An ad is evidence, not a verdict. Clicks and sign-ups can show that someone feels the problem in the language you used. They cannot show that you have a business. The output is a decision: stop, refine the audience, or keep going.
From talk to creative
After the voice agent drafts the problem statement and hypotheses, Microphone generates headlines, bodies, and angles you can reject or ship. Expect to throw away half. The half you keep should be the half you would defend to someone who thinks the idea is bad.
Angles, not adjectives
The useful axis of variation is the reason someone should care: the cost of the problem today, the state after it is gone, the identity of the person who has it, the mechanism that fixes it, or the objection that makes it seem impossible. Running these against one audience simultaneously produces a spread you can actually read.
Reading what comes back
Weight actions that cost the visitor something — sign-ups, replies, booked calls — above raw clicks. A clear winner among angles means people recognize that framing. Flatness across honest variations means the claim is wrong, not the copy. The full method is in how to read ad test results.
Who this is for
Plus is for the first real test, with 500 platform credits a month and extra credits that never expire. When a claim has survived, social media subagents seed the channels where that audience already talks.