How this works

Every fact carries its source. Every page carries the question it set out to answer.

What this is for

Australia's public data has been published for years. It has not been organised. auscyclopedia organises it so Australians can actually use it: to answer questions, to find what is hiding in plain sight, and to make sense of the country's record. Where there are claims to check, auscyclopedia looks for what is actually the case, regardless of who is making the claim.

What we check before publish

Every page has to clear three questions for an ordinary reader: what does this mean, why does it matter, why should you care.

Press releases pick the flattering number. We name the framing and show the raw figure beside it. Where the framing is honest, we credit the honest number.

Where the figures come from

Named public sources only. Government datasets, the Bureau of Statistics, the ATO, Hansard, and other public records as the work calls for them. They are public and openly available, and each finding cites them in line.

Sources are chosen for the question, not the other way around.

How AI fits in

auscyclopedia is written with AI. Drafts are produced by Claude (currently Opus 4.7). The rules and prompts behind them are kept open, so the method is auditable.

Large language models inherit bias from what they were trained on. The offset is the best data, a method open all the way down, and, over time, the same questions run against more than one model.

AI gets things wrong, and so will we. The framework is being built so the work, over time, runs itself, with the method visible at every step.

If something is wrong

Tell us. Corrections are made on the page, not hidden. Where official records contradict each other, the contradiction is named rather than glossed over.