Bridge of Confidence
Ask Grok and Gemini “What’s the safest car today for under $60k?” and you get two very different, very short, lists of careful endorsements. Grok hedges: “No single ‘safest’ car exists”, “safety depends on crash type and features”, and “Here are some of the strongest contenders”. Gemini also hedges at first—”Finding a single ‘safest’ car is difficult”—then dives in: “if you are looking for the absolute best protection available today for under $60,000, the Mazda CX-90...”, a model that didn’t make Grok’s list.
Granted, models aren’t brands, but both the confidence and inconsistency of the endorsements between the two is telling. First, AI is risk averse. It tends to evaluate as low-risk/high-confidence brands that appear across quality sources. Second, those sources change with different models and prompt-intentions. For the CMO, that’s the difference between being THE recommendation, A recommendation, and NOT recommended. Most have no idea how AI makes those choices.
Brand AI Report is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
Over the past few months, I’ve been researching how AI systems make these decisions. I’ve built some diagnostic tools to test why some brands get a confident recommendation, while others get hedged, and the rest are ignored. The pattern that emerged is easier to explain as a scene than with jargony details.
AI’s Bridge of Confidence
Picture the Golden Gate Bridge. On one side is a traffic jam of brands: companies of all sizes trying to reach potential customers on the other side who are prompting AI for product recommendations, comparisons, and “best of” lists. Under the bridge is the cold, dark Bay of Invisibility.
The bridge is supported by two massive cables: one is all the historical data that pre-trained your AI model-of-choice; the other is the architecture and hidden processes AI uses to make decisions. As on the Golden Gate, those cables are untouchable.
But what about the hundreds of smaller vertical cables stretching down, connecting to the roadway? Think of those as the many sources AI relies on for additional information. Those many cables, brands can affect. With an intentional strategy, consistently implemented and monitored, brands can change their fate. For example, they can add to the density and quality of third-party reviews. They can align their product page schema and html text. They can update their Wikipedia pages and link to technical documentation far too detailed or interesting for human consumers. The list is long, but it is manageable. The cumulative effect is to measurably strengthen the ecosystem of sources AI relies on to frame a brand. The higher the confidence, the less likely AI is to wrap it in hedging. The lower the confidence? Bay of Invisibility.
Finally (for this post), the Bridge of Confidence has a strict load limit. If a brand hasn’t built up the support for a confident recommendation, the AI may let it cross, but only with restrictions, wrapping it in hedged language that the brand itself is unaware of – language the AI created to reduce its risk of being perceived as wrong. Or, AI sees the brand as too much trouble and skips over it entirely.
The brands that consistently work on those vertical cables -- building quality 3rd party review coverage, maintaining accurate knowledge graphs, creating structured technical documentation, strengthening the signals AI uses to assess categorical authority – eventually build sufficient evidence to support crossing with lower restrictions.
And speed matters. Brands that structure and strengthen their ecosystem of sources will cross the bridge before those relying on legacy signals alone.