The realization of how important GEO is to my IR program came to me while gardening
I probably shouldn’t admit this, but while I was gardening the other day – pruning a rather unruly hydrangea – I found myself mentally reviewing my IR program for the year.
Yes, I do have a life outside of investor relations, but sometimes the brain wanders where it pleases. There I was, shears in hand, quietly checking off the usual suspects: upcoming press releases and investor campaigns, roadshows, the media strategy, stakeholder engagement, ESG reporting, etc. It all felt satisfyingly under control – until I realised there was one thing I didn’t know much about: generative engine optimization (GEO).
I’ve since come to appreciate that this little acronym is very important in the life of an IRO.
It struck me that in the not-so-distant past – though it now feels like another era entirely – we could more or less manage visibility. The pre-AI world was reassuringly linear: search engines behaved predictably, inputs were fixed and a well-planned SEO strategy could reliably push our company name up the rankings.
But today, that neat equation is not as clear. There isn’t just one search algorithm anymore; there are hundreds of large language models, each trained differently, each drawing from its own data universe. Two people can ask the same question and receive completely different answers depending on whether they’re using ChatGPT, Gemini, Claude or Perplexity.
I don’t think most people quite grasp the real challenge here – it’s not just about what information audiences are asking, it’s about what information the large language models already have about your company.
A recent MuckRack study found that roughly 27 percent of AI responses were drawn from journalism, over 30 percent from blogs and around 11 percent from paid or owned media. On the surface, that sounds balanced – until you think about what that means. If the internet is full of opinion pieces, commentary or outdated blog posts about your company, there is a very real risk that AI could turn those opinions into facts, depending on how a question is phrased.
Imagine an investor asking a generative engine Which retailers have had reputational issues in recent years? If one old blog post or speculative article gets picked up in the model’s training data, you could find yourself unexpectedly ‘featured’. Not ideal.
GEO is hard to measure – but AI visits are not
As Christopher S Penn pointed out on his Almost Timely News blog, it is possible to measure referral traffic from generative AI tools – essentially, to see when humans land on your site after interacting with an AI system. You’ll never know exactly what the conversation was (thankfully?) but you can see which pages people arrived on.
There’s also a growing ability to track AI crawler activity – how often various bots from visit your site and which content they consume. You will also need to make your company’s own information easier for AI to find and understand. Think of it as tidying up before the algorithmic house guests arrive.
SEO is not dead – it is simply evolving. Large language models are not trained daily; they rely on search to fill in the gaps between updates. So when someone asks ChatGPT about something that happened yesterday, the answer likely comes from today’s search results.
Interestingly, traditional search is now being used by AI systems to check their own answers, reducing the risk of hallucinations. In other words, while fewer humans are visiting websites directly, more AI systems are. The fantastic (and slightly unnerving) reality is that your company website may now be read more often by machines than by people.
How to structure your website for AI consumption
- Structured, well-organized content. Pages that flow logically are easier for AI to interpret.
- Bullet points. Machines love lists – they’re clear, direct and unambiguous.
- Q&A sections. Generative models seem particularly fond of modular, conversational structures.
- Summaries and detailed subheadings. Redundant for us, perhaps, but helpful for an AI scanning your site.
- Plain, conversational language. The clearer the phrasing, the less likely it is to be misinterpreted.
- Test your content through AI. Convert your website into text form and ask an AI tool to summarize it. You’ll quickly see what information stands out, what’s missing and how clearly your key messages come through.
And remember: it’s not just your website. Don’t overlook your wider digital ecosystem – YouTube transcripts, LinkedIn posts, social media content, corporate blogs and even podcasts. If it’s online and publicly accessible, it’s part of what AI could use to define you.
In short, if you don’t feed the AI the right information, it will happily feed itself.
Because in this new era of generative discovery, visibility is still earned – it is just that the audience has changed. It is no longer only humans reading your story. The machines are reading it too, and it is probably time we made sure they get the right version.
Isabel Vilela is head of IR and corporate communications at GoviEx, a London-listed uranium mining company.
