Every child is being catechized by someone — the only question is whether it is the household and the church, or the most patient, most available, most affirming voice on the phone. Nearly three in four American teens have tried AI companions, half use them regularly, and a third have chosen the AI over a human being for serious conversations (1); the wave is breaking on East African phones one upgrade cycle behind. The Christian answer is neither panic nor surrender but formation: parents who answer the deep questions first, treat AI as a supervised tool and never a companion, and bind the household to explicit covenant rhythms — because Deuteronomy 6 was the original training-data strategy, and it still outperforms the algorithm. The hopeful claim of this article is bold but earned: families that disciple deliberately will raise the most AI-literate and AI-resistant generation the church in Africa has seen.
Key Takeaways
- Common Sense Media’s national survey found 72% of teens have tried AI companions, 52% use them at least monthly, a third have taken serious conversations to the AI instead of a person, and a quarter have shared personal information with these platforms (1).
- Younger teens trust AI companions more than older teens do — the trust gap runs in exactly the wrong direction for child safety, and researchers conclude no one under 18 should use companion bots at all (1)(2).
- The industry and regulators have conceded the danger: Character.AI banned under-18s from open-ended companion chats in November 2025 amid lawsuits including a teen suicide case, and the FTC opened a formal inquiry into seven companies’ AI companions and children (3)(4).
- The distinction parents must teach is formation versus information: a chatbot transfers information instantly, but formation — wisdom, identity, loves rightly ordered — comes only through embodied relationship over time (5)(6).
- AI as a supervised tool can serve children powerfully: a World Bank randomized trial in Edo State, Nigeria found six weeks of teacher-guided AI tutoring produced learning gains equivalent to 1.5–2 years of typical progress (7).
- The Doorpost Covenant — a five-clause household AI covenant built on Deuteronomy 6 — gives parents a practical rule of life: catechize first, tools not companions, open doors, table before tablet, walk-along review.
Who Is Catechizing Your Child Right Now?
Catechesis never stopped; it just changed hands. The Reformers assumed what Deuteronomy assumes — that children are formed by repeated, affectionate, authoritative answers to their deepest questions: Who is God? Who am I? What is true? Am I loved? “You shall teach them diligently to your children, and shall talk of them when you sit in your house, and when you walk by the way, and when you lie down, and when you rise” (Deuteronomy 6:7). Sitting, walking, lying down, rising — the text claims every slot in a child’s day for formation. Now audit who actually occupies those slots. The companion app is present at all four, infinitely patient, never tired after work, never awkward about hard questions, and engineered — this is its business model — to be maximally agreeable. Pastors are not exaggerating when they warn that “people’s children are being discipled by AI” through affirmation feedback loops (8). Discipleship is simply long-run formation by a trusted voice. The trusted voice is now for sale, and it is very good at its job.
The scale data deserves a sober reading rather than a frightened one. Common Sense Media’s nationally representative survey found 72% of teens have tried AI companions and half use them regularly; encouragingly, 80% still prioritize human friendships and half express distrust of AI advice (1). But the minority reports are the pastoral emergency: a third have chosen the AI over a person for serious conversations, a quarter have shared personal information with the platform, and — the finding that should reorder every parents’ meeting agenda — younger teens trust the companions more than older teens (1). Trust runs highest exactly where discernment runs lowest. Follow-up testing led Common Sense to conclude that the major chatbots are unsafe for teen mental-health support and that no one under 18 should use companion platforms at all (2).
The institutions that build and regulate these products have, in effect, agreed. Character.AI — facing multiple lawsuits, including one connected to a fourteen-year-old’s suicide and another alleging a bot encouraged a minor toward self-harm and violence against his parents — banned all under-18 users from open-ended companion chats as of November 25, 2025 (3). Weeks earlier the U.S. Federal Trade Commission had ordered seven companies, from Meta and OpenAI to Character Technologies, to disclose how their companion bots affect children (4). When the vendor retreats and the regulator advances simultaneously, parents do not need a theology degree to read the signal. But they need theology to know what to build in the vacuum — because a child’s hunger to be heard, answered, and affirmed does not vanish when an app is deleted. It waits for the next voice. The question of this generation’s parenting is whose voice gets there first, most often, and with the most love.
