Martin Bayton
Every now and then someone sends you something that makes you stop, read it properly, and think “yeah… this is worth sharing.” That’s exactly what happened when one of our Pureinsights customers pointed me to Chapter 2 of The AI Search Manual from iPullRank.
Now, full disclosure: this is firmly rooted in SEO-land rather than enterprise search. But stick with me – because the behavioural shifts it talks about are absolutely relevant to anyone building, buying or relying on search systems today.
The chapter is titled User Behaviour in the Generative Era: From Clicks to Conversations, and the headline message is pretty clear: people aren’t searching the way they used to – and they’re not clicking the way they used to either.
Less Clicking, More Synthesizing
One of the big themes is the rise of AI-generated answers (Google AI Overviews, AI Mode, ChatGPT, Perplexity, etc.) and what that does to user behaviour. Instead of bouncing between blue links, users increasingly scan a synthesised answer, get what they need and move on. We’re starting to see the same expectations show up inside enterprise environments, where users want answers, not lists of documents.
In practical terms, that shift looks like:
- Fewer result-list journeys
- More answers delivered directly in the interface
- And higher expectations that search will actually solve the task, not just point to documents
Whether you see that as progress or a problem, the shift in how users interact with search systems is hard to ignore.
Prompts Are the New Queries
One idea that resonated strongly is that prompt quality is quickly becoming a foundational search skill.
We’re moving beyond keyword shorthand toward richer, more contextual questioning. The more clearly you express intent, constraints, and situational context, the more relevant the response becomes. Search is evolving from a transactional lookup into an interactive dialogue — with major implications for how content is discovered, interpreted, and reused by AI systems.
How Prompt Quality Impacts Output
As prompts become more sophisticated, output relevance improves dramatically across AI systems compared to traditional search.
Trust, Bias, and “Good Enough” Answers
The chapter also digs into trust – and this is where things get uncomfortable. Users are increasingly willing to accept AI-generated answers at face value, even when they’re incomplete or occasionally wrong. If the answer looks confident and well-structured, many people simply stop searching.
For organisations building AI-driven search, that raises a different set of questions:
- Which sources are being used to generate the answer?
- How do we make those answers explainable and trustworthy?
- And how do we stay in control of the knowledge being surfaced?
Worth Your Time
A quick thank you to the customer — and professional peer — who shared this chapter with us. It sparked some great discussion within our Special Interest Group on Search and AI, and it’s exactly the kind of perspective-sharing that makes that forum so valuable.
You don’t need to agree with everything in the chapter to get value from it. What iPullRank does well here is clearly articulate how search behaviour is changing — not just at the algorithm level, but at the human one.
If you work anywhere near search, discovery, content or AI-driven experiences, Chapter 2 is well worth a read.
– Martin
Additional Resources
- 7 Tech Trends in AI and Search for 2026 – Pureinsights
- Generative Engine Optimization (GEO) – Wikipedia
- A Complete Guide to AI SEO in 2026 – Eric Siu on YouTube
- SEO Experts Build Me an AI Search Strategy (GEO) – Ahrefs on YouTube
- Generative Engine Optimization (GEO) Rewrites the Rules of Search | Andreessen Horowitz