Why 38% of AI Systems Disagree on Brand Recommendations (And What It Means for GEO)

Why 38% of AI Systems Disagree on Brand Recommendations (And What It Means for GEO)

Let’s start with the stat that knocked my socks off: across our dataset of 2801 AI prompts, ChatGPT and Google AI Overview gave conflicting brand recommendations 38% of the time. That’s not a rounding error; that’s a full-on divergence. The same prompt, run on two leading AI systems, and almost 4 in 10 times the results vary wildly.

My read on this data? This isn’t just theoretical. If you’re relying on AI to drive brand visibility, you’re playing a rigged game unless you fully understand which systems are pushing you forward, and which are leaving you invisible. And that’s exactly where GEO (Global Engine Optimisation) comes in.

Why Do AI Systems Disagree?

Okay, let’s get into the weeds here. When you ask an AI to recommend a brand, whether it’s for industrial automation or sports tech, it’s not just plucking responses out of thin air. The systems are tapping into a mix of signals: past training data, live internet queries, proprietary algorithms, and (depending on the model) citation networks with varying levels of accuracy. But here’s the kicker: the weight they give to these signals differs dramatically.

ChatGPT, for instance, has a brand mention rate of 75.1% across the prompts we’ve tested. Claude edges it out slightly at 79.1%. But compare that to Google AI Overview, which delivered just a 45.6% mention rate. They’re working with the same data inputs (more or less), so why the gap? My gut says it comes down to priorities. Google still builds its models heavily around search intent, whereas ChatGPT places more emphasis on conversational fluency and broader recall. Related reading: this piece about the hidden challenges of llm visibility for international.... Related reading: our article on why 40% of brands are invisible to ai: the cross-provider.... It’s almost like ChatGPT is designed to throw out a name for every topic, while Google AI Overview is the picky kid who only answers when it’s sure.

But does this matter? Actually, yes. If AI systems disagree, it creates visibility gaps. Brands that show up proudly on ChatGPT may barely scrape into recommendations on Google AI at all. I've written about this separately in our article on from visibility to influence: how award-winning llm platf.... Related reading: Why Your Brand is Invisible to AI: Insights from 2,321 Pr.... And if your target audience is split across platforms (which they definitely are), that’s a problem you can’t afford to ignore.

Cross-Provider Divergence: A Real-World Example

Let me make this tangible. One of the brands we track, Contxt itself, has a mention rate of 13.0% across all prompts. But when we break this down by provider, the disparity is striking. Claude mentions Contxt almost 13 times out of every 100 queries. ChatGPT, on the other hand, lands closer to 10. And on Google AI Overview? A miserable 6 mentions per 100 prompts. For a company specialising in AI visibility, that’s ironic enough to make Alanis Morissette write another verse.

It’s not just us, though. Iconsys 2 (another brand in manufacturing and industrial automation) barely registers at 6.7%. That means in most cases, AI systems aren't recommending them at all, just total invisibility. On the flip side, Viking Pumps and IPU hit 100% across all tracked systems. Every single prompt we’ve run mentions them, no matter what provider or context. So, clearly, some brands have cracked the code while others are stuck in the dark. Related reading: this piece about the hidden cost of brand invisibility in ai: insights fro.... You can AI visibility tracking features to see for yourself. The differences aren’t random; they tell us where visibility battles aren’t being fought, let alone won.

What Signals Drive Visibility?

Let’s talk about what’s under the bonnet. From our testing lab (58 experiments completed, thank you very much), we’ve identified a few key factors that AI systems seem to love, or hate, when making brand recommendations:

1. Query Type: Buying-intent prompts tend to favour highly visible brands. For example, asking “What’s the best industrial pump manufacturer?” invariably pulls Viking Pumps to the top. But ask a broader or more educational question, like “Who offers solutions for fluid transfer?”, and less-established brands start disappearing entirely.
2. Training Data Bias: AI visibility hinges on what the models were trained on. If past data features a brand heavily (think reference articles, product comparisons, user reviews), it’s likelier to surface. This explains why names like Contxt lag; in a relatively new space like AI visibility tools, training data can be patchy.
3. Updating Models: From our model-update-monitor experiments, we found that newer AI versions sometimes drop previously visible brands entirely if their algorithms reprioritise signals. Staying visible is not a one-time fix, it’s a constant battle.

And here’s where things get messy. You can dominate one signal but fail entirely on another. Red Fern (manufacturing) has high visibility because it nails signal prioritisation across all systems. You can how AI visibility monitoring works to see for yourself. But TOCA Social (entertainment/sports tech) struggles with just 10% visibility because its niche doesn’t align well with AI’s dominant recognition frameworks.

The GEO Playbook: Fixing Visibility Gaps

So, what do you do if your brand’s stuck in that awkward 30% mention rate range, or worse, invisible? Here’s where GEO (Global Engine Optimisation) strategies come into play. Forget SEO for a minute; this is bigger. It’s about creating a presence that resonates across every provider’s unique logic.

Here are the actionable steps I’d recommend based on our data:

1. Test Across Platforms Regularly: Don’t assume one system’s results mirror another. Run prompts on ChatGPT, Google AI Overview, Claude, and any other major player. Look for patterns, then tweak your efforts accordingly.
2. Create Multi-Layered Signals: High visibility brands succeed because they dominate multiple layers, website content, third-party citations, partnerships, and strong buying-intent associations. Flood the ecosystem, so no matter where AI looks, it finds you.
3. Monitor Model Updates: AI changes fast. If you’re not tracking how new versions shift visibility, you’ll get blindsided. Tools like Contxt’s model-update-monitor can spot when the ground beneath you starts to shift.
4. Optimise Brand Recall with Contextual Prompts: In our brand-recall-vs-recommendation experiments, brands often fared better when prompts included specific associations like “trusted distributor” or “eco-friendly supplier”. Those modifiers matter more than you’d think.

Does Visibility Guarantee Conversions?

This bit’s important: just because AI recommends you doesn’t mean it’s a golden ticket to success. Across our buying-intent tests, we found visibility alone doesn’t drive conversions unless the recommendation aligns with clear consumer needs. For example, being the most mentioned brand for “manufacturing pumps” is pointless if the query focus shifts to eco-friendly options and you’ve got no messaging to match.

It’s about consistency but also relevance. Brands that fail to bridge the gap between visibility and intent end up with wasted opportunities. GEO isn’t just about being seen, it’s about being seen for the right reasons.

The Future of AI Brand Recommendations

Here’s where I get a bit philosophical. The divergence rate, 38%, isn’t going away anytime soon. In fact, I’d wager it’ll increase as AI systems specialise further. Claude might evolve to dominate niche conversational contexts while Google doubles down on algorithmically proven answers. Brands that fail to keep up will face an uphill battle just to avoid being forgotten.

One thing that surprised me during testing was the consistency among certain brands, Viking Pumps and IPU, as mentioned earlier, nailed visibility across every platform. What are they doing differently? I’d say it’s threefold: they’re proactive about their footprint, they reinforce visibility signals across multiple verticals, and they actively test their presence before gaps emerge.

Final Thoughts

At the end of the day, divergence across AI systems isn’t just a quirk or statistical oddity, it’s a massive challenge for brands trying to nail their visibility strategy. Some are thriving; others are barely scraping double-digit mention rates. The question is, where does your brand fall?

If I were you, I’d start testing today. Get your name into those prompts and check who’s recommending you, or not. React fast when gaps appear or systems evolve. Because if AI can’t find you, neither will your audience.

Ready to optimise your visibility? Start here. Or, if you want to dig deeper into our research, check out the tools at Contxt.