All Five Policy Areas Decline as Housing Visibility Reaches Zero
Ten weeks into the CiteCompass NZ Election 2026 AI Visibility Monitor, a new pattern has emerged in how generative AI surfaces political-party policy information.
For the first time in the study, all five monitored policy domains recorded a decline in average AI Visibility Score during the same reporting week.
The overall average across the 30 monitored party-policy profiles fell from 10.3/100 in Week 9 to 10.0/100 in Week 10.
That places measured AI visibility almost exactly where it began, with the Week 1 baseline at 9.9/100.
However, the headline average is only part of the story.
Housing recorded an average AI Visibility Score of zero, with all six monitored party profiles returning zero for the first time.
Across the wider dataset, 26 of the 30 monitored profiles now record scores below 20/100, compared with 24 in Week 9.
Individual profiles continued to move in different directions. Eight increased, 11 decreased and 11 remained unchanged, generating 61 points of absolute movement despite a net decline of just nine points across the dataset.
The Week 10 findings highlight two important characteristics of the measured AI information environment:
Overall political-party AI visibility remains persistently low, and even as the aggregate score returns towards baseline, the distribution of visibility across parties and policy areas continues to change.
Week 10 reporting period: 19–26 October 2026 UTC.
AI Visibility Scores are CiteCompass research metrics measured on a 0–100 scale within the defined research framework. They measure observed visibility within generative AI research activity, not policy quality, electoral support or voter preference.
Week 10 at a Glance
| Measure | Week 10 |
|---|---|
| Average AI Visibility Score | 10.0 / 100 |
| Week 9 average | 10.3 |
| Week 1 baseline | 9.9 |
| Change vs Week 9 | -0.3 points |
| Profiles below 20 | 26 / 30 (87%) |
| Profiles at 20 or above | 4 / 30 |
| Profiles recording zero | 7 |
| Highest measured Week 10 score | 26 |
| Profiles increasing vs Week 9 | 8 |
| Profiles decreasing vs Week 9 | 11 |
| Profiles unchanged | 11 |
| Combined absolute movement | 61 points |
Several changes distinguish Week 10 from earlier reporting periods.
The number of profiles recording a score of zero increased from five to seven, the highest count recorded during the study.
At the same time, the number reaching 20 or above declined from six to four, while the highest individual score fell from 33 to 26.
Although the aggregate average declined by only 0.3 points, 19 of the 30 profiles changed.
This reinforces a consistent finding from the longitudinal research: relatively small changes in overall AI visibility can conceal materially different patterns across individual policy profiles.
Week 10 AI Visibility Scorecard
The following table shows the measured AI Visibility Score for each of the 30 monitored party-policy profiles.
| Policy domain | National | Labour | Green | NZ First | Te Pāti Māori | ACT |
|---|---|---|---|---|---|---|
| Cost of Living | 26 | 22 | 17 | 9 | 9 | 7 |
| Crime & Public Safety | 13 | 6 | 13 | 4 | 4 | 7 |
| Economy, Jobs & Growth | 14 | 12 | 19 | 1 | 0 | 8 |
| Health | 18 | 26 | 26 | 9 | 15 | 16 |
| Housing | 0 | 0 | 0 | 0 | 0 | 0 |
Important: Bold figures identify the highest measured AI Visibility Score within each policy-domain row. Ties are shown where applicable. These scores describe visibility within the monitored research environment and do not indicate policy quality, electoral support or voter preference.
Insight 1: All Five Policy Domains Declined for the First Time
The most distinctive Week 10 finding is the simultaneous decline across all five monitored policy domains.
During the previous nine weeks, increases in some policy areas generally coincided with decreases or unchanged results in others.
Week 10 is different.
Every domain recorded a lower average AI Visibility Score than in Week 9.
| Policy domain | Week 9 | Week 10 | Change |
|---|---|---|---|
| Health | 18.7 | 18.3 | -0.3 |
| Cost of Living | 15.3 | 15.0 | -0.3 |
| Economy, Jobs & Growth | 9.5 | 9.0 | -0.5 |
| Crime & Public Safety | 8.0 | 7.8 | -0.2 |
| Housing | 0.2 | 0.0 | -0.2 |
Values are rounded to one decimal place. Changes are calculated from unrounded averages.
Economy, Jobs & Growth recorded the largest decline in domain average, falling by 0.5 points.
