Every number, and where it comes from

Every MP profile is built from official public records. Voting data comes from TheyVoteForYou.org.au, which parses Hansard — the official parliamentary transcript — to extract every recorded division vote.

A division is parliament’s formal recorded vote: every member’s name goes on the record. Votes are shown as “Aye” (for), “No” (against), or absent. We never interpret how a member “would have” voted, and we never infer a position from silence.

Attendance is the share of the divisions held in a member’s own chamber over the last 365 days in which they cast a vote. A member who took their seat part-way through that window shows a low figure for that reason alone, so we show no attendance share at all for a member with no votes in the window rather than printing a zero. Pairing arrangements and approved leave are not distinguishable in the public record, so they are not adjusted for — ministers and leaders often show lower figures too.

Party loyalty is the share of divisions in which a member voted the same way as their own party’s majority in that same division. Independents have no party line, so it does not apply to them.

We publish no floor-crossing count. The floor-crossing flag is empty on every division record we hold, so any total would be an absence of data printed as a zero — and “never crossed the floor” would be a claim about a member’s conduct that our data cannot support. Where a member’s loyalty share is below 100%, they voted against their party’s majority in the remainder of those divisions; we show the share and stop there.

Donation data comes from Australian Electoral Commission annual returns and election campaign disclosures. We match declared receipts to members and, where relevant, show donor industries alongside the member’s voting record.

Industry labels are applied from donor organisation names. The AEC publishes annually and disclosure thresholds apply, so this is the declared record — not a complete picture of political money.

A donation from an industry and a vote on a bill affecting that industry is a fact, not an accusation. Verity shows correlation and never asserts causation.

Bill data is collected daily from the Australian Parliament House bills system: title, sponsor, portfolio, status, progress dates and links to the official documents.

Plain-English summaries and the “supporters say / critics say” panels are AI-assisted and labelled as such wherever they appear. We never fabricate bill text or parliamentary quotes.

To answer “how did my MP vote on housing?” we tag divisions with the policy issue they bear on, and record which way an “aye” cuts on that issue. Tagging is AI-assisted and each tag carries a confidence score.

Only tags above 60% confidence are displayed. An issue summary is shown only where a member has at least two tagged votes, so a single procedural vote never becomes a “position”.

Speeches come from Hansard, the official parliamentary transcript, via OpenAustralia. A member’s speech count is the number of Hansard contributions recorded for them in the last 365 days.

Hansard runs behind. The most recent Hansard entry Verity holds is dated 2 July 2026, so speech counts do not yet include anything said after that date. We print the cut-off next to the figure rather than letting a count look current when it is not. Division records are separate and are current to 17 September 2026.

Verity does not pair a speech with a vote to suggest a member is inconsistent. Both records are shown on their own terms; drawing a conclusion from them is the reader’s job, not ours.

News articles are ingested from Australian outlets via news APIs and RSS feeds, then clustered into stories by topic similarity.

Source bias ratings (left, lean left, centre, lean right, right) come from Media Bias/Fact Check, an independent media monitoring organisation. Counted when this page was last built (21 September 2026), 73 sources in our database carry a published rating. We display those ratings as published — Verity does not assign its own bias labels.

We show each article’s source and that source’s published rating, and nothing beyond it. Verity does not currently publish a coverage-imbalance or “blindspot” measure in the app or on this site.

Verity does not conduct its own public opinion polling. We aggregate polls published by others over a rolling window and show the aggregate, the window and the number of polls in it, always dated — never a single poll presented as the state of play, and never an undated figure implied to be today’s. The most recent aggregate carrying a two-party-preferred figure is dated 12 August 2026.

Where no such aggregate exists, the page says so instead of hiding the section or showing the last number we happen to hold.

In-app questions are a different thing entirely: they record what Verity users say, are aggregated by electorate only once a minimum number of responses exists, and are never presented as representative of an electorate or of the country.

AI drafts plain-English bill summaries, the case-for-and-against panels, issue tags on divisions, and the Daily Brief. Every AI-assisted element is labelled where it appears.

The Brief is checked against Hansard and the day’s coverage before publication; if a draft fails that check, we publish the last verified edition and say so rather than publish something unverified.

AI is never used to generate a vote, a donation, a quote or any other primary record. Those are ingested from the sources above or they are not shown.