A FIELD GUIDE FOR PARENTS & CAREGIVERS

Big possibilities.
Better boundaries.

Help your family navigate AI with curiosity,
confidence, and a little healthy caution.

Understand the tools. Know what happens to your data. Build habits that keep your child’s well-being at the center.

Explore the AI tool guide
Free to explore · No account needed · Sources included
A simple language for
thoughtful choices.
✓

Green · A positive useBounded, purposeful, with oversight

!

Amber · Pause & checkReview the settings and the context

×

Red · Set a boundaryAvoid the activity or age mismatch

01 / UNDERSTAND THE BASICS

You don’t need to be a tech expert.
Just an involved grown-up.

AI literacy means knowing what a tool can do, where it can fail, and when a human needs to step in.

✳

It predicts. It doesn’t know.

A chatbot generates a plausible response from patterns. A confident answer can still be wrong, biased, or entirely made up.

♡

Friendly isn’t the same as a friend.

A bot can sound caring without having human feelings or responsibility. Real relationships and qualified support come first.

⌑

A conversation leaves a trail.

Prompts, photos, voice clips and activity can reveal a lot. “Not used for training” does not mean “not stored or processed.”

Further reading: Pediatric guidance · Youth well-being guidance. Practical recommendations here are our editorial synthesis.

THE KNOWLEDGE BUCKET · A SIMPLE ANALOGY

How does AI actually get better?

Picture human knowledge and creations as a reservoir. Training pours selected examples into a model that learns patterns.

Human-created material pours into an AI learning bucketA tilted bucket of selected training material pours into an AI model. Move the slider to fill the model bucket and explore why equal additions can bring smaller gains.HumancreationsAI modellearned patternsBooks · text · images · codeSelected material—not everything
Illustrative improvementMore training material →Smaller extra gains
A few examplesMuch more material

The takeaway: More data can help, but equal additions may bring smaller gains when other factors stay fixed. Better examples, feedback, model design and computing power also matter. An AI searching the web is retrieving information, which is different from retraining its model.

Where the analogy stops: A model is not a bucket storing all human knowledge. This curve is illustrative, not measured performance, a prediction of progress, or proof of a fixed intelligence ceiling. It does not show that AI is becoming safer.

Research: neural scaling laws ↗ · Balancing data, model size and compute ↗
MEET THEM WHERE THEY ARE

Different ages. Different boundaries.

02 / KNOW THE TOOL

What’s behind the chat box?

The benefits, the blind spots, the data, and the controls.
Choose a tool for the full picture.

↻Research snapshot27 September 2026

These are editorial use recommendations, not safety certifications or scores. Age eligibility is a provider rule, not a developmental endorsement. Features vary by country, account, device and app version.

03 / RECOGNIZE THE RISKS

Know what to look out for.
Know what to do next.

The goal is a prepared family.
Open a concern for practical warning signs and a next step.

01Privacy & profilingSmall details can add up.+

School names, routines, faces and worries can build a detailed picture of your family. Personalization, advertising profiles and model training are distinct uses of data.

Watch for: requests for personal details, remembered secrets, unnecessary photo uploads or new app permissions.

Try this: replace names with fictional examples; review training, memory, history and permissions separately.

02Attachment & emotional relianceAlways available isn’t always healthy.+

Friend-like responses can feel deeply personal. A child may trust a bot with distress it cannot responsibly handle.

Watch for: lost sleep, withdrawal, distress when disconnected, or feeling responsible for a bot’s “feelings.” These signs invite a conversation, not a diagnosis.

Try this: ask what the chat provides, protect time with people, and seek a qualified professional for persistent distress. Avoid shaming or abrupt punishment.

03Wrong answers & hidden biasConvincing doesn’t mean correct.+

AI can invent quotations, citations and facts, or repeat unfair assumptions. Search-enabled answers can still misread sources.

Watch for: a claim without an accessible source, a stereotype, or medical and legal advice presented as certainty.

Try this: open the original source, check its date and author, and compare against a trusted reference. Ask a qualified person for consequential decisions.

04Deepfakes, bullying & exploitationA fake image can cause real harm.+

Voice clones and altered images can impersonate someone, humiliate a child or enable extortion. A child can be affected without using AI themselves.

Watch for: urgent money requests, threats, secret-keeping, sexual content or requests for more images.

