For twenty years the goal was simple: rank. Get onto the first page, ideally near the top, and traffic followed. That game still matters, but a second game has started underneath it. When someone asks ChatGPT, Gemini or Google's AI overview a real question, the machine does not hand back a list. It hands back an answer. One paragraph, stated with confidence, assembled from whatever the model believes to be true.
That shift changes the job. You are no longer only competing for a position on a page. You are competing to be the source the model trusts enough to repeat. Generative Engine Optimisation, or GEO, is the discipline of making your brand accurate, quotable and citable to those systems.
The entity problem
Ask five assistants who a mid-sized brand is, and you will often get five different answers. One invents a founding year. Another confuses it with a competitor. A third describes a product it stopped selling. This is an entity problem: there is no single, authoritative description of the business for a model to anchor on, so each one guesses.
Fixing it is unglamorous and foundational. A canonical brand description. Clean Organisation structured data. Consistent facts across the site, the profiles and the places machines read. Once the model has one accurate source to hold onto, its answers stop drifting.
If a machine cannot describe you accurately, it will describe you anyway, and be wrong in front of the customer.
Write for retrieval, not just ranking
Models like to lift clear, self-contained statements. Plain-language definitions of what you do. Honest question-and-answer content that resolves a real query in a sentence or two. Claims that are specific enough to quote and true enough to stand behind. This is where GEO and good writing meet: the same clarity that helps a person skim helps a machine cite.
There is a discipline to it. We hold one rule throughout: accuracy before rankings. A claim about a technique, a material or a result is only made when the underlying page confirms it. It is tempting to stuff pages with the phrases you wish you ranked for. Do that and the model learns to distrust you, which is the opposite of the goal.
How you know it is working
GEO is measurable, even if the metrics are newer than a rankings report. Track how often assistants mention you for your category and technique prompts. Watch whether the facts they repeat are yours or a competitor's. On one heritage-label engagement, that visibility index moved from 14 to 25 in the first phase, a 79 percent lift, simply by giving the engines one accurate entity to cite and a body of citable content to draw from.
The blue links are not going away. But a growing share of buying decisions now begins inside a chat window, with an answer you did not get to write. The brands that win the next few years are the ones the machine already trusts. Be the source, not the guess.