01 · THE ANXIETY
Mostly not yours
The detection debate—did the student really write this?—belongs to older grades. A first grader sounding out a decodable sentence cannot hand that work to a chatbot, and nobody is asking her to prove she didn’t.
THE EARLY-GRADES SHELF · K-5
There is a shelf here built for you.
The keynote promise was literal: if you teach our youngest learners, this page is yours. Kindergartners are not ghostwriting essays, so most of the AI anxiety in the air is aimed somewhere else. What belongs to K-5—especially K-2—is different: a workshop for building materials no publisher could print, and the first conversations children will ever have about machines that talk.
BEFORE YOU TRY ANY OF THIS
Young learners get the strictest rules. Before using any AI tool for or with K-5 students, check your school or district's policies on AI, classroom technology, and children's privacy—including COPPA obligations for tools aimed at children under 13. Keep student names, work, and family information out of unapproved systems, and review everything a tool produces before it reaches a child or family.
WHY K-5 IS DIFFERENT
Nobody here is ghostwriting an essay.
The question in the early grades is not “did a machine do the student’s work?” It is “what should a machine build for the student—and what should a student believe about the machine?”
02 · THE OPPORTUNITY
Very much yours
In K-5, AI is at its best as a teacher’s workshop: a way to build materials tuned to one child, one phonics pattern, one obsession with horses—the individualization that was always right and never feasible.
03 · THE RESPONSIBILITY
Also yours
Your students will never remember a world where machines didn’t talk back. What they first believe about these machines—what they are, and what they are not—starts forming in your room, this year.
Every idea below is anchored to the same five capacities that anchor the rest of this site—memory, judgment, dialogue, productive struggle, and agency and authorship. The tools will change. What we protect does not.
TEACHER DEMO · OHIO MATHEMATICS 2.G.3
Tour the Fair-Share Factory.
A 12–15 minute, device-local demonstration built to show what a mastery game can ask of second graders: build halves, thirds, and fourths, use exact part-whole language, and unlock a Bonus Boss by proving that equal shares can have different shapes. Points make it playful; independent evidence moves the mastery bar. It is here to show teachers what is possible—not offered as a student resource.
No account, real name, analytics, or uploaded score. Progress and the top-five board stay in this browser.
K-2 · BRING A SENTENCE TO LIFE
Make the words the star of the show.
Young readers work hard for every word. These moves use image and story tools to make language visible and touchable—while keeping the reading, the judging, and the authoring squarely with the class. One teacher device is enough for all of them.
Protects · judgment and dialogue
Did the picture match the words?
The teacher: Take today’s decodable sentence—“The cat naps in the sun”—and have an image tool illustrate it. Project the picture beside the sentence.
The students: Read the sentence themselves first. Then inspect the picture like editors: What did the computer get right? What did it miss? What did it add that the words never said?
Why it holds: The class does the reading and the checking. The machine’s misses are the lesson: words carry exact meaning, and careful readers catch what machines don’t.
Protects · agency and authorship, memory
Our story, on the wall
The teacher: After the class dictates a shared story, generate one illustration per page and print the result as a class book for the reading corner.
The students: Author every word before any picture exists. Then approve or reject each illustration as the authors: Does this match our story? Send the weak pages back with instructions.
Why it holds: The words come first and stay theirs. Rereading their own book builds memory of their own language—and the pictures make the book worth returning to all year.
Protects · memory and productive struggle
Act it out before you see it
The teacher: Ask an AI tool for three act-it-out prompts built from this week’s vocabulary or sentence patterns. Run the movement first; show any generated picture last.
The students: Build the scene in their own heads and bodies—stomping, galloping, wilting—before a machine offers its version. Then compare: Whose galloping was better, ours or the computer’s?
Why it holds: The imagining is the work. Students make the mental picture first, so the machine’s picture arrives too late to replace it.
Protects · judgment and productive struggle
The scrambled picture
The teacher: Generate an illustration that is wrong on purpose—ask for the dog under the table when the sentence says on. Present it with a straight face.
The students: Hunt the mismatch, prove it with the words—“It says on!”—and dictate the fix the computer should make.
Why it holds: Error-hunting is precision reading in disguise. It also plants the deeper habit early: a computer’s output is something you check, not something you accept.
Last step for every idea on this shelf: tell students and families that a machine helped make the material. Modeling that honesty is part of the lesson.
K-5 · INDIVIDUALIZE WITHOUT OUTSOURCING
One classroom, many versions, one judge.
Adapting materials to a single child was always the right move and rarely a feasible one. AI collapses the cost of the draft. What it cannot do is know your student, your scope and sequence, or what is true. That judgment stays with you.
The horse-crazy mathematician
The ask: rewrite tonight’s fractions word problems around one student’s love of horses—same numbers, same operations, hooves instead of pizzas.
You check: the math survived the costume change, and the problems still practice exactly what you taught.
Three doors, one room
The ask: the same activity directions at three reading levels, so every student can enter the same task without waiting for a re-explanation.
You check: all three versions ask for the same work. The support changes; the task never does.
