How Eartha works

Eartha's answers are AI-generated. A large language model writes each reply inside a fixed set of rules about tone, topic and safety, then Eartha shows it on screen.

What happens when a child asks a question

The question is sent to our server with a system prompt that defines Eartha's character and limits: nature, science, weather, space, animals, recycling, caring for the planet, and the classroom topic a grown-up has set.

The model replies in two to four short sentences aimed at a child aged 5-11. Off-topic, rude, personal or unsafe questions are not answered; Eartha playfully suggests a nature question instead.

What safeguards exist

• Topic fencing: Eartha only answers nature, science and planet-care questions.
• No personal information is ever requested from a child.
• Climate content is required to be hopeful and action-focused, never alarming.
• A visible AI disclaimer sits on every page of the site.
• Published question pages in our questions library are reviewed by a person before they go live, and carry a "last reviewed" date.

What Ask Eartha is not

It is not a search engine, a homework marker, a medical or safety advisor, or a replacement for a teacher. AI can be confidently wrong, so live answers are best read together with a grown-up.

If an answer looks off, tell us on the accuracy page and we will check it.