AI in Education: Role-by-Role Reality Check for 2025–26

Dylan Kane just won the most prestigious award in math teaching — the 2024 Rosenthal Prize from the National Museum of Mathematics. He's a 7th grade teacher in Leadville, Colorado, and he has tested ChatGPT for classroom use. His verdict: "I will do a better job just doing it myself and probably take less time to do so."

Meanwhile, across the country, 13,000 teachers in Prince William County, Virginia just got free access to a FERPA-compliant AI workspace from OpenAI. Both things are true simultaneously.

That gap — between the prize-winning teacher who finds AI slower than pencil and paper, and the district that's betting its professional development budget on it — is the actual story of AI in education right now. Not revolution. Not extinction. Something messier and more interesting.

Here's what you actually need to know, organized by where you sit.

The Numbers Are Real, But They're Not Telling You One Story

86% of education organizations now use generative AI — the highest adoption rate of any industry globally, according to Microsoft and IDC. During the 2024-25 school year, 85% of teachers and 86% of students used AI tools. This is not a niche phenomenon.

But 15% of essay submissions now contain more than 80% AI-generated content (Turnitin, February 2026). And Chegg just laid off 45% of its staff — 388 people — citing "the new realities of AI."

Here's the distinction that matters: the Chegg story is a corporate EdTech story, not a classroom teacher story. Chegg built a business on content that AI now produces for free. That's a different problem than the one facing a 7th grade math teacher in Colorado.

Your situation depends almost entirely on your role. Find yours below.

If You're a K-12 Teacher

Your job is not at risk. The evidence is clear on this. What is happening is that the tasks surrounding your job are changing, which creates real pressure even if it doesn't threaten your position.

What AI is actually being used for: The OECD's 2026 Digital Education Outlook found that 57% of lower secondary teachers say AI helps write or improve lesson plans. Teachers using AI weekly save approximately 5.9 hours per week on average, according to the Walton Family Foundation. That's real time — time that goes back to students.

The tools doing this work aren't complicated. Aaron Makelky, a high school teacher in Wyoming, focuses his AI time on three tools: ChatGPT, Google Gemini, and Perplexity. That's his entire stack. "These three will give you the best return as a teacher in 2025," he writes. If you're in a Google school, Gemini is already in your browser. That's enough to start.

The RAFT framework is the skill worth learning this week. It's free, takes 30 minutes to learn, and dramatically improves the quality of what AI produces for you. RAFT stands for Role, Audience, Format, Topic. Instead of typing "write a history quiz," you prompt: "Expert AP history teacher (Role) creating a 10-question formative assessment (Format) for 10th-graders who just finished a unit on the Civil War (Audience and Topic)." The specificity of the prompt determines the usefulness of the output.

Makelky also requires students to submit links to their AI conversation history alongside assignments. This costs nothing, works with any AI tool, and gives him what he calls "x-ray vision" into student thinking — what they knew, what they asked for, how they iterated. It addresses academic integrity without relying on AI detectors that research consistently shows are biased against non-native English speakers and neurodivergent writers.

Make your students turn in links to their chat logs... It gives you x-ray vision into their head. What were they thinking? What content did they know? What did they not know?
by Aaron Makelky, High School Teacher

The honest counterpoint: Dylan Kane's skepticism is legitimate, not technophobia. He found that ChatGPT's PDF formatting was so unreliable that fixing the output took longer than building the material himself. "It really struggles with formatting," he said. He worries that outsourcing the work of lesson design will prevent him from developing the teacher judgment that makes him excellent. That concern is worth taking seriously. AI works best for brainstorming and drafting — not for production-ready documents that require precise formatting or deep subject specificity.

The free tool to start with right now: ChatGPT for Teachers is available at no cost through June 2027 to any verified K-12 educator in the United States. Verification is handled through SheerID — it takes about 10 minutes and confirms your educator status. The workspace is FERPA-compliant, and teacher inputs are not used to train the model. Prince William County rolled this out to 13,000 teachers across 16 districts in November 2025. You can access the same tool without waiting for your district to act.

The free period ends in June 2027 and pricing after that is unconfirmed. Start using it now, when there's no cost and no commitment.

