The first force?
Fourteen months after winning six percent of the country, One Nation leads the federal primary vote. It has swallowed a third of the Coalition's 2025 support, holds the most loyal vote in the Commonwealth — and yet the same electorate rejects a Trump-style prime minister by fifty-six points and still prefers Anthony Albanese in the Lodge. This is the July read, told as it happened.
Six to twenty-nine
The surge did not build gradually in public view — it arrived almost fully formed. Between the May 2025 landslide and the first DemosAU read in April 2026, One Nation went from 6% to 26%. Since then it has traded in a narrow band at the top of the field: 26, 28, 30, 29. The time axis below is honest — eleven silent months, then four monthly reads.
One line goes up. Two lines pay for it.
First-preference vote share on a true time axis. Hover any wave for the full field.
July is the Coalition's first uptick of the year — 18 to 21 — and it came from the minors, not from One Nation, which eased just one point from June's peak.
DEMOSAU / AEC · MAY 2025 – JUL 2026 · UNDECIDED 7% EXCLUDED
A 22.6-point rise; an 19-point majors bill
Percentage-point change in primary vote since 3 May 2025.
DEMOSAU · JUL 2026
The month the Coalition finally moved
Month-on-month change — Taylor's party takes three points off the minors.
DEMOSAU · JUN–JUL 2026
The defection machine
Where did twenty-three points come from? Follow the ribbons: more than a third of the Coalition's 2025 voters have walked, a quarter of the minor-party vote followed, one Labor voter in six crossed — and One Nation gives almost nothing back: 98% of its 2025 voters are staying put, the most loyal cohort in this poll.
Watch thirty-two Coalition voters break apart
Each dot is one voter in 100, stacked into its party's pile. Press play to send every voter to where they sit today — or click a pile to follow only its voters.
Dots are allocated within each 2025 cohort by largest remainder, so the 2025 piles reproduce the poll's election baseline exactly (ALP 35 · L/NP 32 · OTH 15 · GRN 12 · ON 6). The July piles show only where 2025 voters went — they exclude the 2025 non-voter pool charted below, so ALP and GRN each sit one point off the published topline. Exact flow percentages are in the table view; slivers unlabelled in the report (all ≤3%) are estimated from chart geometry and marked est.
DEMOSAU · JUL 2026 · 2025 BASELINE, SHARE-WEIGHTED
98 v 57: the loyalty gap
Share of each party's 2025 voters still with it in July 2026.
The Coalition keeps barely more than half its own vote. No opposition has rebuilt from that number without first answering where the other 43% went.
DEMOSAU · JUL 2026
Non-voters break heavily right
July intention of people who didn't vote in 2025 or don't recall their vote.
One Nation takes 45% of the disengaged pool — three times Labor's share of it. Compulsory voting makes this cohort matter.
DEMOSAU · JUL 2026
The anatomy of the vote
Twenty rooms of a hundred Australians, one for every cohort the poll publishes. One Nation leads fourteen of them. Its strongest are regional Australia (39), TAFE-qualified voters (35) and the mid-life 35–54s (32). Its weakest are just as sharp — graduates and the under-35s at 21, and multicultural Australia at 19, where Labor polls nearly double its share. The Coalition leads one room in twenty.
One Nation leads fourteen rooms. The Coalition leads one.
The same hundred circles, a different hundred people in every room. Pick a party to follow it from here to the end of the story, then pick any published cohort — or press play to walk the six that carry the argument, marked with a dot. Only the circles that make up the difference between two rooms move.
Each screen is a different hundred people, not the same hundred changing their minds — so no circle crosses the field here. Only the piles resize, and the circles that change hands at a pile's edge are the difference between that room and the country. Every published segment sums to exactly 100, so a hundred circles is the figure itself, with none of Act II's rounding. All twenty published cohorts are selectable here and in the table view. The Coalition leads exactly one of them — outright homeowners.
DEMOSAU · JUL 2026 · SUBGROUP MOES ARE LARGER
The waters it swims in
Ask the country its biggest problem, unprompted, and 44% say the cost of living — nearly three times housing, almost four times immigration. The surge's supposed engine, immigration, ranks third. And the idea that this is imported Trumpism fails its own test: Australians reject a Trump-style PM by 72 to 16.
