Here's why your AI singing ads are failing.
Ours didn't win either.
Then I watched two brothers doing $300k/day explain why their first song ads flopped.
They made them R&B and pop. Their buyer is a woman over 60. She doesn't listen to that.
The song has to be the genre your buyer already plays in the car.
Their other rule: if you're not bopping your head, it's a loser. Don't ship the first Suno take. Hit generate until it's stuck in your head.
Next song we make, the genre gets picked before a single line gets written.
1
31
One of our best GLP-1 hooks starts by telling people NOT to buy.
"Don't buy semaglutide for $50. That's a trap."
Sounds backwards, right?
But a warning stops people faster than another list of things they should do.
We tested it next to 4 other hooks. Same 3 women, same body, same offer. Only the first line changed.
"You're paying double" got 0 checkouts.
The trap hook got 13 in 2 days.
We scaled it. $182,617 spent, off one 30-second video.
So don't just test different ads. Test different ways of opening the SAME offer.
Sometimes the problem isn't your product. It's your first line.
Which one makes you stop scrolling: "Here's how to..." or "Don't buy this until..."?
2
29
Our cheapest clicks came from our worst ads.
We went back through every ED ad we ran since June.
One ad got clicks for 36 cents.
1,454 people clicked.
2 bought.
Another ad cost $3.55 a click. 10x more.
773 people clicked.
14 bought.
The cheap one pulled in curious people. The expensive one pulled in buyers.
A click is not a sale. If you reward cheap clicks, Meta will find you all the cheap clicks you want.
So our kill rules don't look at CTR or CPC at all. Just two things:
→ Down $150 on an ad? It's off.
→ 20 checkouts and 0 sales? It's off.
Try it today. Open Ads Manager and sort by cost per click, cheapest first.
Then check how many of those actually bought.
1
1
93
Our first winning GLP-1 angle came from a complaint.
Everybody assumes you have to convince people it works. You don't. They already believe that.
What we kept seeing was this:
"Any hidden fees?"
"What's the catch?"
"Is the price going to go up later?"
Same questions. Over and over.
So we quit selling the result and answered the question.
One price. Every fee listed. Same price at every dose.
That was the first angle that actually worked for us.
We didn't invent it. People were already asking for it.
Now we pull real customer words before every script. 2,000+ posts, comments and reviews so far.
No brand yet? No customers to ask?
Go on Reddit. Search whatever you're about to sell. Read what people are mad about with the brands they use now.
When you see the same complaint 20 times, that's your angle. Answer it in the hook.
A sleep() command is the most profitable line of code in my ad account.
Not the ad generator. Not the kill rules. A 2-second pause.
June 11. I'm pushing a 133-ad statics build through the Meta API. Full speed, ~400 calls back to back. Ad #97 goes in and Meta hands back error 190, subcode 459.
My Facebook account was suspended 90 seconds later.
Nothing in the ads was wrong. Same creative, same copy, same page. The pattern was wrong. 400 writes in a few minutes looks like a bot, because it was one.
Got it back the same night. Then I wrote 4 rules that every script I run now obeys:
→ 1.5 to 2 seconds between every write. No exceptions
→ 50 ads per run, max. Bigger builds get chunked 30 min apart
→ First code 190, stop everything. No retries, no "try once more"
→ Zero calls while an account is under review. Not even reads
Boring. Unsexy. Saved the account that runs the business.
Everyone's posting the AI that writes 100 ads in 10 minutes. Nobody's posting what happens when you launch them in 10 minutes.
1
2
142
David Laniado retweeted
$80k+/month ad accounts are quietly testing animated object ads
not influencers
not classic ugc
just tiny characters explaining problems
a germ on your teeth
a bump in the road
a creature hiding in your beard
you understand the idea in seconds
no long explanation
no expert talking
the animation carries the message
ai creates the characters
ai builds the scenes
ai generates dozens of hooks
one concept
multiple videos
endless variations
rt + comment “objects” and i’ll send the framework
(follow for dm)
David Laniado retweeted
$130k+/month ads are starting to look like casual podcasts
not product demos
not influencer promos
not scripted testimonials
just two people talking
one asks a normal question
the other shares a quick experience
the product slips into the conversation naturally
it feels like a podcast clip, not an ad
that’s why people keep watching
ai now generates the hosts, the studio, and dozens of conversation variations
one format
new question every video
endless ads
rt + comment “podcast” and i’ll send the breakdown
(follow for dm)
David Laniado retweeted
This AI influencer pulled $61k last month using one repeatable UGC format.
No filming days.
No brand negotiations.
No creator management.
Just a content system:
- Same AI face
- Same bedroom setup
- Same lighting and framing
The AI simply swaps the product in her hands and updates the script.
To viewers it looks like one normal creator.
In reality it’s one automated UGC engine posting daily across dozens of offers.
RT + comment “CLONE” and I’ll DM how it’s built.
David Laniado retweeted
$92k/month ads are now using AI “doctor commentary” formats
a skincare demo plays on screen
a doctor appears in the corner explaining what’s happening
the format feels like expert analysis, not an ad
viewers think they’re watching a professional break down a real treatment
that’s why it holds attention
problem on screen
expert commentary below
solution explained in real time
ai now generates the doctor, the explanation, and endless variations of the same format
one concept
dozens of videos
volume finds the winners
rt + comment “doctor” and i’ll send the structure
(follow for dm)
Listening to Kanye again is like hitting a juul on a plane. I know it’s wrong but feels so right.
1
201






