From Keywords to Context: Why Generative AI Optimisation Depends on Relevance, Not Rankings

From Keywords to Context: Why Generative AI Optimisation Depends on Relevance, Not Rankings Back in the day, you could rank on Google by stuffing keywords everywhere. Title tags, meta descriptions, blog posts, just cram in the focus keyword and hope for the best. It wasn't pretty, but it worked. Fast forward to now, and AI systems like ChatGPT are rewriting the rulebook on visibility. Here’s the thing: these generative engines don’t behave like traditional search engines. They don’t think in keywords; they interpret context. And if your brand isn’t showing up in their answers, it doesn’t matter how good your site SEO is. You're invisible in the conversations people are having, where decisions are made and recommendations are acted on. Visibility in AI prompts isn’t about ranking; it’s about relevance. The shift from rankings to relevance In my experience, the penny dropped on this when we started running experiments across generative AI systems in our Contxt network. Across 2881 prompts (a hefty dataset), the results were crystal clear: AI visibility isn’t as simple as being “optimised” for search queries anymore. In fact, 39% of the time, ChatGPT and Google’s AI Overview gave conflicting recommendations. So, if you’re visible on one, you’re far from guaranteed a mention on the other. Here’s an example of what I mean. Say you’re a manufacturing company (like Iconsys 2, one of the brands we track). In traditional SEO, you’d target keywords like “industrial automation solutions” and try to rank for that phrase. But in our AI data, Iconsys 2 had an abysmal mention rate, just 6.7% overall, despite fitting the criteria perfectly. Related reading: this piece about top 25 llm visibility tools compared: features, pricing &... . For a deeper look, have a read of this piece about content, links, technical: the seo trinity took 25 years ... . Why? Because the AI systems are responding to the broader context of the prompts, not just plucking out keyword matches. Why context mat