Your Creative Is Your Targeting: How Meta Actually Decides Who Sees Your Ads

Two ads. Same ad set. Same broad targeting, no interests, no lookalikes, nothing touched at the audience level. One spent 94% of its budget reaching men. The other spent the clear majority of its budget reaching women.

Nobody told Meta to do that. The targeting boxes were identical. The only thing that differed was what was on the screen.

That one number is the whole argument of this post. If you understand why it happens, you stop fighting Meta's delivery and start steering it - and you do it with your creative, not your settings.

Where the targeting actually happens

Most founders still picture targeting living at the ad set level. You set an age, a gender, maybe an interest, and that box decides who sees the ad. That model is years out of date.

The targeting happens at the ad level. And not just the ad level - each individual ad builds its own audience.

When you upload an ad, Meta reads it. Hundreds, thousands of signals: the visual, the on-screen text, the body copy, the hook, the pacing. Roughly 80% of that read comes from the creative itself. From all of that, it forms a guess about who this specific ad is for, shows it to a small sample, and watches. These people scrolled past. These people stopped. These people bought. So it shifts toward the ones who behaved like buyers, and away from the ones who didn't. It's doing this constantly, even when the ad just says "active" and looks like it's sitting still.

So inside one broad ad set, every ad is quietly assembling a different audience. One ad might appeal to a wide slice of people and comfortably spend big. The ad next to it might resonate with a narrow group and never scale, no matter how long you leave it. Same settings. Different creative. Different audience. That's the mechanism the 94/6 split is showing you.

Meta's newer delivery system leans into this even harder. The feed is being optimised to be as relevant as possible to each person, which means ads are getting sub-segmented down to who's actually in them and what they talk about. Put a plumber, doing plumber things, talking about plumbing, into an ad, and it finds plumbers. The flip side: cost per result tends to drop as the match gets tighter, because relevance is cheaper to deliver.

Your creative is a casting call

Here's the thing - if the creative does the targeting, then every creative decision is a targeting decision, whether you meant it that way or not.

Who's on screen is a targeting decision. Put a woman in a golf skirt in the ad and you'll skew male, because of who stops to watch. Put a 60-year-old in the ad and you'll skew older. Show a cross in the background and talk about faith, you'll reach more faithful people. The message matches the prospect, and the prospect is who shows up.

The words on the screen are a targeting decision too, and this is the part you can control most directly. When your hook calls out a specific person, that person stops and everyone else keeps scrolling. Someone described this as a dog whistle, and it's the right image. "This is for my girlies who need a girls' night soon" is a dog whistle. The people it's for hear it instantly. Nobody else does. And the moment the right people stop and engage, Meta starts building the audience around them.

So the brands that upload a clean product photo with no copy, or a pretty video that says nothing specific, are throwing away their single best targeting tool. They've left the casting call blank and then they're surprised Meta cast the wrong people.

The four questions that make a call-out land

You can't write a sharp call-out from your logo and a vibe. You write it from understanding the person you want Meta to find. I run every brief through four questions before a word goes on screen.

• What do they actually desire? Not the product - the outcome they're chasing. More energy, a flatter stomach, a gift that makes them look thoughtful, a pre-workout that doesn't give them the jitters. Name the desire and you've got your hook.
• Where do they start the conversation? Some people already know the solution they want and are searching for it. Others just have a problem and no idea what fixes it. You speak to those two people differently. One needs "the easiest greens you'll actually drink"; the other needs to be told the mess and the taste have a fix at all.
• How many products have they tried before yours? A first-timer and someone who's been burned twice need different openers. "A pre-workout without the jitters" only lands for the person who's already had the jitters and hated them. That line is doing the targeting - it's looking for the burned buyer.
• What else is true about them? Not "people into fitness". Go specific: a 25-year-old who trains bodybuilding-style, wants lean muscle, values owning less, and is grinding to be a better version of himself. The more exactly you can describe them, the more exactly your creative can call them out.

Answer those four honestly and the on-screen line writes itself. You're not being clever. You're being clear enough that the right person can't misread who the ad is for.

A practical note on how this rolls up: in a broad, dynamic setup, you let Meta control the budget and mix the creative combinations, and you stack ads built around different avatars. Different angles will deliver differently - one might spend a little at a strong return, another might spend a lot at a slightly lower return. The instinct is to pour everything into the highest-return one. Resist it. Because each angle is reaching its own audience, they're not really stealing from each other. As long as an angle clears the return you need, it earns its place. That's how you scale without collapsing back onto one fragile winner.

A hook rewrite that moved delivery without touching a setting

Let me make this concrete with an anonymised example. Numbers and details are invented to illustrate the point, not a guaranteed result.

A supplement brand we worked with had a hero ad for a collagen product. The hook on the video was soft and product-led: something like "Meet your new daily collagen." It was spending, but the cost per purchase had drifted up to around $58, and when we pulled the delivery breakdown, the ad was landing on a broad, slightly younger audience that wasn't converting well. The brand assumed the fix was audience-side. They wanted to add interests, build a lookalike, narrow the age. Classic settings thinking.

We didn't touch a single targeting box.

Instead we rewrote the on-screen hook to call out the actual buyer. The real customer wasn't a 25-year-old curious about supplements. It was a woman in her 40s who'd noticed her skin changing and was quietly bothered by it. So the new hook said as much, plainly - the kind of line that makes a 44-year-old stop and a 24-year-old keep scrolling. Same product, same video footage, same offer, same ad set.

Within about ten days the delivery had shifted noticeably toward women in their 40s, the people who'd actually buy. Frequency on the wrong audience dropped because Meta stopped wasting impressions on people the hook now repelled. Cost per purchase settled back to roughly $39. Nothing changed in the campaign structure. The creative re-cast the audience, and the account followed.

That's the whole lesson in one move. When delivery is wrong, the first place to look isn't the targeting tab. It's the first three seconds of your ad.

How to read your account through this lens

Once you accept that creative is the targeting, your reporting habits change. A few that have served us well:

• Check the delivery breakdown on your top ads. Look at the age and gender split per ad. If an ad is reaching the wrong people, the creative is mis-cast - the answer lives in the hook and the visual, not the settings.
• Use frequency and CPM as a read on audience size. A high CPM and a climbing frequency on a specific ad usually means it's resonating with a small pocket of people and you're hammering them. That's fine if it's hitting your numbers, but it tells you the creative is narrow.
• Make more ads, sliced finer. Instead of one ad for "plumbers", make ten for different slivers of plumbing. Cost per result tends to fall as the call-out sharpens, so you can afford to burn through more of them. That's the nature of it now.
• Feed in your organic winners. A piece of content that already pulled the right people organically has effectively pre-cast its audience. Dropped into the account with a simple call-to-action overlay, it often keeps doing exactly that.

The settings are not where the fight is anymore. Meta's machine is better at finding buyers than any audience you'd hand-build, and it makes its decision by reading what you put in front of it. So the real question on every ad isn't "who am I targeting". It's "who did I just cast - and is it the person who buys"?