Why we publish our method
Anyone can claim to be neutral. Publishing the method — and a permanent corrections log — is how we let you check instead of asking you to believe.
Australia Speaks asks you to do something unusual for a political app: trust it. So it seems only fair to start by saying that we don’t think you should — not on our word, anyway.
Every outlet, feed and commentator says they’re balanced. The claim is free, so on its own it can’t carry any weight. What can’t be faked is the working: how a summary is made, what checks it passes, who signs off, and what happens when it’s wrong. So that is what we publish, in full, on our method page.
The short version
Every bill summary is generated from the actual text of the bill as introduced to Parliament, under instructions to explain only — no opinion, no framing, not left, not right, not centre. The three perspective panels are clearly labelled AI opinion, all written by the same model with the same word budget so no side gets a smarter or louder voice. A second, independent AI model then checks every piece for balance and grounding, and a human editor approves all four pieces before anything is published. No auto-publishing, no exceptions.
Your vote works the same way: the database has no column linking a person to a choice, so even a complete copy of our data could not reveal how you voted.
Why write it down?
Because a published method is a standing invitation to catch us out. Our method page is kept under version control — any change to how content is made or checked is a public, dated change to the standard, not a quiet edit. And when we get something wrong, the correction goes on a permanent public log. You can flag any summary or perspective as unfair from inside the app, and every flag goes to the human review queue.
None of this makes us neutral by decree. It makes us checkable — which we think is the only version of neutral worth offering.
Australia Speaks is in early access. If you’d like to hold us to all of the above, join the list.