Health and Cost of Living continued to register the highest domain averages, despite both declining.
Housing returned to zero.
The combined effect brought the overall AI Visibility Score back to 10.0/100.
This is the first simultaneous decline across all five domains, but it does not establish that political-party information has become systematically less visible across every AI platform or every possible voter query.
The findings describe the observed results within the defined CiteCompass research framework.
Whether the simultaneous decline represents a continuing pattern or a temporary movement will become clearer in subsequent reporting.
Insight 2: Housing Records Zero Visibility Across All Six Parties
Housing produced the most distinctive individual policy-domain result in Week 10.
For the first time in the ten-week study, all six monitored Housing profiles recorded an AI Visibility Score of zero.
| Party | Week 9 | Week 10 |
|---|---|---|
| National | 0 | 0 |
| Labour | 0 | 0 |
| Green | 0 | 0 |
| NZ First | 1 | 0 |
| Te Pāti Māori | 0 | 0 |
| ACT | 0 | 0 |
| Domain average | 0.2 | 0.0 |
Housing has consistently recorded the lowest average AI visibility across the five monitored policy domains.
Its ten-week trajectory illustrates this persistent pattern:
1.0 → 2.2 → 0.2 → 1.3 → 1.8 → 2.0 → 0.2 → 0.2 → 0.2 → 0.0
Although individual profiles have occasionally registered measurable visibility, Housing has never achieved a domain average above 2.2/100.
The Week 10 result is therefore not an isolated low point. It extends a sustained pattern of limited measured Housing visibility throughout the study.
Importantly, a zero AI Visibility Score does not mean that a political party has no Housing policies or that AI systems can never retrieve information about those policies.
It means that the monitored profile returned a score of zero within the defined CiteCompass measurement framework during this reporting period.
The distinction between the existence of policy information online and its measurable representation in AI-generated research responses remains central to the study.
Insight 3: Health and Cost of Living Continue to Account for Most Measured Visibility
Although every policy domain declined in Week 10, the distribution of measurable AI visibility remains concentrated.
Health recorded the highest domain average at 18.3/100, followed by Cost of Living at 15.0/100.
Together, these two domains accounted for approximately 66% of the combined AI Visibility Score points recorded across all 30 monitored profiles.
Yet they represent only two of the five policy domains.
Economy, Jobs & Growth recorded an average of 9.0, Crime & Public Safety 7.8, and Housing zero.
This distribution has remained a recurring characteristic of the research.
Health and Cost of Living have generally generated considerably greater measured visibility than Housing, while Economy, Jobs & Growth and Crime & Public Safety have shown more variable results.
The concentration is also evident among individual profiles.
Only four of the 30 monitored profiles recorded AI Visibility Scores of 20 or above in Week 10. All four were in Health or Cost of Living.
These were:
- National – Cost of Living: 26
- Labour – Cost of Living: 22
- Labour – Health: 26
- Green – Health: 26
This concentration raises an important research question:
Why does measurable AI visibility remain concentrated in particular policy domains while other monitored areas consistently record substantially lower scores?
The available data establishes the distribution but does not determine its cause.
Differences in published information, available citations, third-party coverage, AI-platform behaviour and dynamically generated research queries may all contribute.
Insight 4: Individual Profiles Continue to Move Despite Lower Domain Averages
While all five policy-domain averages declined, individual party-policy profiles did not move uniformly.
The most substantial decreases were recorded in:
- National – Cost of Living: 33 to 26 (-7)
- Te Pāti Māori – Economy, Jobs & Growth: 7 to 0 (-7)
- Labour – Cost of Living: 26 to 22 (-4)
- Green – Health: 30 to 26 (-4)
At the same time, several profiles recorded increases:
- Labour – Crime & Public Safety: 1 to 6 (+5)
- Te Pāti Māori – Health: 10 to 15 (+5)
- National – Economy, Jobs & Growth: 10 to 14 (+4)
- ACT – Cost of Living: 3 to 7 (+4)
These movements demonstrate why domain-level averages require careful interpretation.
For example, Economy, Jobs & Growth declined from 9.5 to 9.0 overall.
Yet three of its six monitored profiles increased, while the other three decreased.
Similarly, Health declined slightly despite Te Pāti Māori’s monitored Health profile increasing by five points.
Across all 30 profiles, the total absolute week-on-week movement was 61 points, while the combined net movement was a decline of only nine points.