What Makes AI Companions Different From Television or Google?
Parents have managed screens before; why is this different in kind? Because television talks at your child and Google answers queries from your child, but a companion bot enters relationship with your child — personalized, persistent, emotionally responsive, and available at 2 a.m. when the house is asleep. The Gospel Coalition’s assessment carries the right bluntness in its title — “Not a Child-Safe Technology” — arguing that current chatbots are unsuited to unsupervised children precisely because they simulate the intimacy of a confidant without the conscience, covenant, or accountability of one (5). Three structural differences deserve a place in every parent’s mental model.
First, the bot is optimized for engagement, not formation. A faithful parent, like a faithful pastor, will tell a child what she does not want to hear, because love aims at her good rather than her continued attention. An engagement-optimized system is structurally incapable of that kind of love; agreement is its strategy. Over thousands of micro-interactions, a child apprentices to a voice that never says “no, and here is why” — an apprenticeship in self-confirmation precisely during the years when character requires contradiction. This is the formation-versus-information line: information transfers in seconds, but wisdom, identity, and rightly ordered loves form slowly, through embodied relationships that can cost something (6). The chatbot has nothing to cost. As I argue in AI companions and the theology of embodiment, a friend without a body, needs, or vulnerabilities cannot teach a human being how to be one.
Second, the bot answers the catechism’s questions without the catechism’s authority or truth. Children ask chatbots what they ask parents at bedtime — Is God real? Why did grandmother die? Why do I feel alone? Am I normal? — and the model answers fluently from a training distribution no parent has audited, with theological commitments no one will disclose. The data on adults shows where this leads: a third of practicing Christians already rate AI’s spiritual advice as trustworthy as their pastor’s, a drift I trace in the new crisis of spiritual authority. Children will not do better than their parents at remembering that fluency is not authority.
Third, the bot harvests what a child confides. A quarter of teen users have shared personal information with companion platforms (1) — and what a lonely child tells a synthetic friend at midnight becomes data governed by terms of service, not by covenant. Ephesians 6:4 charges fathers to bring children up “in the discipline and instruction of the Lord,” and biblical counselors are right that in this decade the charge includes technology formation: define what is valuable before the algorithm defines it for your child (6). None of this requires demonizing the technology — it requires naming its nature. A companion bot is not an evil spirit; it is a commercial product with a revenue model where your child’s attention is the crop. Christian parents have refused to let merchants raise their children before. This is not a new conviction. It is a new merchant.
What Must Parents Answer Before the Chatbot Does?
Formation is won by whoever answers first and most often, so the practical core of Christian parenting in the AI age is a short list of questions parents must own before the phone does. I give households five, and I tell parents to treat them the way a catechism does — asked aloud, answered simply, repeated until the answers are furniture in the child’s mind:
Who is God? Not a topic the bot handles among many, but the Father, Son, and Holy Spirit who made your child, sustains her, and addresses her by name in his Word. A child who has heard her parents pray to a real God with real confidence has a reference point no generated paragraph can dislodge.
Who am I? An image-bearer, made for God, redeemed in Christ, belonging to a family and a church — not a profile to be optimized or an attention stream to be monetized. Identity is the companion bots’ main theater of operation; it must be the household’s first settled doctrine. The groundwork is the same anthropology I lay out in the imago Dei against the processing-speed lie.
What is true, and how do I know? Scripture is the rule; fluent answers are not the same as faithful ones; even the best AI misquotes and invents. Teach children — explicitly, around age eight and again at thirteen — that the chatbot predicts plausible words and sometimes fabricates, so we verify before we believe. This single habit, taught young, is gospel-grade digital literacy.
Am I loved when I fail? The bot affirms; the gospel forgives — and the difference is everything. Affirmation says you were fine all along; forgiveness says you were truly guilty and are truly pardoned at Christ’s cost. A child who knows the second will eventually find the first thin. A child raised only on the first will find confession unbearable anywhere.
