The Cross-Provider Visibility Gap: Why 39% of AI Systems Disagree on Brand Recommendations

The Cross-Provider Visibility Gap: Why 39% of AI Systems Disagree on Brand Recommendations Here’s a stat that stopped me in my tracks: across our dataset at Contxt, 39% of the time, ChatGPT and Google AI Overview give conflicting answers to the same prompt. Let that sink in for a second. Two of the biggest AI systems out there are nearly four out of ten times at odds about which brands to mention for identical queries. That’s not a gap, it’s a canyon. But what does it mean for businesses? Well, if you’re optimising your digital presence for one AI system, your competitors could still be eating your lunch on another platform. It’s no longer just about ranking on Google Search or getting your brand into human-written content. You need to tackle the AI visibility gap head-on. Because if AI can’t find you, your potential customers won’t either. So, why do AI systems disagree so often? The good news is, we’ve got some answers. Contxt ran 2,641 prompts across 30 businesses, testing how major AI tools, like ChatGPT, Google AI Overview, Perplexity, and Claude, handle brand visibility. Here’s the headline: mention rates vary wildly. ChatGPT mentioned brands 74.4% of the time in our tests. Google AI Overview? Only 44.1%. Perplexity was even lower at 40.6%, while Claude managed a surprisingly strong 77.1% mention rate (though we only tested it with 70 prompts). The numbers don’t lie. A brand that’s front and centre on ChatGPT might be completely ignored by Google AI Overview, or vice versa. This isn’t just a fluke. It’s systemic. Related reading: this piece about which sources do chatgpt, perplexity, and google ai actua... . Related reading: our article on why 39% of ai systems disagree on brand recommendations (... . And it’s all down to the way these systems retrieve and rank information. comparison-chart-mention-rates Here’s my read on it: Large Language Models (LLMs) rely on training data and algorithms that prioritise different sources. ChatGPT, for example, seems to favo