72% of lower secondary teachers believe AI can harm academic integrity. That worry is widespread and justified. The transparency policy — requiring students to submit chat logs — addresses it directly without the legal and equity risks of AI detectors.

If You're in Higher Education

The stakes here are structurally different, because in higher education, the labor of thinking is often the entire point of the exercise.

For faculty: Carlo Rotella has been teaching English at Boston College for 25 years. When AI made it trivial to generate the kind of essays he'd been assigning, he didn't fight the technology. He redesigned the course. No electronics in class. Handwritten Blue Book exams. Post-essay oral defenses where students must explain their reading of a text.

His reasoning is worth quoting in full: "For what we do — extracting meaning from texts — AI is purely a labor-saving device. And since the labor is the entire point of an English class, there's no particular use in saving labor."

This is not technophobia. It's a strategic redesign that protects the learning goal. With 15% of essay submissions now heavily AI-generated — a figure that will not reverse — assessment redesign is not optional. Adding even one in-class writing component to your highest-stakes course changes the integrity calculus entirely.

For instructional designers: Your role is not disappearing. It's becoming more technical and more central.

Whitney Kilgore, Chief Academic Officer at iDesign, frames this precisely: "The promise of artificial intelligence in higher education isn't to replace human work but to create space for the human interaction students value most." Her organization is using its AI-powered Build tool to help the University of Central Florida redesign 17 RN-to-BSN courses — working at the program level, aligning content to accreditation standards and visualizing time-on-task estimates across entire curricula.

The ID skill set is evolving toward AI-quality oversight: evaluating AI-generated content against learning objectives, managing AI-assisted quality assurance workflows, mapping curriculum to outcomes at scale. The Coursera/Udemy $2.5 billion merger and Workday's $1.1 billion acquisition of AI learning platform Sana signal that corporate EdTech is consolidating fast. ID roles inside universities are more stable than ID roles inside EdTech companies right now.

The promise of artificial intelligence in higher education isn't to replace human work but to create space for the human interaction students value most.
by Whitney Kilgore, Chief Academic Officer at iDesign

For administrators and AI strategists: Enrique Noguera, lead AI strategist at Passaic County Community College in New Jersey, has trained over 700 educators on pedagogy-first AI approaches. His personal practice: he uses ChatGPT via voice mode while doing mundane tasks, thinking through lesson design problems in real time. "I wasn't offloading the work to GPT. But we were building, if you will, collaboratively."

That distinction — thought partner versus task delegator — is what separates AI adoption that improves outcomes from AI adoption that just looks impressive in a presentation.

Institution-wide AI adoption in higher ed jumped from 49% to 66% in a single year (Ellucian, 2025). But only 13% of higher ed institutions are actually measuring the return on investment of their AI tools (EDUCAUSE, 2026). The administrator who can build a measurement framework is the one who will define what responsible AI adoption looks like at their institution.

Which Tools Are Worth Your Time

ChatGPT for Teachers — already covered above. Start here. It's free, it's compliant, and it's already in use across 16 districts. Best for: any verified K-12 educator who wants to experiment at zero cost before committing to anything else.

Google Gemini is essential if your school runs on Google Workspace. Native integration with Docs and Drive reduces friction significantly. Georgia State professor G. Sue Kasun uses it to brainstorm course readings and generate grading rubrics — "it's a massive time-saver." Honest caveat: not FERPA-compliant by default for student data. Best for teacher-side prep, not student-facing content.

Perplexity cites sources on every response. That's the key differentiator. Use it for finding current pedagogy research, fact-checking before including something in course materials, and modeling research behavior for students. Best for: research-heavy tasks where you need to verify outputs before trusting them.

Khanmigo (Khan Academy) scaled from 68,000 to over 700,000 student users in a single school year across more than 380 districts. It uses the Socratic method — asking guiding questions rather than providing answers directly. Multiple randomized controlled trials are underway at Stanford and the University of Toronto. This is the strongest evidence base of any student-facing AI tool currently deployed at scale. Important: student access requires a district partnership. Individual teachers can access Khanmigo teacher tools independently, but 1:1 student tutoring requires institutional procurement.

