AI versus Marshy - Optimising your website for LLMs

I’ve been talking about the changing landscape of search and SEO for a while now, and I wanted to dive deeper into one aspect that’s caught my attention: optimising your website for Large Language Models (LLMs). If you’re new to this topic, don’t worry - I’ll try to keep it simple. But if you’re already familiar with the basics, let’s get straight into it. The rise of Google, the golden era, adapting to algorithm changes, the introduction of Bing and local market competitors, and bouncing between white hat, grey hat, and black hat worlds. Even with this in mind - SEO right now is having a time . I’ve never subscribed to the school of thought (probably to my detriment) that gaming an algorithm for “free” traffic is a long-term and viable strategy for growth. Having said that - I know plenty of businesses that have built empires from following theses patterns ( HotelsCombined was a client back in the day), and plenty more that were reliant on this channel and got deleted into obscurity over night. My old adage was: just focus on being useful and usable to your customers and the rest follows. Regardless of where you stand - there’s something happening for fast-moving companies that are optimising their web presence for picking up by LLMs. The reason why it works is obvious - people are flocking to the search experiences within a ChatGPT or Perplexity, and enjoy the fact they can back-and-forth, probe more deeply, and actively investigate a topic with nuance and their own lines of questioning. So far - people are willing to make the trade-offs with a bit of random hallucination, in return for this search power. Also so far - there are companies noticing SUPERIOR conversion rates from online enquiries that come from: ​ ?utm_source=chatGPT ​ ^ that’s just nerd speak (and what you see in Google Analytics) for the traffic source From July 2025 We’ve also seen this with a startup we’re advising - Appacca - while small numbers for now, there is no question its “better” traffic. So why is that happening? Two reasons spring forward in mind: Your conversational search history in an LLM seems more “private” and bespoke to what you’re looking for. So when you’re asking a series of 5, 10, or 20 questions looking for a recommendation (and then receive one). I think there’s a natural trust element that you no longer award to Google. ​ I was working at a startup called UseVerb, and one of their (indirect) competitors was SEEK (Australia’s largest job board).We were investigating a small football club that had advertised a job on SEEK.A search for their company profile on SEEK got weird when I searched “seek mackay football club”.The company didn’t only serve up the company page.It served up every single spelling and misspelled variant of that company and led it back to SEEK.This sounds stupid. ​ Originally appeared in newsletter : AI versus Marshy 68 - AI engine optimisation, recognising patterns, and community building question