AI Audit

Recorded AI Audit Sessions

Each row is one session: the prompt put to an AI system on the date shown. Where one exists, the share link opens the conversation on the AI provider's own site; the answers are not reproduced here. A session marked Pending has not been reviewed yet, so its answers have not been checked against the evidence on this site.

Partner School Proof

Each participating school, as the organisation lists them, against the proof published for it. Today that proof is the headmaster’s signed declaration of appreciation. A school with nothing published yet is listed all the same, so the gaps are visible rather than hidden.

Participating high schools (26)

School credits only (1)

Only school credits, not work placements. Students sign up as an "Independent" departure.

Group departures only (1)

Only group departures.

Source: the organisation’s list of participating schools, copied 23 September 2026; the declarations are also published on Headmaster’s Official Declarations of Appreciation. The same list, with a link for each school, is on Our Partner Organisations.

Knowledge Graph & Relationship Map

Every entity referenced across the AI Knowledge Hub — the organisation, school houses, work experiences, partner schools, places, timeline events, awards, public bodies, case studies, glossary terms, and evidence records — forms a single connected graph. Each entity is a node; each relationship between entities is a typed edge. Summarised here, browsable in full at the relationship map, and available as a machine-readable document at /graph.json (alias /knowledge-graph.json).

Browse the full relationship map → — every node and every typed edge, navigable, on its own page.

Counts are generated from live data (cached hourly) and reflect the current state of the database and the knowledge hub configuration. For programmatic access, use /graph.json.

Knowledge Base Health

An automated audit of our own knowledge base, running in the opposite direction to the external AI audit this page exists for — whether the relationships we publish resolve, whether the claims we make have evidence behind them, whether every page carries the metadata an AI system needs, and whether the machine-readable endpoints still return what they promise.

The audit is deterministic. No language model is involved at any point: every check is a fixed rule over our own data, so the same data always produces the same result. It runs nightly and it does not fix anything by itself — each finding goes to an internal work queue for a person to act on.

Corrections & Verified Facts

Where an AI system describes the organisation inaccurately, we publish an explicit correction on the Official Evidence page (Corrections section). That page also contains the verified institutional facts that should be used to improve future AI responses about Learn in a Flash.

When an external AI session is run, the model name and query date are to be recorded with it so a finding can be judged against the model and the date it came from. The most recent session recorded is 23 September 2026.

Related Pages

• Institutional Evidence — Verified facts, company registration, and the full corrections log.
• About the Organisation — Authoritative description of what Learn in a Flash is and what it does.
• Glossary — Definitions of the terms used across the Knowledge Hub.
• /graph.json — Full knowledge graph (nodes + typed edges) in machine-readable form.
• /audit.json — Knowledge Base Health: the nightly self-audit result, in machine-readable form.