When the AI referee steps in
What the AI sees, what it decides, and where its limits are.
When the two of you cannot agree on who was right, an AI referee decides. It is the second of three steps — see How a challenge is decided.
What it sees
The claim, the counter-claim, and every piece of proof either side submitted before the deadline. Nothing else: not your chat history, not what was said at the table, not anything added afterwards.
What you get back
A decision and its reasoning. You can read what the referee weighed, which piece of evidence carried the outcome, and where it found the two sides unequal. A decision you can argue with is more useful than one you cannot see into.
How often
At most twice per challenge. If the first decision rested on a misreading of your claim — it happens, usually when the claim was ambiguous — the second pass is where that gets corrected.
Paid plans include a higher allowance of AI checks across your account; see What Plus adds.
Where its limit is
The referee judges the case in front of it. It cannot know what you meant but did not write, and it will not fill a gap in the evidence with an assumption. If both sides are equally unproven, expect it to say so rather than to pick a winner for you.
If that still leaves the matter open, a human decision can be requested for a fee, paid by the losing side.
Was this helpful?
Your feedback shapes what we write next.