# Parent AI-Literacy Age Matrix

Companion to [AI literacy by age: what to expect from a 6-year-old to a 16-year-old](/articles/ai-literacy-by-age-for-parents).

These bands are starting points for a conversation with your specific child, not a developmental
assessment. When a band and a product's actual minimum age disagree, the stricter rule wins.

## The four bands, in full

### ~6: "It's a machine that guesses words. It can guess wrong."

- **Supervision:** Adult present and driving the keyboard or voice input; child watches and asks.
- **Privacy:** Adult's own account only. No child data, images, or names entered under any
  circumstance.
- **Verification:** Adult checks every factual claim aloud, in front of the child, so the checking
  itself is visible.
- **Creation:** Adult-led only — a story or picture made together, narrated as "let's see what it
  comes up with."
- **Prohibited:** No solo account. No images of the child. No conversation about people the child
  knows by name.
- **Sample phrase to use out loud:** "It's guessing what word comes next based on lots of books and
  websites it read. Sometimes its guess is wrong. Let's check."

### ~10: "It learned patterns from lots of text. It doesn't know things the way a person knows things."

- **Supervision:** Adult nearby, on a shared or supervised device, not necessarily reading every
  line.
- **Privacy:** No school records, home address, or photos of other children. First name only, and
  only if genuinely necessary for the task.
- **Verification:** Child names one way they would check a claim (search engine, textbook, ask an
  adult) before the adult confirms it.
- **Creation:** Child drafts with AI assistance; an adult reviews before anything is shared, printed,
  or submitted anywhere.
- **Prohibited:** No unsupervised account. No unsupervised image generation. No companion-style or
  chat-persona apps.
- **Sample phrase to use out loud:** "It doesn't know things — it predicts likely words. How would we
  check if this is actually true?"

### ~13: "It's a statistical system with training-data cutoffs and known failure modes I can name."

- **Supervision:** Own account possible where the product's terms allow it (see minimum ages below).
  Periodic check-ins about what they're using it for, not shoulder-surfing every session.
- **Privacy:** Understands that chat logs and account data are not a private diary, and that a
  provider or a future data breach could expose them.
- **Verification:** Independently cross-checks anything used for schoolwork, health, or a decision
  with real stakes.
- **Creation:** Independent creative and study use, with the household's disclosure rules
  (see the [family AI agreement template](/downloads/articles/family-ai-agreement-template.md))
  understood and followed.
- **Prohibited:** No companion-style "AI friend" products (see
  [do not give a child an AI friend](/articles/do-not-give-child-ai-friend)). No use for identity
  documents. No covert surveillance of siblings, friends, or partners. No bypassing a teacher's
  stated AI-use rules for an assignment.
- **Sample phrase to use out loud:** "You can use tools for school within the rules we and your
  teachers set. An AI that pretends to be your friend is still a product — that is off-limits."

### ~16: Can explain hallucination, training-data bias, and sycophancy in their own words

- **Supervision:** Occasional conversation about how they're using it, framed as interest rather than
  monitoring.
- **Privacy:** Manages their own account privacy settings; understands retention and deletion
  options for the specific product.
- **Verification:** Applies verification habits without being prompted to.
- **Creation:** Full independent use within the platform's own terms of service.
- **Prohibited:** Still no companion-style "AI friend" products. Same other legal and safety
  boundaries as an adult, plus whatever minimum-age terms the specific product still enforces at 16
  (several major products require 18).
- **Sample phrase to use out loud:** "I am not pretending I have this all figured out. I do want to
  know what you are using, where school rules sit, and how you would ask for help if something felt
  off."

## Real minimum ages — check and date-stamp this section yourself

Products change these terms. The dates below are when this table was last checked; re-verify before
you rely on it if more than a few months have passed.

| Product | Stated minimum age | Supervised path for younger users | Checked |
| --- | --- | --- | --- |
| OpenAI ChatGPT | 13, or local digital-consent age if higher; under-18 needs parent/guardian permission | Optional parent-teen account linking; automatic "teen experience" content safeguards | 2026-07-30 |
| Anthropic Claude | 18, no exceptions on the consumer product | None — Claude.ai has no supervised or teen mode | 2026-07-30 |
| Google Gemini Apps | Family Link supervised access from under 13 where locally permitted; unavailable in EEA/UK/Switzerland | Parent must explicitly enable per child in Family Link | 2026-07-30 |

- [ ] Re-checked OpenAI's terms on: ____________________
- [ ] Re-checked Anthropic's terms on: ____________________
- [ ] Re-checked Google Family Link's Gemini availability on: ____________________

## One-session literacy check

Before moving your child to the next band, run one supervised session using this script:

1. Say the mental-model sentence for their current band out loud, before opening the app.
2. Ask them to repeat it back in their own words after the session ends.
3. If they can restate it accurately, they are ready to try the next band's supervision level next
   time.
4. If they cannot yet, stay at the current level for a few more sessions rather than moving up on a
   fixed schedule.

Record what happened:

- Date: ____________________
- Band attempted: ____________________
- Could they restate the mental model? Yes / Not yet
- Next session's supervision level: ____________________

## Related reading

- [A child's first AI conversation](/articles/child-first-ai-conversation) — the facilitated script
  for the actual first supervised session.
- [Talking to teenagers about AI](/articles/talking-to-teenagers-about-ai) — the two-way conversation
  that replaces close supervision once a child has their own account.
- [What ChatGPT remembers, sees, and shares](/articles/privacy-101-what-chatgpt-remembers) — the
  underlying data-retention mechanics behind the privacy rule in every band.