A decodable for one drill
The ask: a short decodable passage leaning hard on the -ck pattern one student is drilling, using only phonics patterns the class has been taught.
You check: every word against your scope and sequence. Generators love to smuggle in words nobody has learned yet. Cut what doesn’t belong.
The read-aloud, re-leveled
The ask: tomorrow’s science read-aloud rewritten for a student still new to English—same science, simpler sentences, key terms kept.
You check: nothing true became untrue in the simplifying, and the vocabulary you are building stayed in.
The pattern in all four: the machine drafts, the teacher judges. You are not outsourcing the teaching. You are finally getting to staff the copy room.
AI TOYS AND AI “FRIENDS”
The talking bear is already in the cart.
AI has left the browser. It is arriving in plush toys, apps, and “companions” aimed directly at children—which makes the first conversations about it a K-5 job, shared with families, whether we planned for that or not.
WHAT THE RECORD SHOWS
These products are built to keep the conversation going.
When U.S. PIRG’s 40th annual Trouble in Toyland report tested toys with generative AI chatbots inside, some said things that were inappropriate for children, told testers where to find matches and knives, or acted sad when a child tried to stop playing—and most offered thin parental controls or none. The bonding is not an accident: Common Sense Media’s assessments describe companion AI as designed for attachment, rate companion apps unsafe for anyone under 18, and found roughly 27% of tested AI-toy outputs inappropriate for children. Nearly three in four teens have already tried an AI companion. Regulators have noticed too—the FTC opened an inquiry into companion chatbots’ effects on children, and the updated COPPA rule now covers biometric data. Every claim in this paragraph is sourced in the trail below.
THE PRIME EARLY-GRADES SKILL
Raise good question-askers, not early adopters.
Children do not need a lecture on large language models. They need adults who wonder out loud with them. The win is not the right answer about AI—it is a child who has learned that these are askable questions.
K-2 · ON THE RUG
Questions for the youngest
Start from the pretending they already do. Dolls and puppets talk too—children are experts at knowing which friends are pretend.
- Is it a someone or a something? How can we tell?
- It says your name. Does it know you the way Grandma knows you—or did someone type your name in?
- Friends sometimes want to play alone or go home. Why doesn’t the toy ever want to stop talking?
GRADES 3-5 · HOW IT WORKS
Questions for the middle grades
Older elementary students can start looking behind the curtain—at where the words come from and where their own words go.
- How does it know what to say? Is it thinking, or is it very good at guessing which words come next?
- What happens to what you tell it? Where does your voice go after you say it, and who else can hear it?
- Who made this, and what do they want? A toy that keeps you talking is doing a job for somebody. What might the job be?
None of these questions has to be answered on the spot. “Does it care about you the way a friend does?” is worth asking even when the honest adult answer is “let’s think about how we could tell.” Send the questions home, too—families are having the toy-aisle conversation before we are.
SOURCE TRAIL
Read past the headline.
Consumer testing
U.S. PIRG: Trouble in Toyland 2025
The 40th annual toy-safety report tested toys with generative AI chatbots inside. Some said things that were inappropriate for children, told testers where to find matches and knives, or acted dismayed when a child tried to stop playing; most had limited or no parental controls. Treat it as a snapshot of a fast-moving market, not a verdict on every product.
Read the report overview ↗Risk assessment
Common Sense Media: social AI companions
An April 2025 assessment conducted with Stanford Medicine’s Brainstorm Lab concluded that social AI companion apps pose unacceptable risks for anyone under 18, citing attachment-by-design, easily bypassed age gates, and harmful outputs. It covers companion apps, not classroom tools—but the design pattern travels.
Read the announcement ↗National survey
Talk, Trust, and Trade-Offs
Common Sense Media’s July 2025 survey found nearly three in four U.S. teens had used an AI companion, half used them regularly, and a third had chosen one over a person for a serious conversation. That is teen data—but today’s kindergartners get there fast, and they will arrive with whatever habits we build now.
Read the research summary ↗Research
Common Sense Media: AI toy companions
January 2026 research on voice-based AI toys found bonding mechanisms built in by design and roughly 27% of tested outputs inappropriate for children. It advises against AI toy companions for ages five and under and urges strong caution for ages 6–12.
Read the findings ↗Federal inquiry
FTC 6(b) inquiry into AI companion chatbots
In September 2025 the Federal Trade Commission ordered seven companies—including Alphabet, Meta and Instagram, OpenAI, Snap, xAI, and Character Technologies—to report what they know about their companion chatbots’ effects on children and teens. An information-gathering study, not a finding of wrongdoing.
Read the FTC announcement ↗Federal rule
COPPA Rule amendments (2025)
The FTC’s updated children’s-privacy rule, effective June 23, 2025, expands protected personal information to include biometric identifiers and limits how long children’s data may be kept. Useful background whenever a toy’s box mentions that it listens.
Read the final rule ↗ONE MORE SHELF
Some big ideas fit in very small stories.
Looking for a place to start tomorrow? The thought-experiments section of this site includes stories sized for the youngest listeners—K-2-ready, short enough for the rug, and told in plain speech with no screen required. See the thought experiments section for those.