« Creative is the new targeting ». No, not really. Here’s how Meta actually uses creative signals to find your audience.
From 2021-2025 privacy changes destroyed explicit signals like interest targeting, so Meta had to find new ways to match ads to the right audience.
Now Meta uses computer vision to identify objects and scenes, and then encodes that info as a numeric fingerprint. The system matches that fingerprint to people who view and engage with similar content.
Your ads no longer need to explicitly target « men 18-45 who like Runner’s World magazine ». Instead Meta sees your video with running in it and infers who might be interested.
No, Meta doesn’t read income, emotions, or brand colors.
No, you don’t need to tag ad copy with « Hey runners… »
Yes, you want lots of creative diversity.
But if ten « different » ad versions look similar to Meta’s learning model and result in the same fingerprint, your ads will get shown to the same people over and over.
Focus on truly distinct variations across scene, setting, talent, use case, and headlines. Learn from which versions of ads get the most delivery, and build on the winners.
Creative is not the new targeting. Creative is the signal Meta uses to infer relevance. Design accordingly.
cc: Nick Shackelford Patrick Coddou David Herrmann Jess Bachman