As more people use GPT tools and Google’s market share continues to slide, I’ve had more and more people asking me about ‘GEO’ — Generative Engine Optimisation — and how they can get their content or brand showing up in tools like ChatGPT and Perplexity.
Whilst I can completely understand where this is coming from, a small part of me dies inside as even afters years and years of beating the drum that we ‘optimise for users not engines’, when the new platform comes along the first question comes from a minipulative mindset of how to chase an algorithm rather than understanding what people want.
The Engine Is Not Your Audience
What we’ve learnt over the years is that these systems always evolve. And with every iteration, they get better at aligning with what people actually want rather than what’s been gamed into their results. The gap between the tricks and trust gets smaller each time.
Still, the temptation is understandable. A site might get a spike in visibility from something that seems to work well. But if the focus is on manipulating the system instead of creating something genuinely useful, that generally crashes pretty spectacularly. As a consultant I literally can’t afford short term gains with crashes because I’d be fired.
But this ‘designing for the engine instead of the person’ mindset is still here. Maybe its human nature. How do I get the edge?
Whether it was Google in 2011 or GPT-5 in 2025, the mindset has been the same — find the flaw, fire your shot, and hope it lands. Like Luke Skywalker taking out the Death Star by nailing the exhaust port. Maybe I could coin a new phrase ‘DSEO’… Death Star Engine Optimisation.
And well…back in 2011, it worked. I remember one Saturday morning I bought an exact match domain, spun up a templated site with some basic copy, threw a few spammy links at it, and by lunchtime, I was ranking #1 and I’d made an affiliate sale by the end of the weekend. It was absurd. But the engines were simple, and it was too cheap and easy to cheat them.
Those days are over. Today’s platforms, including Google search, are smarter, faster, and learning directly from how users interact with our content. Chasing the quirks of these models is a fools game.
The Pace of Change Is Only Accelerating
Outside of some intense periods, we’ve generally become accustomed to Google algorithm updates every few months. There’s been a few occassions where the bigger updates forced us to adapt, but we had time. We could run tests, observe patterns, reverse engineer, reflect and gradually shift strategies.
That’s gone.
With AI systems, we’re not dealing with quarterly updates anymore. These models are being retrained, fine-tuned, and reinforced at a rate we’ve never seen before.
These models adapt not only on a system basis, but per user, based on how people interact with their outputs. If someone skips a response, rephrases their question, or asks for a source, that’s feedback. The model learns. The loop is tighter, the turnaround faster. And that means even if you manage to ‘optimise’ your way into visibility today, you’ll drop out just as fast if the content doesn’t deliver value to the user.
Trying to stay one step ahead by guessing what the model wants is a race you’ll lose — because what the model wants is what people want. And people don’t reward fluff, misdirection, or content built purely to game a system. They reward usefulness.
Mid-Funnel Is the New Battleground
For years, content strategies have obsessed over top-of-funnel visibility — ranking for broad, high-volume queries to capture attention — and bottom-of-funnel conversion — optimising landing pages to close the deal. The top and bottom funnel are easy sells for budgets due to a simpler sell of more eyeballs or more leads.
But AI is collapsing the top and bottom funnel stages. The middle, the mid-funnel, and messy middle - that’s where the focus should be. That’s the advice I can try to give people wanting to understand what to focus on.
The middle is where our audience are comparing, validating, and getting closer to making decisions. In the past we would chase users in those top funnel queries where they’re asking ‘what is X?’ etc because these were high volume terms where we wanted to be seen and to be considered. But AI systems can do the top funnel stuff all day long and way better than Google.
They’re also stepping in to answer those layered, contextual, comparative queries. They’re scanning, aggregating and presenting condensed insights from multiple sources.
So, if your content doesn’t help someone move forward in the decision-making process — if it’s vague, thin, or transparently self-serving — it won’t surface. In short, if an LLM can describe something better than you can then its not worth putting pen to paper. You really need to get that expertise across.
This shift is a huge opportunity for those willing to invest in useful, trustworthy content. No clickbait, no keyword-stuffed filler. Clear, relevant, human-first, useful content that helps people. That’s the stuff AI models are looking for because it’s what people actually want (and its also what Google wants btw).
So, What to Do Next?
Firstly, ignore the hype and don’t panic.
Forget trying to reverse-engineer language models or squeeze your way into GPT through any tricks. Sorry, I can’t help you with that because the fundamentals haven’t changed — they’ve just become more important and harder to fake.
Focus on creating content that earns its place — content that answers real questions, guides real decisions, and builds real trust. That means understanding your audience deeply, showing up in the messy middle with clear, in-depth advice, and being useful in a way that is difficult for others to replicate.
You don’t need to optimise for the engine. You need to optimise for the person the engine is trying to serve.
That’s the work that lasts. Always has, always will.
