Keywords to conversations: How to optimise websites for the LLM era

Since the advent of ChatGPT and other widely used large language models (LLMs), I, like many others managing websites or digital comms, have been trying to stay up to date on the impact these tools may have on content and SEO efforts. I have read everything I can find (or anything that the algorithms regularly send my way on social platforms) and discussed with experts about what this shift could mean for our day-to-day work.

Will Google die? Is Google already dead? Is SEO really over? These questions often pop up, in headlines left and right. Underneath the obvious clickbait, or hook, if you prefer, lies a serious question: as more and more people turn to LLMs for information (We can definitely sense it but what do the data say?) what should we be aware of, and how do we pivot to ensure our content and brand remain visible, and be cited, by these tools? And crucially: how can we tell if it is working?

If you are wondering on how to update your content and SEO+GEO guidelines, I am sharing below some of the latest insights with links to the respective articles I have recently read on these topics.

Let's start with some data: The state of search behaviour today

People are still online to find information. In fact, this remains the primary reason for using the internet, with 60.9% of people stating so in late 2024, an increase from 60.2% in 2021 (Source: DataReportal, April 2025).

Search behaviour is indeed evolving, as it always has been. In her article“WTF is GEO? Plus, how to cope with the LLM ranking panic” SEO and marketing specialist Jenna Kamal reminds us all that Google remains the most used search engine, with around 67 percent of internet users globally still relying on it throughout 2024 (Source: GWI Core Flagship Study, May 2025). However, usage of generative AI platforms is increasing. ChatGPT, for example, grew from 17% to 22% user adoption in the same year. In another survey of 12,000 consumers, 58% (vs. only 25% in 2023) reported having turned to Gen AI tools for product/service recommendations.

A new search paradigm? From SEO to GEO and back

SEO helps you control your own channel, or your online HQ or storefront if you want, and curating every detail inside (form, tone, structure, content). GEO, or Generative Engine Optimisation, is about your reputation outside of your own channel. You do not fully control what is said, but it affects how AI perceives you. LLMs pull from a much wider set of digital signals: Reddit (which has grown and grown in importance lately), Substack, Wikipedia, media articles, reviews, podcast transcripts, and more.

Here is how Jenna Kamal from GWI puts it: “Marketers naturally want their brand to show up wherever their audience is searching. GEO should be considered part and parcel of SEO now.”

What LLMs optimise for: Resolution over attention

Search engines aim to capture attention. LLMs aim to resolve questions. In the words of Dubois, Dawson and Jaiswal writing for the Harvard Business Review: “LLMs are not optimising for attention; they are optimising for resolution.”

This distinction is essential. A vague marketing slogan is less useful than a clear answer supported by trusted sources. To earn visibility in generative engines, brands need to provide structured, solution-focused content.

“On ChatGPT, unlike Google, there is no page two.”

A new metric to watch? "Share of Model" (SOM)

Share of Model (SOM) is a new metric and platform proposed by the LLM analytics agency Jellyfish to measure how often a brand is surfaced by LLMs and in what tone. I recently read about this in a Harvard Business Review article “Forget What You Know About Search. Optimize Your Brand for LLMs” (I have some obvious reservations about the headline, by the way :) ). One brand, for example, was cited in 24 percent of prompts on LLaMA but under 1 percent on Gemini. Others were completely missing from some models. As the authors point out: “On ChatGPT, unlike Google, there is no page two.”

If you want to learn more about the SOM metric and how it compares with Share of Voice or Share of Search, the article above offers a handy comparison chart.

Michael Brito from Zeno Group highlights that around 90 to 94 percent of brand citations in ChatGPT responses come from earned media, not owned content or ads. That makes your visibility on external sources absolutely critical.

In practice: How to optimise for LLMs

So what can we do to appear in AI responses? Here are some essentials:

• Structure for scanning: Use bullet points, FAQ pages, specs and schema markup. LLMs prefer content they can easily summarise.
  
  

• Answer real questions: Explain not just what the product, service or position is, but why and how it helps in practical scenarios.
  
  

• Sound human: Write the way people speak. Use natural phrases, real names, references and statistics. (And please never use references and statistics provided by LLM models without cross-checking them first. They are often inaccurate or outright fabricated.)
  
  

• Provide trusted sources: Include links to studies, reviews, and credentials. Tools like OpenAI and Anthropic are crawling for credibility.
  
  

• Be visible in off-site sources: Earned media will also earn you citations in branded AI searches. Focus on digital PR, third-party mentions, and high-authority sources.
  
  

• Match model behaviour: Each LLM has a different style. For example, Perplexity might reward content that is versatile in format, uses accessible language, and clearly adapts to varied question types. ChatGPT, on the other hand, has shown a preference for local relevance and depth of explanation. Taking all these model-specific behaviours into account can further improve your chances of being referenced.
  
  A recent study by Profound, analyses the different citation patterns by ChatGPT, Google AI Overviews, and Perplexity.
  
  

• Ask the LLM what it wants: Jemma Kamal offers practical tips on how to do this. Do read her article, it’s pure gold.

Notice anything? The majority of these guidelines above have been part of our good digital content practices even before the arrival of LLMs. For Google it was the very definition of 'helpful content'. In our guidelines for writing for social media, we have always emphasised that a human, conversational tone is what people engage with the most. The fundamentals have not changed, but the context in which they matter has expanded.

To be in the answer, make sure your brand is part of the conversation.

Don’t Panic! Traditional search still matters

Despite the hype around AI, it is not yet time to abandon Google.

AI platforms often cite the same sources that already rank well in Google. If a site is not already visible in traditional search, it may not be quoted in LLM responses either.

Build for both worlds

LLMs are changing the digital landscape by layering new behaviours and expectations on top of established ones.

Your website is still one of your most important digital assets and its role in the digital ecosystem continues to evolve. And just as important is the way your brand is talked about across the wider internet. Your communication strategy as a whole matters. Think SEO and GEO as complementary forces. Build structured, helpful content, and make sure you are being mentioned in the places LLMs are watching.

Generative search is as much a technical shift as it is a cultural one. People are asking questions differently. To be in the answer, make sure your brand is part of the conversation.

Optimise for being remembered.