Most people think their ad is broken. It usually is not. You said tacos. The ad showed a sandwich. The page said catering. Nobody lied, you do sell all three. But a stranger had to re-learn what you do at every step, and most of them will not do that work for you. They just leave. Three jobs, one promise: 𝗧𝗵𝗲 𝗰𝗼𝗻𝘁𝗲𝗻𝘁 makes the promise. The one idea that earned the stop. 𝗧𝗵𝗲 𝗮𝗱 repeats it. Same words, same claim, same face. Not a rewrite. 𝗧𝗵𝗲 𝗽𝗮𝗴𝗲 delivers it. First screen, no scrolling, no new vocabulary. Here is the test. Open your own funnel in an incognito window and walk it like a stranger. Write down three lines. 1. The post's promise, in the exact words you used 2. The first line of the ad 3. The page headline, above the fold If those three lines do not match, you just found your leak. And it costs nothing to fix. On the page itself: three things and nothing else on the first screen. Who it is for, the outcome, and one piece of proof a stranger can picture. On paid, the question is not whether to run ads. It is whether the post has earned a budget. Three gates, and you need all three. 1. Two of the three signals: retention, saves and shares, buyer comments. Views are not a signal; they only tell you the hook worked. 2. About one save or share per 500 views. 3. It passes the three-line test. A congruent post scales. An incongruent one just leaks faster and costs more. Then spend about $500 per ad, not per ad set. Most people turn off ads too soon. They see one expensive conversion, but the next 5 coming down the pipeline would reduce "costs" by about 50% or more as the account stabilizes. Scale 20 to 30 percent at a time, not double. On the AI side, what I hand to a machine and what I never will. Hand over the structure. The outline, three hooks off one idea, the re-angles, the caption written last, the repurpose into five doors. Never hand over the promise or the proof. Which idea you are betting on, the specific client story, the real numbers, the one-line next step, and the last ten percent that sounds like you.