01 logo

Generative Engine Optimisation: The New Game Hiding Inside the Old One.

Why getting found in 2026 looks less like ranking and more like being remembered.

By Paul SeraPublished 6 months ago • 4 min read
Generative Engine Optimisation

A client rang me last month with a problem I've been hearing more often. Her traffic was flat, her rankings were fine, but enquiries had dropped off a cliff. We dug into her analytics. Nothing structural had changed. Her competitors hadn't moved either. Then I asked her to open ChatGPT and ask it where to find what she sold. Her business wasn't mentioned. The third name on the list was a competitor she'd been outranking for two years.

That's the bit nobody saw coming. The traffic was still arriving, just to someone else, through a door we'd never thought to lock.

Generative engine optimisation, or GEO if you'd rather a tidier acronym, is the discipline of getting cited, named, quoted, and recommended inside answers generated by large language models. ChatGPT, Perplexity, Gemini, Claude, the AI Overviews creeping further down Google's results page. They're all doing the same thing in different liveries. A user asks a question. The model writes an answer. Some businesses get a mention; most don't. The ones that don't might as well not exist for that query.

This is not SEO with a new hat on, although there's plenty of overlap. SEO has always been about earning a place on a list of links. GEO is about earning a place inside a paragraph. The difference matters more than it sounds, because paragraphs don't have ten slots. They usually have one or two, sometimes zero, and the model picks them based on a process nobody can directly inspect.

What's strange is how much of the new game is built from old bricks. Authoritative sources still win. Clear, well-structured content still wins. Domains with established trust signals still get cited far more often than upstarts shouting from behind a thin layer of AI-generated waffle. If you've spent years building a credible site with proper internal linking, accurate information, and genuine subject-matter depth, you've already done most of the heavy lifting. You just need to point that work in a slightly different direction.

The slightly different direction is where it gets interesting. Large language models don't crawl the same way Googlebot does. They build their answers from a stew of training data, retrieval results, and live web fetches depending on the system. That means three different rooms to be visible in. Your site needs to be referenced enough across the open web that the model has absorbed your brand into its weights. Your pages need to surface in the retrieval layer when someone asks a relevant question. And your content needs to be the sort of thing a model can lift cleanly into an answer without mangling it.

The third point trips up most businesses. Models are looking for sentences they can quote with confidence. They reward writing that makes a statement and supports it. They punish, in their quiet algorithmic way, content that buries the answer five hundred words deep under throat-clearing introductions. If you write a paragraph that begins with "The cost of solar installation in Melbourne typically ranges between $4,000 and $8,000 for a residential system," and you back that up with sourcing, you're giving the model a gift. It will hand that paragraph straight to the next person who asks. Write the same information as "There are many factors that influence the cost of solar installation" and you've handed it nothing.

Statistics earn citations. Quotes from named experts earn citations. Specific numbers, dates, and proper nouns earn citations. The research that's emerged over the past eighteen months keeps pointing at the same thing: models trust content that sounds like it knows what it's talking about, and they identify that confidence through structural signals that good journalists and good technical writers have used forever. Hedging hurts you. Vagueness hurts you. A confidently wrong sentence will sometimes still get picked up, which is its own problem, but a vaguely correct sentence usually goes nowhere.

There's a second front opening up that most agencies aren't paying attention to yet. People are starting to do the kind of research inside AI tools that they used to do across a dozen browser tabs. They ask the model for comparisons. They ask for recommendations. They ask for the best provider in their suburb. The model answers, and the conversation moves on. The user often never clicks anything. Your job is to be the answer, because the click-through is no longer guaranteed.

That changes how you think about content strategy. The page that ranks well for "best accountant Melbourne" is no longer just competing with other pages. It's competing with the model's summary of all those pages. If your page is structured so that a model can pluck a clean, specific, attributable claim out of it, you've got a fighting chance of being named. If your page reads like every other page on the topic, you're invisible at the layer that matters most.

The agencies that figure this out first will spend the next two years quietly eating the lunch of the ones still optimising for the 2018 SERP. The work isn't harder. It's just pointed somewhere new. And like everything else in this industry, by the time it has a proper name and a conference circuit, the easy wins will be gone.

tech news

About the Creator

Paul Sera

Owner of Next Gen Digital Agency in Melbourne, Australia.

Entrepreneurship Courses - paulsera.com

Restaurant Near Me - restaurantnearme.com.au

Generative Engine Optimisation - GEOPT.com.au

Car Events & merchandise - radiallife.com.au

Enjoyed the story? Support the Creator.

Subscribe for free to receive all their stories in your feed.

Subscribe For Free

Reader insights

Comments

There are no comments for this story

Be the first to respond and start the conversation.

Sign in to comment
    Written by Paul Sera