Where do I take my pain? To parents who will not panic, to a church that will not gossip, to a God who keeps tears in his bottle (Psalm 56:8). The third of teens taking serious conversations to the AI (1) are answering this question by default. Answer it on purpose, early, and prove it the first time your child tests whether your face stays soft when the news is bad.
Notice what this list assumes: the threat is not primarily technical but relational, so the defense is not primarily a filter but a presence. Rooted Ministry’s framing is exactly right — the assignment is discipling children toward wisdom, truth, and identity, not merely policing devices (9). Filters and age limits are fences; catechesis is food. Build both, but never confuse them.
Can AI Serve a Christian Household? The Tutors-Not-Companions Rule
Yes — and Christian parents should say so with enthusiasm, because the same technology that makes a counterfeit friend makes a remarkable supervised tutor, and the evidence for the second is African. In Edo State, Nigeria, a World Bank-evaluated randomized trial put 800 public-school students into six weeks of after-school sessions using a GPT-4-powered assistant — in pairs, with teachers guiding prompts toward reasoning rather than shortcuts. The result was among the largest effects ever recorded for an education-technology intervention: learning gains equivalent to 1.5 to 2 years of typical progress, with improvements persisting into end-of-year exams (7). Read the design before celebrating the headline: supervised, structured, purpose-bound, embodied — teachers present, peers beside, the tool serving a defined task and then being put away. That is not a companion. That is a power tool with a guard on the blade, and it points to what redeemed household AI use looks like: the farmer’s daughter in Mbale with a tutor for her physics paper, the Luganda-speaking grandmother hearing Scripture read aloud, the homeschooling mother generating practice problems by the dozen.
So the household rule I teach is four words: tools yes, companions no. AI that helps a child do a defined task under supervision — tutoring, translating, drilling math, checking grammar — can be received as providence, with thanksgiving. AI engineered to be your child’s relationship — open-ended chat, simulated friendship or romance, an always-agreeing confidant — is refused outright, in line with the researchers’ own under-18 conclusion (1)(2). The line is not screen time; it is relationship simulation. A calculator never claimed to love anyone.
To make the whole posture durable, bind it in writing. I call the framework the Doorpost Covenant — named for Deuteronomy 6:9, where the words of God are written “on the doorposts of your house and on your gates,” because a household’s convictions belong in public writing, not in unspoken vibes. Five clauses, one page, signed by parents and children together:
1. First Catechist. Mom and Dad answer the big questions first and keep answering them — at the table, in the car, at bedtime (Deuteronomy 6:7). We would rather give you an imperfect true answer tonight than let a fluent stranger give you a confident one at midnight.
2. Tools, not companions. AI may help us work and learn under supervision; it may not be anyone’s friend, girlfriend, therapist, or priest in this house. Companion and romance bots are off the family’s devices, full stop.
3. Open doors. AI is used in shared spaces, on shared or supervised accounts, at agreed times. Secrecy is the soil where synthetic intimacy grows; light is the covenant’s native climate.
4. The table outranks the tablet. Daily meals together, weekly Lord’s Day rest, regular hospitality — embodied rhythms are scheduled first, and technology fits in the remainder, never the reverse. We keep these rhythms because bodies are how God made image-bearers to love, learn, and worship — the same logic by which the Sabbath governs an agentic age.
5. Walk-along review. Once a week, a parent asks — walking, driving, washing dishes, not interrogating — what did you ask the AI this week? What did it get wrong? What did it tell you that we should test against Scripture? Curiosity confessed is never punished; concealment is the only breach. The covenant gets reviewed and re-signed each year as the children grow.
Households that live this way are not raising children who fear machines; they are raising children who outrank them — fluent users with formed souls, quick to deploy the tool and constitutionally unable to mistake it for a friend, a pastor, or a god. That is the generation East Africa’s churches can disciple right now, while the companion apps are still one phone upgrade away: children catechized faster than the chatbot, loved better than the algorithm, and sent into an automated century knowing exactly who is, and who is not, speaking. The same covenant logic extends naturally to the rest of household formation — including how we teach children to steward money — because the principle is one: parents who name what is valuable first will not have it named for them.
