Methodology
AI use policy
FamilyPlans uses AI assistance in some editorial and programmatic workflows. We are transparent about where it is used, what it is allowed to do, and the safeguards that keep published pages grounded in checked sources.
No production LLM dependency
The public app does not need live LLM API calls to render pages or answer users.
Facts before prose
AI-assisted copy must be based on structured facts already held by FamilyPlans.
Recorded and reviewed
Approved fragments carry provenance, review status, and links back to this policy.
What AI may assist with
- Drafting answer-first paragraphs from structured data we already hold.
- Summarising data observations into readable editorial fragments.
- Copy-editing human-written drafts.
- Generating page-specific FAQs from facts already on the page.
- Writing parser code while keeping the parsers themselves deterministic.
- Assisting data collection inside editorial sessions, with deterministic imports, provenance records and review before publication.
What AI must not do
- Generate generic city or postcode paragraphs by swapping place names.
- Invent local context, parent quotes, recommendations or safety claims.
- Determine legal, status, or closure classifications for the Sure Start research layer.
- Run live LLM calls inside the production app.
- Auto-publish editorial without review.
- Generate medical, legal, safeguarding, or financial advice.
How we keep this auditable
Every AI-assisted content fragment is recorded in the content ledger with the prompt path, prompt hash, input hash, tool used, review status, and the structured facts the fragment is grounded in.
If asked whether AI was used to write a page, and which facts it was based on, we can answer from that ledger.
Editorial AI disclosure
Editorial articles drafted with AI assistance carry a visible footer disclosure. Activity and operator pages that render AI-assisted fragments carry a similar provenance line linking back to this policy.
This article was drafted with AI assistance and reviewed before publishing. See our AI use policy.