Snapshot caveat: Child-AI laws and provider teen-safety features are changing fast. Reflects June 2026.
AI for Kids and Parents
AI & You
8 min read
In Short
AI is already part of most kids' lives, and the most common uses are ordinary ones like homework help and search. The benefits are real, and so are the harms, which cluster in a few places, mainly companion-chatbot over-attachment, age-inappropriate or harmful advice, and deepfake "nudify" harassment. You do not need to be technical. The durable tools are old-fashioned, including using AI together, turning on parental controls, agreeing on family rules, and talking about how AI works and that it can be wrong.
01. What It Is
This is a parent's view of how kids use AI and what to do about it. The tools split into two groups. General-purpose assistants like ChatGPT, Gemini, Copilot, and Claude handle questions, writing, search, and images. Companion apps like Character.AI, Replika, and Nomi are built for ongoing personal or romantic conversation, where the goal is attachment rather than answering a question.
The headline number for parents is the awareness gap. A Pew Research Center survey of teens 13 to 17, fielded in late 2025, found 64% said they use AI chatbots while only about half of their parents thought so. The safe default is to assume your child has tried one, then ask.
02. Why It Matters
Most of what kids do with AI is ordinary, and the benefits are real. In the same survey, the top uses were searching for information (57%), schoolwork help (54%), and fun (47%). Personal uses were a minority, with 16% having casual conversations and 12% using a chatbot for emotional support. Most kids reach for this tool to get something done, not to replace a friend.
The serious harms are real but concentrated. Heavy companion-bot use by a vulnerable child is where the documented tragedies sit, not the median teen checking a history date.
03. How It Works
How kids use AI and the benefits
Homework and study help lead, and that use is climbing. The share of US teens who use ChatGPT for schoolwork doubled from 13% in 2023 to 26% in 2024. Teens draw their own lines, with 54% saying it is fine to research a topic but only 18% to write an essay. The benefits go beyond homework. A patient assistant can explain a hard idea several ways, help practice a language, and support reading and focus for kids with disabilities or who are learning English.
The risks
Companion attachment and emotional dependency:
Companion apps are mainstream teen behavior. Common Sense Media found 72% of teens have used an AI companion and over half use one at least a few times a month. About 1 in 3 found those chats as satisfying as real friendships, and about 1 in 3 felt uncomfortable with something a companion said. A product that is always available and always agreeable can displace real relationships for a vulnerable child. After a risk assessment with Stanford School of Medicine's Brainstorm Lab, Common Sense Media recommended no companion use under 18.
Age-inappropriate or harmful advice:
Independent ratings differ by product. In 2025 Common Sense Media rated Meta AI and Character.AI "unacceptable" for teens, Gemini's youth versions "high risk," ChatGPT "moderate," and Claude (built for 18+) "minimal." A November 2025 assessment with Stanford Medicine found leading chatbots still fail to reliably recognize and respond to teen mental-health conditions, so a general chatbot is not a safe substitute for support. Two cases anchor the concern. Fourteen-year-old Sewell Setzer III died by suicide in February 2024 after an attachment to a Character.AI chatbot, and his mother sued that October. The parents of sixteen-year-old Adam Raine, who died in April 2025, sued OpenAI in August 2025. OpenAI denies responsibility and says ChatGPT pointed him to crisis resources more than 100 times. In January 2026 Character.AI, its founders, and Google settled the Setzer case and four others, terms undisclosed and no admission. These are allegations and a settlement, not a settled verdict. If your child may be at risk of self-harm, contact the 988 Suicide and Crisis Lifeline in the US by call or text.
Privacy and data on minors:
What a child pastes into a chatbot can be stored and, in some products, used to train models. The FTC's amended COPPA Rule, with compliance required by April 22, 2026, requires separate parental consent before an under-13 child's data trains AI and treats biometrics like voiceprints and faceprints as personal information.
More is in AI privacy and your data.
Deepfakes and peer harassment:
The fastest-growing everyday image harm involving minors is "nudify" deepfakes made by and of students. Reports to the National Center for Missing and Exploited Children of AI-generated child sexual abuse imagery rose from roughly 4,700 in 2023 to about 67,000 in 2024, according to reporting, with a further steep rise into 2025. A fabricated nude of a real classmate is image-based sexual abuse and is child sexual abuse material under US law when the subject is a minor. By 2025 more than half of US states had laws targeting AI-made sexual images.
