Five lessons from New Zealand's first AI election, and a checklist for your website

Learn why political parties are struggling to appear in AI-generated answers and what New Zealand businesses can do to make their content easier for AI tools to find, understand, and recommend.

What happens when a voter asks ChatGPT which party has the best plan for housing?

At the six-week mark of the 2026 campaign, the honest answer is that the AI may use almost nothing the parties themselves published. CiteCompass's NZ Election 2026 AI Visibility Monitor scores how visible each party's policies are in AI-generated answers, on a scale of 0 to 100. In week six, no party scored above 6 on housing, and three scored zero.

The study measures visibility only. It makes no judgment about policy quality or voter support. That neutrality makes it useful for marketers, because the election has become the largest public test yet of a question every brand now faces: when someone asks AI about you, is your own information part of the answer?

First, a quick definition

Search engine optimisation (SEO) helps a page rank in a list of links. Answer engine optimisation (AEO), sometimes called generative engine optimisation (GEO), helps your content get used and cited when an AI tool writes the answer itself.

The distinction matters more each month. SparkToro's 2026 study found 68% of US Google searches ended without a click to any website. Pew Research found people clicked a traditional result 8% of the time when an AI summary appeared, against 15% when it didn't. Research on Google's AI answers also shows that searches phrased as questions trigger them far more often, around 60% of the time. Buyers researching a considered purchase tend to ask questions.

Lesson 1: Good intentions don't add up to visibility

The average score across all 30 party and policy combinations was 10.4 out of 100 in week six. Twenty-seven of the 30 sat below 20, and the best single result was 41.
Doug Johnstone, who leads the study, says that commercial brands measuring their AI visibility typically aim for 60 to 70%. The parties are well resourced and at peak promotion, and they are still a long way short. Budget and effort don't produce AI visibility on their own. Structure does.

Lesson 2: Human-readable is not machine-readable

A review of the party policy pages found four patterns that make content hard for AI systems to use. Each one is common on New Zealand business websites.

Generative search tends to reward clear, well-structured text with plenty of context. As Johnstone noted, AI tends to favour content with clear headings, short structured points and FAQ sections.

Lesson 3: Absence is filled by someone else

If AI can't use your page, it still answers, using other sources. A housing query could return none of the parties' own material and draw on third-party sources instead.

National's first home buyers policy shows how this works. Its details now circulates via a press release reposted on Scoop, a copy of the policy PDF hosted on interest.co.nz, and news coverage. For a brand, the equivalent sources are review sites, resellers, directories, journalists, and competitors' comparison pages. The message gets through, but not in your words.

Lesson 4: Measure by topic, and measure often

The study's week six report shows why a single headline number is misleading. The overall average moved by less than a tenth of a point between weeks five and six, while individual scores shifted by a combined 124 points. Some moved 15 points in a week.

Some marketers are now tracking "share of model", their presence in AI answers compared with competitors. To make it useful, break it down by topic or product line, track it weekly and set a target.

Lesson 5: Insight without ownership changes nothing

All six parties were told about the research and invited to correct errors; none had done so by week six.

Inside most organisations, AI visibility falls between teams. Marketing owns the message, digital owns the site, IT owns the platform, and the SEO agency reports on rankings. Unless someone is accountable for the number, nothing changes.

A starter checklist for your website

• List the 10 to 20 questions your buyers most often ask before they buy.

• Put those questions to ChatGPT, Gemini, Perplexity, and Google, and record which sources each one cites.

• Check your key commercial pages for facts held only in images, video, or PDFs, and add them as text on the page.

• Audit every URL changed in the past two years and put redirects in place.

• Add a short FAQ section to each priority page, written in your buyers' words.

• Decide who owns AI visibility, and agree a target and a reporting rhythm.

Labour and the Greens have shown upward trends over the six weeks, from a low base, which suggests visibility responds to effort. The parties have a deadline of 7 November. Brands don't, which makes the problem easy to put off. Your buyers are already asking the questions, and the answers are being written with or without you.

Author:  Rebecca Caroe has more than 25 years' experience as a CMO and marketing director across SaaS, technology and professional services. She consults on go-to-market and AI search visibility through Digital Pivot NZ.