ReKid ™

Four gates, each able to refuse

The design is fail-closed. When a check cannot complete, the answer is withheld rather than shown — an unchecked answer counts as a failed one.

Gate 1

The question

Runs before any model, so a blocked question costs nothing and produces nothing. Disguised spellings are normalised first, so look-alike characters and spacing tricks are caught too.

Gate 2

The writing

The answer is written for the age range on the package, not for an adult. Length, vocabulary and examples follow the reader, not the topic.

Gate 3

The sources

Every cited address is fetched. Blocked and high-risk domains are refused, redirects are re-checked at each hop, and private or loopback addresses are rejected outright. Dead and parked pages are dropped.

Gate 4

The review

A second model reads the finished answer before anyone else does. It approves, corrects or refuses. A refusal — or a review that fails to run — means the answer is not shown.

What happens between the question and the answer

  1. 1

    The question is checked first

    Before any model runs, the question is normalised — leetspeak, look-alike characters and spacing tricks are collapsed — and matched against blocked topics. Roughly thirty milliseconds, and no model is called at all.

  2. 2

    Unclear questions are asked back

    If a question could mean several things, ReKid™ offers a few clearer versions instead of guessing. This costs nothing from your allowance — we asked, so we pay for it.

  3. 3

    The answer is written for the age

    A model writes the answer at your child's reading level, together with a few quick facts. It can write in a different language from the interface — useful while a child is learning one.

  4. 4

    Every source is fetched

    Each address the model cites is actually requested. Ones that do not resolve, return 404, or land on an error or parking page are dropped. A dead link is worse than no link: a child clicks it and thinks they were wrong.

  5. 5

    A second model reviews it

    A different model reads the finished answer against the question and the age. It can approve, correct, or refuse. Refused answers are recorded so a grown-up can see the question was asked — but the text is never shown.

What the safety layers block

The list below is the topic dictionary, not the whole system: pages are also judged by their address and their actual text.

CategoryExamples of what is blocked
Adult contentPornography, nudity, sexual material, adult sites and adult top-level domains
Drugs, alcohol, tobaccoHow to obtain or use, purchase sites, vaping
GamblingBetting sites, casinos, odds, loot-box promotion
Violence and weaponsGraphic violence, gore, weapon purchase, explosives
Hate and extremismHate speech, terrorist or extremist material
Self-harmHandled with care — see below
Dating and stranger chatDating apps, random-stranger video chat
Unmoderated user contentSocial networks, forums, comment sites, chat platforms

Self-harm is handled differently

A child asking about self-harm is not misbehaving. ReKid™ does not return search results for these questions, and it does not scold. It shows a short, warm message encouraging the child to talk to a grown-up they trust, alongside a helpline. Nothing about the question is attached to a name, because ReKid™ does not know any names.