Would you trust an AI that shows its reasoning?

I've been building tactica., a football intelligence product, and one of the assumptions I've become increasingly uncomfortable with is that giving someone a prediction is enough.

So we've gone in the opposite direction.

The product is designed around:

Evidence → Reasoning → Decision → Outcome → Accountability

For a football match, the model doesn't just output a prediction. It exposes the evidence behind the decision, where the evidence conflicts, what qualifies or rejects a recommendation, and then tracks what actually happened afterwards.

We also separate the model's performance from the user's own betting P&L, because they're two different questions:

Was the model's reasoning any good?

and

How are you actually betting?

What I'm interested in stress-testing isn't whether you like football or betting.

It's the product assumption underneath it:

If an AI product is making recommendations, does showing its reasoning and keeping a permanent record of its outcomes actually make it more trustworthy or does it just create more information for the user to process?

What would you want to see before you trusted a system like that?