The overall average therefore remains an important indicator of visibility, but it does not describe the full extent of movement within the monitored information environment.
Insight 5: Ten Weeks of Research Show Little Sustained Growth in Overall Visibility
The longitudinal trend provides important context for interpreting Week 10.
The Week 1 baseline average was 9.9/100.
Ten weeks into the study, the average is 10.0/100.
Despite fluctuations throughout the research period, overall measured visibility remains only 0.1 points above baseline when rounded to one decimal place.
Ten-Week Average AI Visibility Trend
| Reporting week | Average AI Visibility Score |
|---|---|
| Week 1 | 9.9 |
| Week 2 | 9.5 |
| Week 3 | 10.2 |
| Week 4 | 11.1 |
| Week 5 | 10.4 |
| Week 6 | 10.4 |
| Week 7 | 10.0 |
| Week 8 | 10.0 |
| Week 9 | 10.3 |
| Week 10 | 10.0 |
Week 4 remains the highest overall average recorded during the study, at 11.1/100.
Since then, the aggregate average has remained within a relatively narrow range of 10.0 to 10.4.
The ten-week data does not demonstrate sustained growth in overall political-party AI visibility.
Instead, it reveals a relatively stable, low aggregate average accompanied by ongoing changes in the visibility of individual party-policy profiles.
These results also reinforce the distinction between the availability of political information online and the extent to which that information is surfaced within a defined set of AI-generated research responses.
What Week 10 Shows
Ten weeks of research are beginning to reveal consistent characteristics of the generative AI political-policy information environment.
First, overall measured AI visibility remains persistently low, with the Week 10 average almost identical to the original baseline.
Second, all five monitored policy domains declined simultaneously for the first time.
Third, Housing recorded zero across every monitored political-party profile, extending its position as the lowest-visibility policy domain.
Fourth, measured visibility remains concentrated in Health and Cost of Living, while 26 of the 30 monitored profiles score below 20/100.
Finally, individual profile movements continue even when overall changes are relatively modest.
The central Week 10 observation is therefore:
Ten weeks into the study, the overall AI Visibility Score has barely moved from its baseline. What continues to change is where political-party policy information appears within the measured AI research environment.
As the 2026 General Election approaches, the CiteCompass longitudinal study continues to document these changes without attributing them to any particular political strategy, campaign announcement or AI-platform development.
What the Data Does – and Does Not – Tell Us
AI Visibility Scores measure the observable representation of monitored political-policy information within the defined CiteCompass generative AI research framework.
They do not measure whether a policy is good or bad, whether one political party has better policies, electoral support, voter preference or likely election outcomes.
The research is a longitudinal observational study rather than a controlled experiment.
Individual research queries are dynamically generated using a consistently applied methodology to model changing voter information needs and the evolving election information environment.
Consequently, movements in AI Visibility Scores should not be interpreted as evidence that a particular campaign activity, website change, policy announcement or media event caused a specific result.
The research reports observed differences and changes in measured AI visibility, not causal relationships.
Our Governing Research Principle
Observe. Measure. Report. Do not advocate.
CiteCompass is politically independent.
The NZ Election 2026 AI Visibility Monitor does not endorse or oppose political parties or candidates, assess the merits of political policies, recommend how New Zealanders should vote, measure voter sentiment or predict election outcomes.
All six parliamentary political parties are monitored using the same research architecture, five policy domains, equivalent persona methodology, AI platforms, measurement framework and reporting cadence.
About the NZ Election 2026 AI Visibility Monitor
The CiteCompass NZ Election 2026 AI Visibility Monitor is an independent 12-week longitudinal observational study examining how leading generative AI platforms surface, represent and cite publicly available New Zealand political-policy information.
The study covers six parliamentary political parties across five policy domains:
- Cost of Living
- Economy, Jobs & Growth
- Health
- Housing
- Crime & Public Safety
Thirty party-policy profiles are measured each week to observe how AI-generated responses, visibility and source representation evolve over time.
The research measures information visibility, not political merit.
Research information
- Week 10 reporting period: 19–26 October 2026 UTC
- Study duration: 12 weeks
- Profiles monitored: 30
- Methodology: Version 1.0
- Research methodology
- Week 9 research findings
CiteCompass remains committed to publishing politically independent, transparent and data-led findings throughout the research programme.