The Briefing
Meta shipped its own image model, Muse Image, straight into Advantage+ this week. It reads your creative brief, swaps single elements, and spits out on-brand variations with fewer round trips. Free image labor is now baked into the ad platform.
The question isn't "can I use it." It's "does it beat what I've got."
Meta also started auto-labeling ads made or edited with third-party AI tools, using hidden C2PA metadata. If your ad uses a photorealistic AI person, the "AI info" tag now sits right next to "Sponsored." Assume your buyer can see how the sausage was made.
Quietly, Meta removed a privacy opt-out that grows your retargeting pools, and added a 2 to 5 percent fee on ads in six countries. Two small moves. Both hit your cost per result. Check your numbers before you blame your copy.
And Meta's Business Agent now runs for over 1 million businesses on WhatsApp and Messenger, answering questions and qualifying leads while you sleep. Handy. Also one more tool screaming for your attention this week.
The Build
This week I did the boring thing. I killed 14 agents and replaced them with 2. Higher quality. Less noise. Fewer things to babysit.
Then I built one new skill. I saw a post on Facebook about ad writing that actually had merit, so I turned it into a real tool and tested it.
The idea is simple. Line up 100 people who want your thing. They are not all buying for the same reason. Some buy to stop pain. Some buy to become someone new. Some buy out of fear. Some to belong. Some for an edge. Some because they've tried everything and got burned.
So you never write "an ad." You write an ad for one buyer, in that buyer's words, with the guarantee that removes that buyer's exact risk. Six buyers. Six registers. Mix them and the ad reads as "not for me."
This sits right on top of my old ad structure (Hook, Stack, Payoff, CTA). The structure builds the ad. The buyer avatar decides who it's talking to. Here's the prompt I ran:
Write 3 Facebook ad variants for [my offer]. One variant each, for these buyers: Pain-Avoidance, Last Resort, Optimization. For each one: - headline in THAT buyer's register - a hook that matches it - a CTA in that buyer's action words - a guarantee that removes THAT buyer's risk Then run the 6-point pre-launch checklist on each. Flag any "register bleed" and fix it.
Out came three ads for the same offer that read like three different people wrote them. One promised fast relief. One led with proof for the burned buyer. One offered an edge to the buyer already winning.
What worked: The audit caught a relief headline stapled to a "start your journey" CTA. Register bleed. I'd have shipped that ad without ever seeing the mismatch. That catch alone earns the skill a place.
What didn't work: It wants to write all six buyers every time. Most offers have 1 to 3 real buyers, not 6. Six ads spreads your budget thin and half of them talk to nobody. I had to force it to pick. And it can't tell me which buyer my market actually is. Only the data does that.
Which is why I'm not celebrating yet.
The ads are live. I'm waiting on the numbers. I don't trust a thing until the test comes back.
The Play
If you're busy, stretched thin, or a little addicted to every new AI toy, hear this: AI is a tool, not the answer. The test is one question. Does this get a bigger, better result more efficiently? If not, kill it.
Run that test on your ads today. Take one ad you're running now. Ask AI to tell you the truth about it:
Here's my ad: [paste it]. Which single buyer does it actually speak to: Pain-Avoidance, Aspiration, Fear, Identity, Optimization, or Last Resort? Where does the register "bleed" — a headline for one buyer, a CTA for another? Show me the exact lines. Then rewrite it for ONE buyer, clean.
If the headline sells relief and the CTA sells a journey, you found your leak. Fix it to one buyer. Then let the data decide, not your gut.
Try this today. Audit one live ad with the prompt above, rewrite it for a single buyer, and put a date on your calendar to check the result. No date, no test. No test, no trust.
Growth Partner for founders stuck between $1M and $20M | I find the money you're sitting on and build the machine that collects it | $1B+ in client results.
Growth Partner for founders stuck between $1M and $20M. I find the money you’re sitting on and build the machine that collects it. $1B+ in client results.