Copywriting fed my family before anything else did.
Before AdSkills, before the stages, before anybody knew my name, I was the guy getting PAID to write ads. It's the skill I tell everyone to learn first, because it's the one that made me un-cancelable.
So you'd think building an agent that writes my Meta ads would feel like handing a stranger my pen.
It did. For about a day.
Then I remembered I've done this before, back in my agency years, every time I hired a junior copywriter. You don't hand juniors the pen. You TRAIN the hand that holds it. Rules, frameworks, examples, edits.
The only difference with this junior is that it never forgets a rule, never gets bored, and never turns in ad number 40 with less effort than ad number 1.
This is Part 2 of the AI Paid Ads Department we're building all month in Build Notes. Part 1 gave the department eyes... the Daily Ad Brief that reads your ad accounts every morning before you're out of bed.
Today it gets a mouth.
The Volume Lever
On Monday I published the damage report on what AI did to Google Ads and Meta Ads this year. Short version: Meta's new Andromeda and GEM systems rebuilt the engine, your targeting inputs barely matter now, and creative is close to the only lever you still own.
Today is about how hard you can pull that lever. But first, the numbers that changed how I think about ad creative.
A company called Motion (they make creative analytics software for ad teams) published their Creative Benchmarks report in April. The dataset behind it: 578,750 ad creatives across 6,015 accounts and $1.29 billion in spend, from September 2025 through early January 2026.
Here's what's inside...
Only about 5% of creatives become winners. Motion defines a winner as an ad that earns at least 10x its account's median spend (with a $500 minimum). Roughly half of all ads never get meaningful spend at all, and the winners hoover up about 55% of the account's entire budget.
Those three numbers describe a slot machine where 1 in 20 tickets hits.
And nobody can tell you which ticket in advance. Not me, and not Meta.
So how do the big accounts win? They stopped trying to predict the winner and started buying more tickets.
Motion's median launch volume by account size: accounts spending under $10K a month launch 2.8 new ads a week. Enterprise accounts at $1M+ a month launch 18.9. The top quartile roughly TRIPLES those numbers.
And the payoff math is almost linear. An account launching 4 ads a week finds about 0.2 winners per week. An account launching 18 finds about 0.9. Upload 20 ads and expect 1, maybe 1.6 winners. Upload 50 and expect 2.5 to 4.
The hit rate barely moves. The ticket count is the whole game.
I call this The Volume Lever. You can't out-think the auction anymore. But you can out-produce it.
Now for my favorite number in the whole report.
Motion broke down win rates by asset type: text-only ads win 11.6% of the time. Product image with text, 8.75%. UGC (user-generated content, the shaky-phone testimonial videos every brand is chasing) comes in at 7.56%. And high-production creative, the polished agency stuff, wins 6.87%.
The plain TEXT ad wins at nearly double the rate of the expensive video.
For 20 years I've watched businesses convince themselves they can't compete on Meta without a video team. A $1.29 billion dataset just said the cheapest format on the platform is the deadliest one.
Words. Just words.
And words are now the one thing you can produce in unlimited supply.
The Sniff Test
Before I show you the agent, you need the rule that makes it work. Because volume without this rule is expensive noise.
People who click ads are skeptical. They've been burned before. They land on your page with a finger hovering over the back button, and for the first three seconds they aren't reading... they're SNIFFING. Does this page smell like the ad I just clicked? Same promise? Same offer?
If the ad says one thing and the page says another, the nose wrinkles and the back button wins. You paid for that click. The mismatch threw it away.
Match the ad to the page and conversions jump. Instantly. No redesign, no new offer, no bigger budget. This was my biggest secret to success in 20 years of running ads and coaching other ad buyers, and it has rescued more accounts than any targeting trick I ever taught. Comfort the clicker, ease their anxiety, remove every ounce of friction between the promise and the page.
Which is exactly why my ad writer agent doesn't start with a blank page or a creative brief.
It starts by READING your landing page.
One URL In, Three Ads Out
Here's what I built (in MindStudio, naturally)...
You paste in one thing: your landing page URL. That's the entire brief.
The agent scrapes the page (copies all the text off it, like select-all-and-copy on autopilot), studies your offer, your audience, your proof, and your promise. Then it writes three complete Meta ads, each 800 to 1,200 words of long-form primary text (the main body copy of a Meta ad, the part above the image), and each built on a different proven direct response framework.
Three genuinely different angles, not one ad in three fonts. And it delivers them as an internal memo with its own analysis and recommendation on top, like a creative director handing work across your desk.
Every ad passes the sniff test automatically, because the ads are born FROM the page, not checked against it afterward. They can't drift from an offer they were extracted out of.
Cost per run: pennies. Time per run: about a minute.
Run it once, you've got 3 ads. Run it again with tomorrow's coffee, 6. An account with fewer than 5 live ad variants is starving Meta's machine, and two runs of this agent means you'll never starve it again.
Remember Motion's medians? Micro accounts launch 2.8 new ads a week while enterprise teams launch 18.9. Six runs of this agent matches the enterprise median. That's six minutes.
And if you're wondering whether AI-written ads can hang with human copy... a study published in January by researchers from Columbia, Harvard, TU Munich, and Carnegie Mellon (run on Taboola's ad network across 500 million+ impressions) measured AI-generated ads at a 0.76% click-through rate against 0.65% for the human-written versions. One platform, native placements, so season to taste. But the direction is not what the AI-slop crowd predicted.
One caution from the other side of the ledger, and it matters. A vendor benchmark of 50,000+ ad variations (methodology unpublished, so treat it as rumor-grade) found AI ads lifting clicks 12% while conversion rates FELL 8% on offers over $100. Clicks up, sales down?? That's the smell of ads overpromising for pages they never read.
Which is the entire reason this agent reads the page first. The sniff test is the difference between volume and expensive noise.
Continue with Build Notes+
This is the free preview. The complete step-by-step guide and supporting files are available to Build Notes+ members.
