We'd rather show you two AIs disagreeing than hand you one "truth"

When I set out to build a fact-checker, the thing that bothered me about the rest was the confidence.

One model, one score, one verdict, delivered like it settles the matter. That is not how truth works, and it is exactly how you get quietly steered.

So Check does the opposite. A standard check runs two independent AIs (Gemini and Grok) over the same claim, and if they disagree, you see the disagreement instead of an averaged-away number. Every verdict links to real, clickable sources, and each one has a public proof page, so nothing is invented behind the scenes. If the evidence is not there, it says so rather than making something up.

I care about this because I do not want to be told what is true. I want to see the evidence and decide for myself.

Genuine question for this community: when an AI hands you a fact-check, what would actually make you trust it?

Real sources?

Seeing the models disagree?

Something else?