Chapter 2: From Clicks to Conversations – User Behaviour in the Generative Era

Search once meant typing short, functional keywords into a box: “best restaurants near me” or “cheap flights Auckland to Sydney.” You had to think like Google to get useful results. That dynamic is changing fast.

AI referral traffic is still tiny, just 0.54% of sessions in July 2025, but the growth rate is impossible to ignore. That’s a 10x jump in just twelve months. Small base, massive acceleration.

And here’s why that matters: early adopters aren’t casual browsers. They’re the high-intent users leaning into conversational search, knowledge workers, decision-makers, and customers looking for tailored answers right now. For New Zealand businesses, this sliver of traffic is less about the volume today and more about the lead pool it represents tomorrow. The brands that train AI engines to recognise and cite them early will be the ones who own that high-value audience as it scales.

The difference lies in how people now interact with AI-driven search. Instead of treating it like a database, they treat it like a partner. Queries are becoming prompts: longer, natural, and context-rich. But prompts aren’t just replacing queries; they’re rewriting how we interact with search.

The most visible expression of this behavioural shift is in how people talk to search engines. For years, we trained ourselves to type in keywords: “cheap flights Auckland Sydney” or “CRM software small business.” Functional, clipped, stripped of context.

Now, prompts are replacing queries. Instead of reducing intent to two or three words, people write natural, context-rich instructions. This shift matters because generative engines reward specificity. The more detail a prompt contains, the better the output. Prompt fluency, the ability to articulate what you want clearly, with context and constraints, is becoming the new search literacy.

As the saying goes: “Garbage in, garbage out.” Brands must recognise that their visibility will increasingly depend on being part of the answers that stem from advanced prompts. Users who master this new literacy will extract deeper, more personalised value from AI, and brands that align their content accordingly will be the ones surfaced.

One of the biggest shifts under the hood is how generative engines process prompts. Generative engines break prompts into subqueries and rewrite them for clarity before stitching the results back together.

Take this prompt, for example:

• “Compare Air New Zealand vs. Jetstar for flying Auckland to Queenstown in winter and tell me which is better for families.”

Behind the scenes, that single request splinters into half a dozen mini-searches:

• “Air New Zealand Auckland to Queenstown winter flights”
• “Jetstar Auckland to Queenstown winter flights”
• “Family-friendly airlines in New Zealand”
• “Air New Zealand winter delays Queenstown”
• “Jetstar winter delays Queenstown”
• “Best airline for families NZ domestic travel”

Each subquery is a retrieval task on its own. Your content doesn’t need to win the full original query; it only has to answer one of the fragments cleanly to make it into the synthesis. The takeaway? Optimise for fragments, not just full queries. When your content is structured so that even a single passage can stand on its own, you dramatically increase your chances of being pulled into the synthesis.

This interaction is what makes conversational discovery so powerful: it condenses the messy middle of the decision-making process into one interface.

Generative engines aren’t just answering questions; they’re remembering them.

• Google AI Mode draws on search history, Gmail, YouTube, and even Calendar data to personalise outputs.
• ChatGPT has introduced persistent memory across sessions, enabling follow-ups days later.

Different systems create different “information environments.” The same query produced different results across Google AI Mode, ChatGPT, and Perplexity, reflecting not just alignment strategies but also the data each platform prioritises. Personalisation doesn’t just adapt answers, it subtly rewrites your information environment.

For brands, this isn’t just a technical quirk; it’s a lead generation opportunity. Hyper-personalisation means no two users are seeing exactly the same synthesis.

If your content is structured to align with specific contexts, “Auckland small business,” “family travel Queenstown,” “enterprise IT procurement, you can surface in ways that feel tailor-made to each user. That kind of relevance doesn’t just build visibility; it attracts warmer, more qualified prospects who are already halfway down the funnel.

The strategic implication? Brands aren’t competing for page-one rankings anymore. They’re competing for a place in the AI’s memory.

