How GEO Transforms AI Visibility: Lessons from 4,017 Prompts

How GEO Transforms AI Visibility: Lessons from 4,017 Prompts

Let me ask you something: how confident are you that your brand shows up when someone asks an AI about what you sell? Not on Google, not in ads. I’m talking about AI systems like ChatGPT, Google AI, and Claude. Have you even checked?

Here’s the awkward truth. Across the 4,017 prompts we’ve tested at Contxt, nearly 20% of brands couldn’t break a 30% visibility rate. Think about that for a second. If your target customer is turning to an AI assistant instead of a search engine, there’s a decent chance you’re being, well, ghosted. And the craziest part? Some of the brands in our “low visibility” tier aren’t even small players. They’re established names in their industries. Invisible. Irrelevant. In AI terms, non-existent.

So why’s this happening? And what can you do to fix it? That’s where Generative Engine Optimization (GEO) comes in. If search was yesterday’s battlefield, generative engines are today’s. And GEO isn’t just SEO 2.0, it’s an entirely different game with its own rules, opportunities, and risks.

Why Traditional SEO Isn't Enough in Generative AI

Here’s the issue. SEO was built for search engines, where ranking is dictated by signals like backlinks, on-page content, and keyword optimisation. It’s hierarchical, transactional, predictable. Generative AI doesn’t think like that. It doesn’t rank websites in a neat, numbered list. Instead, it synthesises answers from tons of sources, sometimes citing them, sometimes not, and prioritises different kinds of signals entirely.

A classic example? Brand mentions within contextually relevant prompts. In our dataset, we ran 58 experiments across 14 businesses to untangle what drives AI systems to recommend one brand over another. And spoiler alert: traditional rankings barely factored in. Related reading: this piece about why b2b brands struggle with llm visibility (and how to g.... We've covered this in more detail in Comparing Open-Source vs Commercial LLM Visibility Tools:.... What mattered more was how often a brand, or its products, had already been “seen” in relation to a topic.

One standout finding was just how differently engines treat the same prompt. Across all tracked prompts, ChatGPT and Google AI disagreed on recommendations 38% of the time. That’s like flipping a coin. A brand visible on one system could be totally absent on the other. Related reading: ChatGPT and Google AI Give Completely Different Answers. .... Related reading: Why Your Brand Is Invisible to AI: The Cross-Provider Div.... Right now, no amount of keyword stuffing or shiny link-building will bridge that gap.

What’s Driving AI Recommendation Patterns?

In my experience, this is the bit where people start panicking. “You’re saying SEO doesn’t work anymore?!” No, not exactly. SEO still matters for search engines, but generative engines like ChatGPT and Claude operate on a different logic. They prioritise what I’d call editorial signals over traditional search signals. Here’s what I mean by that:

• Contextual relevance: If your brand consistently appears in content related to a specific question or need, it’s more likely to get cited.
• Structured data: AI loves clean, clear information. Schema markups, detailed product data feeds, and metadata all help.
• Prompt sensitivity: The phrasing of a user’s question plays a huge role. In our buying-intent tests, the same brand could go from centre-stage to completely invisible depending on how the query was worded.

One client I worked with recently runs a B2B SaaS platform. For generic prompts like “best software for X,” they were mostly ignored. But for highly specific questions, “Which SaaS platforms integrate with [specific tool]?”, their visibility shot up. A subtle shift in how users interact with AI unlocked a huge opportunity for them to dominate those niche queries. You can Contxt platform features to see for yourself. You can how AI visibility monitoring works to see for yourself. It’s all about anticipating not just what users search for, but how they phrase it.

