k-12 · principals · buying-guide · EdTech-procurement · seo · ai-tutoring
K–12 AI Tutoring: A Principal's Buying Guide for 2026
AI tutoring is in every EdWeek headline. Here's how to tell what's real, what's hype, and what will actually help your teachers.
K–12 AI Tutoring: A Principal's Buying Guide for 2026
Audience: US K–12 principals, assistant principals, instructional coaches Hook: AI tutoring is in every EdWeek headline and every vendor pitch. Here's how to tell what's real, what's hype, and what will actually help your teachers. Tone: Professional peer-to-peer. Like a fellow principal who's been through three RFPs. Word count target: 1,800–2,200 SEO keywords: K-12 AI tutor buying guide, AI tutoring for schools, principal AI tools, EdTech procurement AI tutor, AI tutoring ROI schools
Why This Guide Exists
Every EdTech vendor has an "AI tutor" now. ChatGPT launched in classrooms and suddenly every platform got a chatbot bolted on. As a principal, you don't have time to demo 40 tools. You need a way to separate signal from noise.
This guide gives you the 8 questions to ask any AI tutoring vendor, the 4 red flags that mean "walk away," and the rubric I use when I'm evaluating these tools.
I've been through three AI tutoring procurements in the last 18 months. Here's what I wish someone had told me before the first one.
The 8 Questions to Ask Any AI Tutor Vendor
1. "What's under the hood?"
If they say "AI" without specifying the model, pedagogy layer, or training data — that's a red flag. A real AI tutor has:
- A structured pedagogy layer (worked examples, retrieval practice, spaced repetition)
- A model that respects student misconceptions (not just generates text)
- Guardrails for off-task questions
If they can't show you this in a 30-minute demo, they don't have it.
2. "Show me what it looks like when a student is stuck."
Anyone can demo the happy path: kid asks question, AI gives answer, kid moves on. The real test: what happens when the kid is stuck?
A good AI tutor:
- Recognizes confusion before the student gives up
- Offers a hint, not the answer
- Tracks the misconception and comes back to it later
- Notifies the teacher
A bad AI tutor just keeps generating more text.
3. "What does my teacher actually see?"
Teachers won't use a tool that buries insights in a dashboard. Ask:
- How many clicks to get to "which students are struggling right now"?
- Can a teacher see a student's last 5 minutes in under 10 seconds?
- Does the AI surface what to reteach tomorrow, or just data?
If the demo requires a tutorial, your teachers won't adopt it.
4. "How do you handle IEPs and 504s?"
This is non-negotiable. The AI must:
- Follow accommodation settings (extended time, read-aloud, simplified language)
- Not collect data outside the accommodation envelope
- Log interactions for IEP review
If the vendor says "we're FERPA compliant" but can't talk about IEPs specifically, they haven't thought about special education.
5. "What's your data privacy story?"
You need to know:
- Where is student data stored?
- Do you train your models on student work?
- What happens if we leave — can we export everything?
- Who's liable if there's a breach?
If they can't answer in writing, don't sign.
6. "How do you measure outcomes?"
Vague "improves learning" claims are useless. Look for:
- Specific metrics (NWEA, MAP, iReady correlation, custom pre/post)
- Published research or district pilot data
- Willingness to do a 90-day pilot with measurable goals
If they won't commit to measurable outcomes, they don't believe in their own product.
7. "What's the pricing model, really?"
Per-student? Per-school? Per-teacher? Site license? Ask:
- Are there hidden costs (training, support, integration)?
- Does the price change if we add a grade level?
- What happens if usage spikes?
Get pricing in writing for 3 scenarios: pilot, single school, district.
8. "What happens if it doesn't work?"
Every EdTool rep promises the moon. Ask:
- Can we exit cleanly after a pilot?
- Will you help us migrate data out?
- Is there a money-back guarantee?
If the answer to any of these is "we'll have to check," walk away.
The 4 Red Flags
🚩 "We use ChatGPT under the hood." ChatGPT is a general-purpose model. It's not a tutor. A real AI tutor wraps a tuned model in structured pedagogy. ChatGPT alone will hallucinate, confuse students, and waste teacher time.
🚩 "Setup takes 2-3 weeks." If onboarding is more complex than setting up a Google Classroom roster, your teachers won't finish it. Look for: Clever/ClassLink roster sync, Google Classroom import, single sign-on.
🚩 "We replace teachers." AI tutoring that claims to replace teachers is lying to you. AI handles practice and feedback. Teachers handle relationships, motivation, classroom culture. If a vendor tells you otherwise, they're selling you something.