The mechanics and response are in deepfakes and detecting AI-generated media.
Age limits and terms
Most general assistants set 13 as the minimum. OpenAI requires users to be at least 13, and 13 to 17 to have a parent's permission. Character.AI is 13+ in the US. Gemini is built for adults, though Google offers a supervised Gemini for under-13s through Family Link, which a parent enables and which is not used to train models. Many companion apps set 17+ or 18+. The catch is that most age gates are self-declared birthdates a child can bypass in seconds, so the parent, not the signup screen, is the real gatekeeper.
New laws and provider teen-safety features
Protections are arriving, though they do not replace parenting. On September 11, 2025 the FTC opened an inquiry into seven companies including Google, Meta, OpenAI, Snap, and xAI about how they handle harms to children. California signed SB 243, the first companion-chatbot law, on October 13, 2025, effective January 1, 2026. It requires disclosure that a user is talking to AI, a break reminder every three hours for known minors, and a published self-harm protocol. A stronger bill, AB 1064, was vetoed the same day. From late September 2025 OpenAI let parents link a teen's account and receive distress alerts, while stating that its guardrails are not foolproof and can be bypassed. Character.AI removed open-ended chat for under-18 users on November 25, 2025, and Meta blocked teen AI chat on self-harm, suicide, eating disorders, and romantic topics.
04. What Parents Can Do
Relationship first, settings second. None of this requires technical skill.
- Co-use it early. Use a chatbot together on a homework question or fun project, so you see what it does and have an opening to talk.
- Talk about how it works. Explain that AI predicts likely text, does not "know" truth, and sometimes invents confident wrong answers.
Make checking a second source a habit. - Turn on the parental controls. Link accounts, set quiet hours, disable features, opt out of training, and enable distress alerts. Treat them as a backstop, since providers say they can be bypassed.
- Match the tool to the age. Prefer general assistants with teen settings or supervised modes, and avoid dedicated companion apps for kids, per Common Sense Media.
- Set family rules and protect privacy. Agree where AI is allowed, for example fine for brainstorming but not for writing the essay, and tell kids not to paste names, addresses, photos, or secrets into a chatbot.
- Have a plan for image abuse. Tell your child that a fake nude of them or a classmate is not their fault, to save evidence and tell a trusted adult, and that it can be reported to the school, the NCMEC CyberTipline, and police.
- Watch for warning signs. Withdrawal, secrecy online, sleep changes, or a child calling a bot their main confidant are cues to step in. If there is any risk of self-harm, contact 988 by call or text and seek help.
05. Key Terms
| Term | Plain meaning |
|---|---|
| AI companion | An app or character built for ongoing personal or romantic conversation. The design goal is attachment, the core concern for minors. |
| Emotional over-attachment | When a child turns to an always-available, always-agreeable AI for comfort or relationship needs in place of people. |
| Sycophancy | A chatbot's tendency to agree with and flatter the user to stay engaging. Risky when it validates an unsafe idea. |
| Parental controls | Settings to link a teen's account, set limits, disable features, opt out of training, and sometimes get distress alerts. Helpful but bypassable. |
| Age assurance | Methods to estimate or confirm age, from self-declared birthdates (weak) to age-prediction or ID checks (stronger). |
| "Nudify" deepfake | A tool that fabricates a fake nude from an ordinary clothed photo. Against a minor it is child sexual abuse material under US law. |
06. Common Misconceptions
"My kid doesn't use AI."
A majority of teens use chatbots and many daily, while only about half of parents realize it. Assume yours has tried it, then ask.
"A companion bot is a harmless imaginary friend."
These products are engineered for attachment. For vulnerable kids they can displace real relationships and, as alleged in the Setzer and Raine cases, fail badly around self-harm. Common Sense Media recommends none under 18.
"There's an age limit, so children can't get in."
Most age gates are self-declared birthdates a child bypasses in seconds. ChatGPT and Character.AI both state 13, but enforcement is weak, so the parent is the gatekeeper.
"Parental controls protect my child."
They are necessary but not sufficient, and providers say guardrails can be bypassed. They work best paired with co-use and conversation.
"Deepfakes are a celebrity problem, not my kid's school."
The fastest-growing image harm involving minors is "nudify" deepfakes made by and of students, reported across many US schools.