08 Jun 2026

GEO as the new performance challenge 

GEO is shifting from chasing visibility to understanding and aligning with real audience intent on LLMs.

Location pin in a futuristic setting
Nisa Bayindir
Share:

We were at a recent generative engine optimization (GEO) summit hosted by Muck Rack, showcasing UK-specific data and insights about what AI is reading. 99% of AI citations on answer engines come from non-paid sources – if we drill down to Earned in particular, the figure is 82%, so measuring performance on LLMs shifts attention from paid media, for now.  

As we went through the panel discussions, one theme became evident: a lot of the organisations embarking on a GEO journey are still fixated on Visibility as a key performance indicator on LLM answer engines (how much an organisation is cited in LLM answers to prompts in a given context). But that only tells part of the story as two more fundamental challenges are starting to emerge: 

  • Audiences use answer engines for unique tasks based on which persona needs are active at that point in time, so the analysed prompts have to be laser-focused on the audience mindset
  • Brands aren’t always aligned to how audiences actually ask questions based on that mindset

Yes – GEO is about being found and surfaced on LLMs that matter to an organisation, but it’s actually more about whether your organisation can show up with consistency and in the language or style that reflect audience needs and motivations. 

Large Language Models side with the audience, not brands 

Data from MuckRack’s analysis of 1 million prompts in the UK confirm that LLMs work for the audience, not brands, by default. Sources with highest citations are very consumer/audience led, proving that LLMs are built to show context, comparison, guidance to benefit the audience.  

In addition, LLMs don’t simply retrieve information; they interpret audience intent: the very reason why GEO is now a “hot” insights practice. Small nuances in phrasing or context can completely change the answers, the citations, and which brands show up within these. That makes audience language and style incredibly important. But this is also where many GEO approaches fall short of providing a full picture.  

In this rapidly evolving information ecosystem which operates across a vast audience mindset spectrum, sources shaping LLM responses are constantly shifting often in a matter of days – from Reddit, to YouTube, to publisher content, often in a matter of weeks. In times of crises, probably hours. So what works for GEO today won’t necessarily work next week.  

What matters most for GEO?  

Organisations face two key risks in GEO-friendly communications: a) visibility on LLMs is insignificant if the analysis is not built on audience perceptions, and b) irrelevant, inconsistent or outdated content quickly undermines authority of an organisation against competitors. Not to mention the need for ongoing, repeated monitoring and adjustment to make GEO actionable in the long-term.  

With generative answer engines, 83% of searches end without a single click to any website – simply appearing is no longer the goal, being the anchor in the context of LLMs response is. What matters more is an organisation’s share of voice in a particular industry, category or issue context; measured through metrics like Share of Voice, Share of Answer and prompt-level and issue-level competition tracking. In other words, moving from visibility to active positioning. 

GEO performance increasingly depends on how content behaves in specific environments – a channel-based citation review. Broad, generic recommendations are lost in the noise of audience perceptions and not strategically effective or impactful. The advantage comes building an audience-focused, ground-up insights pipeline which considers how audiences actually engage in each channel. 

Delivering an audience-led approach to GEO 

At MHP, our Maiven approach brings together the elements that shape how brands are spoken about, understood and trusted ‘in the wild’ on digital platforms, to be able to reflect their unique language used in a given topic on answer engines. 

In practice, that means: 

  • designing prompts around real stakeholder questions – from customers and investors to policymakers and media – grounded in natural, UK-based language 
  • testing systematically across models, formats and moments to see how outputs vary 
  • analysing citations and sources to unpack how credibility is being constructed 
  • identifying potential narrative risks or misinformation and disinformation from outdated claims to misleading framing 
  • and, importantly, translating all of this into clear, actionable recommendations across earned and owned channels 

This is what shifts GEO from an SEO-like reporting exercise to a strategic insights pipeline that adapts to the realities of audiences. 

Talk to us about Maiven: [email protected] 

Tags:

Stay ahead of the curve

Sign-up to our newsletters to stay abreast of the latest news, get insider politics and more…

Subscribe