Try it yourself

Test the evidence analysis on your own text.

The same analysis that runs before a payment, live. You see what the AI observed, with the exact words, and, separately, what the fixed policy would do. No payment is made.

Paste a message, or pick a sample

Samples are synthetic

The conversation · one message per line; “Name: message” sets the speaker0 / 4,000Context · optional, e.g. “phone call before a bank transfer”

Sent to the live model for analysis only. No payment is created and nothing is stored beyond a 10-minute cache of identical inputs. Don't paste real personal data.

What you'll see

What the AI observed: patterns, each with the exact words. Quotes are checked by code to really appear in your text; any that don't are discarded and listed.
Its own estimate, in words: not a calibrated probability.
What policy does with it: a fixed rule, separate from the AI. The AI can pause a payment, never approve one.

Try the legitimate samples too, and try to fool it. It will make mistakes; that's why a person decides.