🚩 "We don't share student data with anyone." Great, now show me the contract. Verbal privacy promises don't protect your district. Get it in writing.
The 4-Week Pilot Framework
If you're evaluating AI tutors, run a 4-week pilot. Here's the structure:
Week 1: Setup
- Roster sync
- Teacher onboarding (2 hours max)
- Pick 2-3 classes to pilot
- Set baseline metrics (pre-test, usage target)
Week 2-3: Run
- Teachers use it daily
- Track: usage, time-on-task, completion rate, teacher NPS
- Weekly check-in with vendor
Week 4: Evaluate
- Compare baseline to post-test
- Survey teachers (5 questions max)
- Decide: expand, switch, or stop
If the vendor won't commit to this framework, they don't believe in their product.
The Buying Decision Rubric
Score each vendor on 100 points:
| Category | Weight |
|---|---|
| Pedagogy quality (model + structure) | 25 |
| Teacher experience (UX + insights) | 25 |
| Student experience (engagement + outcomes) | 20 |
| Privacy + compliance (FERPA, COPPA, IEPs) | 15 |
| Pricing + contract terms | 10 |
| Support + onboarding | 5 |
Any vendor scoring below 70 isn't worth a pilot. Below 50, walk away.
About Grademy
Grademy is built specifically for K–12. Not college, not adult learners. K–12 only. We work with public, charter, and private schools in the US.
- Pedagogy: Tuned model wrapped in retrieval practice, worked examples, and spaced repetition.
- Teacher experience: Daily digest of "who's stuck on what" — not a dashboard you have to log into.
- Student experience: Reading level adaptation, language translation, IEP accommodation honoring.
- Privacy: FERPA + COPPA. We don't sell data. We don't train on student work.
- Pricing: $8/student/month, billed to the school. Teachers free. 30-day free trial, no credit card.
Want to run a pilot? Email pilots@grademy.work.
FAQ
Q: How is Grademy different from Khan Academy or IXL? A: Khan and IXL are content libraries. Grademy is a 1-on-1 tutor that adapts to each student in real time. It's the difference between a textbook and a private tutor.
Q: Does Grademy work for students without home internet? A: Yes. We support offline mode for school-issued Chromebooks and iPads.
Q: What about state testing alignment? A: Grademy aligns to Common Core, NGSS, and state-specific standards (TX, CA, NY, FL). Tell us your state and we'll send the alignment doc.
Q: How long does setup actually take? A: 15 minutes for a single teacher. 1-2 hours for a school. Roster sync with Google Classroom, Clever, or ClassLink.
Related reading
- FERPA + AI Tutors: K-12 Buying Guide — what every principal needs to know about FERPA before signing
- 9 Questions Before Letting an AI Tutor Near Your Classroom — the teacher-side vendor screen
- AI Tutoring in California — SB 942 / LCFF / SBAC — Golden State procurement overlay
- AI Tutoring + HB 1416 — Texas Teacher's Guide — Lone Star State procurement overlay
- FERPA Compliant AI Tutor: Plain-English Guide — the privacy pillar that pairs with this guide
- How 5 US Districts Cut Tutoring Costs 40% With AI — the case-study pillar with district procurement intel
Related reading
- Principal buying guide for AI tutoring — for AP US History: the full 22-week APUSH playbook (DBQ + LEQ + SAQ + MCQ) that turns a 3 into a 5
Related reading
- K-12 AI tutoring principal buying guide 2026 — for AP Biology: the 8 units, 4 science practices, 6 FRQ types, 12 lab skills, and 22-week workflow that lifts a 3 to a 5 on the May 2027 exam
Related reading
- K-12 AI tutoring principal buying guide 2026 — for AP Chemistry: the 9 units, 6 science practices, 4 FRQ types, 16 required labs, 5 math routines, and 22-week workflow that lifts a 3 to a 5 on the May 2027 exam
Related reading
- K-12 AI tutoring principal buying guide 2026 — for AP Calculus AB: the 8 units, 6 FRQ types, 3 calculator-active + 3 calculator-inactive question types, 21 theorems and definitions, and 22-week workflow that lifts a 3 to a 5 on the May 2027 exam
Related reading
- AP Statistics AI tutor playbook 2026 — for AP Statistics: the 9 units, 6 FRQ types, 3 calculator-active + 3 calculator-inactive question types, 24 inference procedures, and 22-week workflow that lifts a 3 to a 5 on the May 2027 exam