Pulled · Policy
AI Use Disclosure
Last updated: 21 May 2026
This page explains exactly how AI is used inside Pulled, because we think you deserve a clear answer before you trust us with information about your child.
The short version
Before your answers are processed by AI, we automatically remove your child's name, any email addresses, and any phone numbers, replacing them with placeholders. A real human on our team reviews and edits the draft by hand, restores your details, and sends the polished version to you. We use Anthropic's commercial API, which contractually prohibits using your data to train AI models.
How the AI step works
When you submit the form, we pass your answers to Claude, an AI made by Anthropic, via their commercial API. Before anything is sent, your child's first name, any email addresses, and any phone numbers are automatically removed and replaced with placeholders (e.g. [CHILD], [EMAIL_1]). A reference library of NZ home education guidance, MoE policy documents, and real approved applications is included with every request so the AI understands what a strong Section 38 application looks like. The AI returns a draft. A real human on our team reads it from start to finish, fixes anything generic or off, restores your real details, and sends you the polished version.
What the AI does and does not do with your answers
We use Anthropic's commercial API. Under those terms:
- Your answers are not used to train any AI model.
- Your child's name, email addresses, and phone numbers are removed before anything is sent to Anthropic.
- The AI has no memory across customers. Your answers are completely separate from anyone else's draft.
The honest limits of AI
The AI is good at structuring sentences and finding the right tone. It is not a substitute for your knowledge of your own child. That is why a human reviews every draft, and why you have the final say. If anything in the draft does not sound like your family, change it. The signature on the application will be yours.
Why this works better than generic AI
We've studied real Section 38 applications the MoE has approved. Our AI knows what they look like. Specific detail, authentic voice, all learning areas covered. The intake form lets us extract what your family needs to say without requiring you to prompt and iterate. A human reviews the draft to ensure it sounds like you and doesn't have generic filler. That's what makes it different from ChatGPT.
Your choice
If you would prefer that AI not be used in producing your draft at all, email us before paying and we will let you know whether we can offer a fully manual version, and at what price.