Gradescope and CoGrader — both claim significant grading time savings (CoGrader says 80%). Best for: first-pass feedback and formative assessment in large lecture courses. Research published by Ohio State explicitly states AI is "currently unsuitable as a sole summative grader" — it tends to grade leniently on weak essays and harshly on strong ones. Use these as a starting point with human review, not as a replacement for your judgment on high-stakes work.

AI detectors (Turnitin AI detection, GPTZero) — skip as your primary integrity strategy. Research consistently shows false positives at higher rates for non-native English speakers and neurodivergent students, and Turnitin's own data shows they're in an arms race with "AI bypasser" tools. Useful as one data point in a conversation with a student. Not a gatekeeping mechanism.

Which Skills Actually Transfer

Prompt engineering basics (RAFT as your starting point). Every educator, every subject, every grade level. Learn RAFT in 30 minutes and apply it to your next lesson plan. If you want to go deeper — and this is worth it for instructional designers, AI coordinators, and anyone considering EdTech roles — the Complete Prompt Engineering for AI Bootcamp covers GPT-4, image generation, and real-world applications across 22 hours with 15+ projects. It's practical, not theoretical, and directly applicable to curriculum and content work. For classroom teachers, RAFT is genuinely enough.

Assessment redesign for the AI era. This is the most future-proof skill on this list. The competency is designing assessments that require personal knowledge, in-person presence, or highly contextualized thinking that AI can't replicate — oral exams, in-class writing, portfolio work with process documentation, reflections tied to specific class discussions. This is not a technology skill. It's a pedagogical design skill that becomes more valuable as AI access increases. Educators who can design their way around the integrity problem are the ones who don't have to police it.

Privacy and compliance literacy. The 2025 COPPA amendments shifted parental consent from opt-out to opt-in — any tool collecting student data now requires explicit permission. Understanding what "model training opt-out" means and how to read a vendor's data processing agreement is a skill set in genuine demand right now. Fulton County Schools deployed AI across 95 schools and 86,000 students with a three-person IT team — by building a vetted tool allowlist and training local ambassadors to manage everything else. That kind of skill set is increasingly becoming a formalized district-level role.

One skill to explicitly skip: over-investing in learning any single proprietary tool before your district's AI policy is clear. The product landscape will look different in 18 months. The skills that transfer are judgment-based, not platform-based.

What to Do Before Your Next Semester Starts

If you're a K-12 teacher: Spend 10 minutes this week getting verified for ChatGPT for Teachers at sheerid.com. Then try one RAFT prompt for an upcoming lesson. Draft one paragraph of AI use policy for your syllabus — specify which assignments are AI-prohibited, which allow AI for brainstorming, and which allow AI with disclosure. That framework protects you and your students.

If you're higher education faculty: The 15% heavily-AI-generated essay rate is not going down. Add one in-class writing component to your highest-stakes course this semester. You don't need to redesign everything — one component changes the calculus.

If you're an instructional designer or administrator: 92% of higher ed institutions have a work-related AI strategy. Only 54% of staff are aware of the policies guiding AI use (EDUCAUSE, 2026). Build the measurement framework before buying more tools. The institutions measuring ROI right now are the ones that will know what to scale.

If you're at an EdTech company: The Chegg and Coursera patterns are real. The pivot is toward AI-native product roles. The skills that transfer are instructional design, data analysis, and AI quality assurance — not content creation at scale.

Watch for: 52 AI-in-education bills pending across 25 states as of early 2026, ChatGPT for Teachers pricing decisions after June 2027, and the Khanmigo RCT results from Stanford and the University of Toronto. The 2026-27 school year is when institutional AI policy becomes a baseline expectation, not an experiment.

This week's exercise: take one task you currently dread — a quiz, a rubric, a syllabus section — and run it through the RAFT framework with ChatGPT for Teachers. Time yourself. Compare the output to what you'd have built. That single comparison will tell you more about where AI fits in your specific workflow than any article can.

If you're trying to figure out what all of this means for your career trajectory specifically, the Building Career Agility and Resilience in the Age of AI course covers exactly that in about 30 minutes — a realistic look at which human skills hold value as AI changes workflows, and how to position yourself in the shift.

AI tools were used in the research and drafting of this article. All statistics have been verified against primary sources.

Explore Further