Method
How Consilens handles contested questions
The internet is full of confident answers to questions that are not actually settled. Consilens puts several top AI models on each question, then shows where they agree, where they disagree, and how strong the evidence really is.
We evaluate claims. We don't cheerlead for them.
Our job is to weigh what the evidence supports, not to push a side. When a claim is well-supported, we say so. When it isn't, we say that too.
We show disagreement, but we don't fake balance.
"Both sides" doesn't mean "both equal." When the evidence clearly leans one way, the page says so plainly. Manufacturing a tie where there isn't one would be its own kind of dishonesty.
We're loud about uncertainty.
Every page tells you how confident the answer is and what it depends on. If something is genuinely unknown or hotly debated, that's the headline, not the fine print.
No one model is the referee.
Different AI makers have different blind spots, incentives, and tuning. We don't treat any single model as the truth. The whole point is to make them check each other.
We won't help cause harm.
We don't give instructions for hurting people, and we don't present dangerous misinformation as settled fact. On health, safety, legal, and money questions, we'll inform the debate, but we'll always point you to a qualified professional for your own decisions.
A human reviews the hard ones.
The most sensitive questions are checked by a person before they're published, and every page cites its sources. We update pages as the facts change, and we date them so you can see how fresh they are.
How to read a page
Start with the short answer and the confidence.
Open where the models disagreed to see the real argument. Check what would change the answer to see how settled it is. If you want to go further, ask your own version of the question in the app.