AI-powered search is no longer a future trend, it’s here. Tools like Google’s AI Overviews, ChatGPT and Perplexity are already reshaping how users find information online. As marketers, we need to adapt.
Welcome to the Generative Engine Optimisation (GEO) world, where success hinges not on keyword stuffing but on crafting content that resonates with how people ask questions. It’s a mindset shift, starting with a new approach to keyword research.
The shift from keywords to prompts
Let’s be honest: Traditional keyword research has had a good run. We spent years targeting concise, high-volume search terms – “best CRM software” and the like with success measured in rankings and clicks.
However, large language models (LLMs) like ChatGPT play by different rules. They’re built for conversation, not just correlation. Prompts are longer (averaging 13 words) and packed with nuance, context and intent.
Consider the difference:
Traditional search: Best CRM software
LLM prompt: What’s the best CRM for a marketing agency with five team members?
It’s the same underlying need with an entirely different format.
Optimising for LLMs means understanding and shaping these prompt structures, focussing less on single keywords and more on the interconnectedness of language, intent and context.
Rethinking your research toolkit
To optimise for LLM visibility, your research approach needs an upgrade. That means shifting from keyword lists to prompt ecosystems. Here’s what that looks like in practice:
Think in entity clusters, not isolated phrases
Explore real conversations using natural language patterns
Analyse prompt auto-suggestions across multiple platforms
Map semantic connections between themes, needs and actions
Test and refine prompt libraries to understand trigger patterns
Specificity wins. So do questions; it’s about getting closer to how people really speak, think and search because that’s what LLMs respond to best.
Prompt research tactics that work
Auto-complete mining is your new best friend. Plug partial prompts like “Is [brand]…” or “How does [product]…” into tools like ChatGPT, Bard or Perplexity and see what emerges. These suggestions reflect real user behaviour, not just volume estimates.
It’s also worth noting that different platforms offer different angles:
Google leans transactional and location specific
ChatGPT offers comparison heavy, decision-making prompts
Perplexity caters to the inquisitive, research first audience
Batch-testing variants are another powerful approach. Play with subtle shifts in language, tone or framing and see what changes. Want to go deeper? Run competitor prompts, too. Where they’re visible, you should be. Where they’re not, you’ve spotted an opportunity.
Why intent still matters more than ever
We’ve always talked about search intent. But in LLM environments, it’s the foundation of visibility. Based on Otterly’s analysis of thousands of prompts, here’s how intent typically breaks down:
Informational (70.3%) – “How does blockchain work?”
Comparative (14.7%) – “Final Cut Pro vs Adobe Premiere?”
Troubleshooting (9.2%) – “Why won’t my Bluetooth connect?”
Opinion-seeking (5.8%) – “Is crypto worth it in 2025?”
Informational prompts are dominant, but comparative ones are where brands can influence decisions. Craft content to match each category clear, cited guides for informational, balanced, data-backed comparisons for commercial research.
From keywords to knowledge graphs
Traditional SEO focuses on keywords like:
Best CRM software
CRM for small business
CRM comparison
But LLMs thrive on entities. For the CRM example, an entity-led approach looks more like this:
Primary entity: CRM software
Related concepts: sales pipeline, Salesforce, lead scoring
Attributes: ease of use, integrations, affordability
Actions: manage contacts, automate workflows, track performance
Your content doesn’t just need to mention these things – it needs to connect them. Build authority through contextual relevance. Become the best answer to the real questions people are asking.
Tools of the trade
A new challenge calls for new tools – or at least new ways of using the old ones:
- Otterly.AI – for monitoring prompt visibility (our favourite)
Semrush AI Writing Assistant – to check the AI-readiness of content
Prompt engineering platforms – to optimise structure and clarity
NLP-based SEO tools – to track conversational patterns
Hybrid keyword tools – blending traditional SEO data with prompt insight
Evaluate tools based on how well they reflect real user behaviour, not just estimated volume.
Focus on:
Prompt volume (how often it’s asked)
Intent accuracy (does the tool get it right?)
Brand visibility (are you being mentioned where it counts?)
Competitive density (where are the content gaps?)
What do we think about it?
We’re entering a new phase of search and it’s a conversational one. The way users interact with information is evolving fast, but the goal remains the same: be helpful, relevant, and discoverable.
At Passion, we believe performance and imagination fuel one another. And nowhere is that truer than in AI search. Combining rigorous testing with intuitive storytelling, we help brands find their place in a rapidly shifting landscape.
So, imagine better. Start by asking better questions – and creating content ready with the answers.