The GEO Playbook: Closing the Visibility Gap

So how do you make sure generative engines see, and cite, you? Honestly, the first step is accepting that GEO isn’t “one size fits all.” It’s not about ticking boxes like you would in traditional SEO. It’s about experimentation, data, and constantly refining your approach. Let me walk you through some strategies we’ve seen work across our experiments:

1. Embrace Structured Data like Never Before

If SEO was all about content, GEO is all about data. Generative engines aren’t browsing your website like humans. They’re consuming structured information. Make sure you’re implementing schema markups (think FAQs, product specs, local business info). The more structured your site, the easier it is for AI to parse and integrate your data into its responses.

One standout case: Oak&More. In our dataset, they hit a 100% brand mention rate on all tested prompts. How? Their site is an absolute masterclass in structured data. Every product detail, every customer review, it’s all marked up. The result? AI providers treat their site like it’s gospel.

2. Optimise for Prompt Sensitivity

This is where it gets fun (and a bit nerdy). You’ve got to think about how people naturally phrase queries to AI. For example, we ran a competitor-displacement test for a medium-sized manufacturer. When users asked, “What’s the best brand for X?”, they rarely made the cut. But in prompts like, “Which brands specialise in X for [specific use case]?”, they suddenly became an AI darling. It’s proof that aligning your messaging with “real world” questions can move the needle big time.

3. Turn “Low Visibility” into a Case for Editorial Coverage

Remember those 8 brands in our dataset with visibility rates under 30%? (Castle Pumps at 6.3%, ouch!) The quickest wins for them are editorial partnerships. When AI sees your brand mentioned across trusted third-party sites, it strengthens your association with specific topics. Over time, this boosts your odds of getting cited. It’s a commercial opportunity that’s often overlooked because, let’s face it, PR is sexier than schema markups. But both are critical.

4. Experiment Relentlessly

The biggest mistake I see? Assuming what works today will work forever. Generative engines are constantly evolving. Every time OpenAI releases a GPT update, or Google tweaks their model, your visibility can shift overnight. Of the 14 businesses that ran experiments through our testing lab, the ones who tested consistently were the ones who stayed ahead. They pivoted their strategies as new opportunities emerged.

5. Measure Like You Mean It

Data is your north star with GEO. You can’t improve what you can’t measure. That means tracking your brand mention rates across different engines, testing variations of prompts, and monitoring how visibility changes over time. A mate of mine who runs a digital marketing agency recently told me they’ve started running monthly “AI health checks” for their clients. It’s a brilliant idea. GEO isn’t set-and-forget. It’s know-and-grow.

Why GEO Matters More Than Ever

Now, here’s the kicker. Generative AI adoption isn’t just growing, it’s exploding. Businesses are making higher-stakes decisions based on AI-fuelled recommendations. Consumers are skipping search engines entirely. And if your brand isn’t visible where people are looking to buy? You’re handing market share to your competition. No exaggeration, this is about survival in a landscape that’s moving faster than any of us expected.

At Contxt, the most common thing I hear from clients is, “We thought we were doing fine until we saw the data.” Honestly, I get it. I was sceptical about generative engine optimisation myself at first. But when you see an 8% visibility rate for a brand that’s pouring millions into traditional SEO, the penny drops. You might also want to check out our article on why most brands fail at llm visibility (and how contxt is.... It’s not rocket science, it’s about playing the right game.

What’s Next for Generative Engine Optimization?

So where does this leave us? GEO isn’t going away. If anything, it’s going to get more important as AI systems become the default for search and recommendations. But it’s also a bit of a minefield. The rules keep shifting, the signals keep evolving, and there’s no handbook to follow. Yet.

My advice? Start now. Run experiments. Gather data. Test prompts. Implement structured data. Partner with sites that have credibility in your niche. And if you’re feeling overwhelmed, have a chat with us at Contxt. This is what we do. We’ll help you figure out where your brand stands and how to move the needle. You don’t have to do it alone.

At the end of the day, generative engine optimisation isn’t just a techy buzzword, it’s your ticket to staying relevant in a world that’s rewriting the rules. The question is: are you ready to optimise for tomorrow?

Want to see how your brand stacks up in the AI space? Book a Contxt audit today and take the first step towards GEO success.