AI guide · Aotearoa and Australia
Plenty of educators already paste a note into ChatGPT and get a story back in seconds. It works, up to a point. The problems are where the child's details go, what the story claims that nobody saw, and whether it sounds like every other AI story. Here is what the privacy regulators on both sides of the Tasman actually say, and how to use AI without putting a child's information somewhere it should not be.
Start with what actually happened: one action, quote, attempt, change, or question. A polished observation is never required.
StoryLoop drafts and checks. The educator reviews, edits, decides what is true, and chooses whether anything is shared.
Reviewed 13 October 2026 · Written by Leo, who builds StoryLoop, in Ōtautahi Christchurch
What StoryLoop does with a note
A real note an educator typed, and the start of the draft StoryLoop wrote from it, unedited, with the Te Whāriki links, learning and next steps to follow.
Aroha tried to put her shoes on before going outside. Velcro kept folding over. Asked 'help me' then pushed my hand away and did the second one herself. Stood up and stomped both feet.
Putting on her own shoes
Before going outside, Aroha tried to put her shoes on by herself. The Velcro kept folding over, making the job trickier. Aroha asked, "help me", letting us know clearly that she needed support.
When help was offered, Aroha pushed the adult hand away and did the second shoe herself. Once both shoes were on, she stood up and stomped both feet. This looked like Aroha checking that her shoes were on and feeling the result with her whole body.
Type what you saw today. StoryLoop writes the draft here, with Te Whāriki or EYLF V2.0 links, and shows you exactly what it is built on.
No signup for your first try. Use a first name or initial only.
Your draft appears here, with curriculum links that only claim what your note shows, usually in under a minute.
Tip: what the child did, one thing they said, and what happened next. Three lines is plenty.
Want to keep it? Start free: 3 stories a month, no card. You check every word before anything is shared.
Australia's OAIC recommends, as best practice, not entering personal information, particularly sensitive information, into publicly available generative AI tools. New Zealand's Privacy Commissioner recommends not uploading personal or confidential information to generative AI tools unless the provider confirms it is not kept or disclosed.
A general chatbot will happily add what you never saw: feelings, intentions, curriculum links the note does not support. Every claim in a learning story should trace back to something an educator observed.
Families notice when every story sounds the same. Generic AI wording, the remarkable journeys and wonderful explorations, reads as if nobody was really watching their child.
Check your service's policy first. Some services do not allow general AI tools for anything about children, and that decision is theirs to make.
If it is allowed: describe what happened, not who it happened to. Use a placeholder instead of the child's name, leave out ages, family details, health information and anything distinctive, and never paste photos.
Ask for structure, not judgement. "Organise this note into a learning story with the moment, the learning, a Te Whāriki link and a next step" is safer than "Tell me what this child is learning". You are the one who knows.
Then check every sentence against your note. Delete anything you did not see, anything about the child's feelings you cannot back up, and any link the moment does not show.
They invent. Ask for a learning story from three lines and you will often get a child who "beamed with pride" or "was determined to succeed". Neither was in the note.
They overclaim curriculum. A sandpit moment gets linked to four outcomes because they all sound plausible. Assessors and families notice when links are generic.
They mix up frameworks and spelling: EYLF language in a Te Whāriki story, American spelling, outcomes that do not exist in EYLF V2.0.
They drift into diagnosis words. "Shows signs of" and "may have difficulty with" do not belong in a learning story written by a chatbot that has never met the child.
If you write one learning story a month, a careful prompt and a careful read may be all you need. If you write ten a week, the risk adds up: privacy terms you have not read, links you do not have time to check, and a voice that starts to sound like a machine.
That is the gap StoryLoop was built for: your note goes in, the draft comes back with what every claim is built on, the child's words checked against your note, and the right curriculum for your country. You edit it and sign it off. Three stories a month are free, so you can judge it on your own notes.
Official references
These sources inform this guide. StoryLoop is independent and does not claim endorsement.
FAQ
Neither regulator has issued a blanket ban for educators, but both say the Privacy Act applies to AI, and both recommend not entering personal information into public generative AI tools. Your service's own policy may go further. If in doubt, leave names, photos and family details out.
Removing names, ages, family details and anything unusual enough to identify a child lowers the risk a lot. In a small community an observation can still identify a child, and it does nothing for accuracy or voice, which you still have to check line by line.
On the consumer ChatGPT app, OpenAI may use your conversations to improve its models unless you turn that off in the data controls settings. Its business and API services are not used for training by default. Check OpenAI's current terms before relying on either.
StoryLoop uses OpenAI or Anthropic models through their business services, whose terms do not allow your notes to be used for training, and it stores notes in Sydney. Around the model it checks quotes against your note, uses the right curriculum for your country, shows what every claim is built on, and flags privacy risks before anything is shared. You still read and sign off every word.
It helps. Tell it to use only what is in the note, to quote the child exactly, to name links to Te Whāriki or the EYLF only where the note shows them, and to avoid praise words. It will still need checking against what you saw, every time.
No. StoryLoop supports drafting and structure, while educators remain responsible for observation, interpretation, reflection, final editing, and sign-off.
Yes. Stories are editable after generation and saved in history, so educators can add context, adjust wording, copy, export, or regenerate from the original observation.
StoryLoop is designed to avoid generic, poetic AI wording. It asks for real observations, keeps claims evidence-based, and links curriculum only when the observation supports it.
The free plan includes 3 learning stories per month. Upgrade prompts are dismissible, and existing history remains available even if the free limit is reached.