Try this: verify urgent requests through a known phone number. For exploitation, support the child, stop contact and report through NCMEC. Do not forward or download sexual images of minors.

05Unsafe content & persuasive designFilters and age gates have limits.+

Sexual, violent, hateful or self-harm content can slip through safeguards. Notifications, streaks and companion features can pull a child into longer sessions.

Watch for: pressure to keep chatting, purchases, intense role-play, secrecy or recommended content becoming more extreme.

Try this: use accurate ages, turn off unnecessary notifications, set stopping points and keep devices out of bedrooms at night.

06Overreliance & actions on your behalfKeep thinking—and permissions—in your hands.+

Doing the work for a child can displace learning. An AI agent connected to email, files or shopping can also take actions with consequences.

Watch for: answers a child cannot explain, unsolicited links, or requests to connect broad accounts and approve purchases.

Try this: ask for hints instead of finished work. Keep sending, purchasing, account access and downloads behind adult review; never share passwords or one-time codes.

Grounding: AAP parent guidance · APA youth guidance · NCMEC exploitation guidance · FTC companion inquiry.

MAKE ROOM FOR THE GOOD

A little help.
A lot of human thinking.

With the right task and an adult nearby, AI can be a useful starting point for learning and creativity.

01

Learn without skipping the learning.

“Give me one hint, then ask me to explain my next step.”

02

Make something unmistakably yours.

“Suggest three story settings. I’ll create the characters.”

03

Practice a skill, together.

“Help us practice beginner Spanish about favorite foods.”

04

Make information easier to approach.

“Explain this public science article in simpler language.” Check that the meaning stayed the same.

04 / FOLLOW THE EVIDENCE

Promises. Lawsuits.
What’s actually been decided.

A company policy is a promise. A complaint is an allegation. A settlement is an agreement. A verdict is a decision. They tell different stories.

Legal entries show the latest milestone established by the linked source, not a live court docket. No listed lawsuit does not mean no risk. A past case does not prove current noncompliance.

05 / TURN KNOWLEDGE INTO HABITS

Your family.
Your boundaries.
Your AI plan.

A good plan is a conversation, not just a set of restrictions. Work through these steps together and revisit them when a tool or your child’s needs change.

A CONVERSATION STARTER
“Show me something interesting you’ve tried with AI. What surprised you? What didn’t feel right?”

Lead with curiosity. Make it safe to ask for help.

A healthier AI routine

0 of 7

Optional local storage only. No names, accounts, uploads or tracking scripts. Anyone using this browser profile may see saved progress.

BEYOND THE LIVING ROOM

Bring the conversation
to school, too.

A school-managed AI account may have different protections from a free consumer account. Ask before your child signs in.

What should I ask my child’s school?+

Which tools are approved? Who can read prompts and uploads? Are records used for training? When are they deleted? How are harmful responses handled? What non-AI alternative is available? Ask for the provider agreement and the school’s privacy notice.

How do we keep homework honest?+

Follow the teacher’s AI policy. Agree whether hints, brainstorming or proofreading are allowed, disclose AI assistance, and ask your child to explain the final work in their own words. Keep a record of drafts when useful.

What rights do parents have over data?+

In the U.S., COPPA generally gives parents rights concerning information collected online from children under 13 by covered services, including review and deletion. Coverage and school authorization have limits; it does not certify a product as safe. Read the FTC parent guide. Other jurisdictions have different rules.

What about AI in feeds, games and smart toys?+

AI is more than a chat window. Check recommendations, in-game companions, microphones, cameras, purchasing and app permissions. Ask what is processed on the device versus in the cloud. Apply the same privacy, time and trusted-adult boundaries.

WHEN SOMETHING DOESN’T FEEL RIGHT

Pause the chat. Connect with a person.

“You’re not in trouble. I’m glad you told me. We’ll work through this together.”

Concerning content or behavior

Stop the interaction, stay calm, and use the platform’s report tool. Note the date, account and relevant link. Contact a trusted teacher or qualified professional if distress persists.

Image abuse or exploitation

Don’t pay, send more images or forward sexual content. Save account names and URLs without downloading abusive images. Report to NCMEC’s CyberTipline ↗. Take It Down may help on participating services; don’t download an image to use it.

Immediate danger or crisis

If someone is in immediate danger, contact local emergency services. In the U.S., call or text 988 for crisis support. Outside the U.S., find local support through Find A Helpline ↗.

AI LITERACY & GOVERNANCE