A most-contested ranking is a category where AI models disagree the most: different models crown different #1 brands, or brands carry wide spreads between their best and worst model ranks.
Contested categories are where AI visibility work pays off fastest. When all four models agree on a leader, the ranking is entrenched. When they split (ChatGPT crowns one brand, Gemini another), the category's AI narrative is still being written, and the sources that feed the lagging models are worth studying.
How a category becomes contested
Disagreement is usually a sourcing story. Each model reads a different slice of the web at answer time, weights recency differently, and resolves an ambiguous category boundary its own way. A category that spans several jobs — "project management" covering everything from a solo task list to portfolio planning — gives four models four defensible answers.
Entrenchment is the opposite signal. When every model opens with the same brand, that brand is usually the one the underlying sources describe most consistently, and displacing it means changing what those sources say rather than changing a page.
Why contested categories are the opening
A contested board is one where the answer is not settled yet. The models that already name a brand show what evidence works; the models that do not show which sources are missing it. That gap is legible and actionable in a way an entrenched category is not.
The practical read: look at which model disagrees, then look at what that model tends to cite. A brand absent from one model but present in three usually has a source-coverage problem, not a product problem.