To really understand why behaviour is changing, we need to look at the psychology of it. Traditional search trained us to evaluate options. We typed in keywords, opened multiple tabs, compared answers, and weighed credibility. It was effortful, but it gave us a sense of control.

Generative search changes that equation. Instead of weighing 10 links, we get one clean synthesis. Instead of working to filter and decide, we outsource the heavy lifting to the AI. The burden of decision-making shifts from human to machine.

This transformation is accelerating because it aligns perfectly with how people want to search in 2025:

• Trust in AI answers. People tend to treat generative outputs as authoritative and neutral, a phenomenon known as automation bias. People tend to accept AI output without question, particularly when it’s framed in confident, natural language.
• Reduced cognitive load. No more scanning through pages of results, no more weighing up which blog seems trustworthy. The AI does the synthesis, delivering a single answer. That simplicity is irresistible.
• Expectation of immediacy. Younger users, in particular, are growing up with an instinct for instant, personalised answers. Clicking through multiple sites feels slow and outdated compared to asking an assistant who “just knows.”

The risk is obvious: trust can be misplaced. A Choice Mutual audit found a 57% error rate in Google AI Overviews for life insurance queries. Yet people accepted them anyway. The combination of trust, efficiency, and immediacy explains why AI referral traffic, though still less than 1%, is growing at an exponential rate. Users aren’t just searching differently. They’re thinking differently about what “search” even means.

The stats are stark:

• According to Ahrefs, click-through rates for the #1 organic result are down 34.5% when AI Overviews appear.
• Pew Research reports that only 1% of users click links inside AI summaries, while 26% abandon their session entirely after reading them.

The pattern is clear: users are browsing less and bouncing more. The answer itself is the destination. For brands, that means content is no longer the product; it’s the raw material. Your words, data, and insights feed the machine, but the machine owns the engagement.

By now, the trend is clear: clicks are down, AI referrals are up, and user journeys are increasingly contained within platforms like Google AI Overviews, ChatGPT, and Perplexity. For many publishers and brands, this sounds like a death sentence for organic visibility. After all, if the traffic isn’t coming through, what’s left to win?

But here’s where it gets interesting: fewer clicks don’t necessarily mean fewer opportunities.

Google has been quick to argue that AI Overviews actually make the web better for businesses. By answering low-intent, surface-level questions in-line, Google says it filters out noise and pushes through only the users who are truly ready to act. “More valuable clicks” is the phrase they’ve used to describe this new flow. Adobe Analytics data supports this, showing that users who do click after engaging with AI-driven results are more intentional, with lower bounce rates and clearer purchase intent.

Independent numbers tell a tougher story. Publishers are bleeding traffic, and automation bias means many users simply stop at the AI answer: no click, no visit, no data trail. A recent Bain survey found that about 80% of consumers now rely on zero-click results in at least 40% of their searches, reducing organic web traffic by an estimated 15% to 25%.

And yet, hidden in that pain is the upside. AI is doing the filtering for you. The casual, low-value users get satisfied in line. What’s left are the serious ones, the people ready to act. But there’s a catch: you only get access to that pool if your brand is consistently cited in the synthesis.

For Kiwi businesses, that flips the equation. You may see fewer visits, but the ones you do win are more qualified, more valuable, and closer to conversion. The paradox of the generative era isn’t about losing traffic. It’s about shifting from chasing volume to owning the high-intent moments that matter most.

Humans and AI don’t experience your content the same way. UX is for the eye, design, hierarchy, and storytelling. AX is for the algorithm, structure, clarity, and extractable facts. To win in generative search, you need both.

Take a bank’s online mortgage calculator. Sleek for people, but to an AI, it’s just a block of JavaScript, invisible. Now compare an outdoor retailer who lists tramping boot specs in a clean table, weight, waterproof rating, materials, and instantly usable for an AI writing a comparison.

That’s the difference: UX wins attention, AX wins citations. Ignore one and you cut yourself off from half the game. The brands that thrive will layer human-first presentation on top of machine-ready scaffolding. Done right, you’ll surface in AI answers and convert the humans who click through.