One grievance towers over everything
Open-ended responses, categorised. Highlighted bars are the two dominant answers.
Multiple categories can apply to one response, so totals exceed 100. Categories under 2% and miscellaneous omitted from the chart; the table view lists everything shown in the report.
DEMOSAU · JUL 2026 · OPEN-ENDED, AI-ASSISTED CODING
Not Trumpism — and proud of it
Agree / don't know / disagree, whole electorate.
63% call Australian democracy something to be proud of. The revolt is against prices and politicians — not the system. Any response that treats it as imported culture war misreads it.
DEMOSAU · JUL 2026
The ceiling
Leading the primary is not the same as winning power. Put names on ballots and Albanese still beats Hanson — three-way and head-to-head — while Angus Taylor fights him to a 38–38 draw. And when One Nation's platform is tested line by line, exactly one policy commands majority support. The first state-level test of that ceiling is Victoria's election on 28 November 2026, and every Victorian poll ahead of it is on the Victorian election polls 2026 tracker.
Albanese leads a field nobody loves
Switch the matchup — the three-way and both forced head-to-heads.
DEMOSAU · JUL 2026
One policy flies. Six sink.
Net agreement (agree minus disagree) with One Nation positions, whole electorate.
Immigration reduction (+40) is the platform. Everything else — monoculturalism (−26), gutting the ABC (−33), abolishing SBS (−46) — loses the very electorate currently voting for the party.
DEMOSAU · JUL 2026
A ladder with no top rungs
Twenty-four public figures were tested, and the best net rating in Australian politics is Jacqui Lambie's +5. The top-right quadrant — known and liked — is empty. Hanson is the most recognised and most polarising figure in the field: highest positive rating (34%), highest negative (42%), and just 1% who haven't heard of her.
Everyone famous is underwater
All 24 figures plotted by how many voters know them against net favourability. Hover any dot; click a legend chip to isolate a party's figures.
The pattern is brutal: the better Australia knows a politician, the less it likes them. The only positive nets belong to the semi-known — Lambie, Minns, Crisafulli — and fame's far end holds Albanese and Morrison at −22.
DEMOSAU · JUL 2026
Albanese −23 · Hanson −9 and falling · Taylor −3 on a sea of shrugs
Positive / neutral / negative, centred on neutral, with change since June.
Taylor's near-parity rests on 49% neutrality — half the country has no opinion of the alternative PM in a dead-heat race. That is either his ceiling or his runway, and July's +3 says it is not yet decided.
DEMOSAU · JUL 2026
Methodology
Question wording
Voting intention: "If a Federal Election was held today, which one of the following would you vote for?" — Labor (Anthony Albanese) / Liberal National Coalition (Angus Taylor) / The Greens (Larissa Waters) / One Nation (Pauline Hanson) / Any Other Candidate / Don't Know (order randomised, Don't Know last). Initial undecideds were asked a leaner; remaining undecideds (7%) excluded.
Preferred PM: "Who is your preferred Prime Minister?" — Albanese / Taylor / Hanson / Don't Know. Respondents choosing Taylor or Hanson were then asked a forced two-way against the remaining pair.
Opinion ratings: "What is your opinion on the following people?" — Positive / Neutral / Negative (leaders); the 24-figure battery adds "I haven't heard of them". People and answer order randomised.
Statements: agree/disagree batteries covering the Trump and democracy statements and seven One Nation policy positions; statements and answer order randomised.
Biggest issue: open-ended; categories generated with the assistance of AI, multiple categories per response possible.
Estimated figures in the flow chart
How this poll was conducted
- Fieldwork
- 3–8 July 2026
- Sample
- 2,694 respondents
- Margin of error
- ±2.5%
- Population
- Australia
- Method
- Internet panel survey, rim weighted by age × gender, education, income, tenure, language, state × region, 2025 federal vote and left–right self-identification
- Commissioned by
- Capital Brief