groups or organizational units.
Collect feedback
Ask teachers and students what helped and what went wrong, including wrong answers.
End the pilot with a decision: approve, approve with controls, or block. A pilot that drifts into permanent use without a decision is the most common gap in AI tutoring policy.
Questions to ask an AI tutoring vendor
Send these before a demo. Clear written answers save time and reveal quickly which vendors have thought about schools.
Learning and safety
- Can teachers stop the tutor from giving final answers?
- What topics does the tutor refuse, and can the school add more?
- What happens when a student mentions self-harm or abuse?
- What can teachers see, and what can parents see?
- How do you measure and correct wrong answers?
Data and accounts
- Will you sign our state's student data privacy agreement?
- Is student data used to train any model, yours or a provider's?
- Which AI model providers process student data, and where?
- How long do you keep prompts and responses?
- Which domains does the service use, so we can allow them?
Keep the answers with the signed agreement. They become the baseline for the yearly review and for any later complaint.
The last question matters more than it looks. A tutor that loads content from several domains will break in odd ways if only its main domain is allowed in the filter.
When an approved tutor needs a fresh review
Approval is not permanent. AI products change faster than most school software, so set triggers that send a tool back through the rubric.
Terms or privacy policy change
Any change to data use, age limits or model training. Ask the vendor to notify you in the agreement.
New features appear
Voice chat, image upload, open conversation or character modes added in an update.
The company changes
A sale, merger or change of model provider can change where student data goes.
An incident is reported
A harmful response, a data breach or repeated wrong answers reported by teachers.
Use spreads beyond the plan
The tool is used in grades or subjects it was not approved for.
The yearly review date
Even with no changes, recheck terms, domains and teacher feedback once a year.
Approving AI tutors with and without a hosted filter
A tutoring policy has to survive the arrival of new tools every month.
| Policy step | If you handle it manually | If the service handles it |
|---|---|---|
| Block AI by default | Add each AI site to a list yourself | Switch on AI / LLM Tools once per school |
| Allow only approved tutors | Exceptions scattered across device groups | One allow list per school, edited in the console |
| Hold new tools until reviewed | Notice them in logs weeks later | New domains arrive already sorted into categories |
| Different rules per age group | Duplicate configurations | Separate settings for each school or branch |
| Show the board what is allowed | Compile a document by hand | Activity reports and a policy record to export |
| Tutors on take-home laptops | Rules stop at the school gate | Private encrypted link keeps the policy on the device |
AI tutoring tools policy: FAQ
Should schools allow AI tutoring tools?
Many allow a small number of vetted tutors with controls. Data terms, age limits, safeguards and teacher visibility decide which ones.
What age can students use AI tutors?
It depends on the tool. Many consumer services set 13 as the minimum and need parent permission under 18; some school products serve younger students.
Can a school consent for students under 13 under COPPA?
Under FTC guidance, yes, when data is used only for the school's educational purpose. The 2025 amendments did not change that.
What should a data agreement with an AI tutoring vendor include?
Data collected, purpose limit, no model training or ads, retention and deletion, subprocessors, breach notice, location and audit rights.
Can federal funds pay for AI tutoring?
July 2025 guidance says AI uses such as high-impact tutoring can be allowable, if program and privacy rules are met. Confirm with your state.
How does the web filter support an AI tutoring policy?
It allows approved tutors for the right grades, blocks other AI tools by category and holds new tools until they are reviewed.
We price by school and device counts and send a quote within one business day. A pilot on a single site lets you test your approved-tutor list before it goes district-wide.
Talk to us about your district
Tell us which AI tutors your teachers want and which grades they teach. We will show how the filter can allow the approved ones and hold everything else for review.
Talk to us about your district See the AI tools category data