Here's the answer, and it's not the one most founders want: you don't ask anyone's opinion, including AI's. "Is my idea good?" is a useless question, because AI is agreeable by default and your friends are worse. The move that works is reading what real customers already wrote.
Somewhere on the internet right now, people are complaining about the products in your category. The reviews sorted one star first, the forum threads, the "why does nobody make one that..." posts. For a physical product, this is the richest complaint data on earth: thousands of reviews on every existing brand, telling you exactly where the category breaks its promises. You are not inventing demand. You are reading it.
The disciplined version takes three days. Day one: map your five competitors and their most common complaints. Day two: collect a hundred or more real customer comments and let AI sort them into the top pains, the top wishes, and the exact phrases people use. Those phrases later become your headlines. Day three: message ten real people, ask about their problem without pitching, and listen.
Then the part that protects your savings: write your pass-or-kill rule before your heart gets a vote. If you can't find ten people actively complaining and already paying for a workaround, change the idea now, while the only cost is three days of reading.
The full method is Volume 1 of our Founder series, from QYRAI™, an AI-driven creative marketing agency.
Every no before launch is a refund you never have to issue.