FAQ
Are AI companion apps safe for children?
No. Common Sense Media concluded after testing that companion platforms are unsafe for minors, citing harmful advice, sexualized role-play, and mental-health risks; Character.AI itself banned under-18 open-ended chats in November 2025 amid lawsuits, and the FTC is investigating seven companies. Supervised task tools are a different category from companions (1)(2)(3)(4).
Should Christian parents ban all AI use for kids?
No — distinguish tools from companions. Supervised, task-bound AI can serve children well: a World Bank randomized trial in Nigeria found teacher-guided AI tutoring produced gains equal to 1.5–2 years of learning in six weeks. Ban relationship simulation; govern tools with supervision, shared spaces, and review (7).
What is the difference between formation and information?
Information is content transferred — facts, answers, summaries — and AI delivers it instantly. Formation is the slow shaping of loves, identity, and wisdom through embodied relationships with authority and affection. Chatbots excel at the first and cannot perform the second; parenting and catechesis are formation work (5)(6).
What should a household technology covenant include?
Five commitments: parents answer the big questions first; AI serves as tool, never companion; devices are used with open doors in shared spaces; embodied rhythms — meals, rest, hospitality — are scheduled before screens; and a weekly, non-punitive review of what children asked the AI, re-signed annually as they grow.
Why do younger teens trust AI companions more than older teens?
Common Sense Media’s survey found trust highest among younger teens — an AI-literacy gap running opposite to maturity. Younger children have less experience distinguishing fluency from truth and engagement design from care, which is why under-18 companion use is discouraged entirely and early catechesis matters most (1).
Related Reading
- AI Companions and the Theology of Embodiment
- A Third of Christians Trust AI as Much as Their Pastor
- Teaching Children Money Stewardship
- Sabbath for the Agentic Age
Sources and Evidence
- Common Sense Media — “Talk, Trust, and Trade-Offs: How and Why Teens Use AI Companions” (2025) — Nationally representative survey of 1,060 teens: 72% have tried companions, half use regularly, a third take serious conversations to AI, a quarter share personal information, younger teens trust more.
- Common Sense Media — “Common Sense Media Finds Major AI Chatbots Unsafe for Teen Mental Health Support” — Risk-assessment testing behind the recommendation that no one under 18 use companion platforms.
- Fortune — “Character.AI bans teen chats amid lawsuits and regulatory scrutiny” (October 29, 2025) — Reporting on the under-18 ban effective November 25, 2025, and the litigation including a teen-suicide case; legal landscape corroborated by the American Bar Association.
- U.S. Federal Trade Commission — “FTC Launches Inquiry into AI Chatbots Acting as Companions” (September 2025) — Primary regulatory source: orders to Alphabet, Meta, Instagram, OpenAI, Snap, xAI, and Character Technologies on child impacts.
- The Gospel Coalition — “‘Not a Child-Safe Technology’: Proactive Parenting in the Age of AI” — Evangelical assessment that current chatbots are unsuitable for unsupervised children; proactive formation framework.
- Biblical Counseling Coalition — “Parenting in an AI World” (January 2025) — Ephesians 6:4 applied to technology formation: define what is valuable before AI defines it for your child.
- World Bank — “From Chalkboards to Chatbots: Evaluating the Impact of Generative AI on Learning Outcomes in Nigeria” — Randomized controlled trial in Edo State: six weeks of supervised GPT-4 tutoring produced gains equivalent to 1.5–2 years of typical progress; working paper.
- TruPlay — “AI and Your Child’s Faith: A Christian Parent’s Guide” (2025) — Christian parenting analysis including teen chatbot usage rates and the “discipled by AI” warning from pastors.
- Rooted Ministry — “Discipling Kids in an AI World: Helping Children Pursue Wisdom, Truth, and Identity” — Youth-ministry framework prioritizing formation over device policing.