For a decade, media buyers earned their keep building audiences. Interest stacks, lookalike percentages, layered exclusions, the whole connoisseurship of the Meta ads manager circa 2017. I built plenty of them, and I was proud of some. That craft is now mostly archaeology. The platforms took targeting back, the algorithms run broad, and the settings screen where skill used to live has maybe four decisions left on it.
It is worth being precise about why this happened, because the reason tells you it is not reversing. Signal loss from privacy changes degraded the data those handcrafted audiences were built on, and the platforms' machine learning got better at finding converters from conversion data than any human got at describing them with interest categories. Meta's own advice flipped from celebrate the targeting options to leave the targeting alone, and the advice is self-interested and also correct: in test after test, broad delivery with strong conversion signal beats the artisanal audience. The levers were not taken from us because we misused them. They were retired because the machine outgrew them.
The skill moved into the ad
The skill did not disappear. It moved. When distribution goes broad, the ad itself becomes the audience filter: the algorithm shows your creative around, watches who leans in, and finds more of those people. Which means the hook, the first frame, the claim, and the offer are now doing the job the audience settings used to do. A campaign with generic creative and broad targeting is not targeted at all. It is a flyer in a hurricane.
This changes what good creative means, and it is not what most brand teams optimize for. Creative that targets is specific enough to repel the wrong people, and repelling is the underrated half of the job. The mortgage ad that opens with the monthly payment qualifies harder than any income filter ever did. The B2B ad that names the job title in the first line is doing audience selection in public. The skincare ad that names the exact condition finds the person who has it and loses everyone else, which is precisely the trade you want. Brand teams trained to maximize appeal will fight this, because their instinct is creative that offends no one, and creative that offends no one selects no one. The words were always upstream of the media; broad targeting just raised the stakes on it.
Portfolio over masterpiece
Practically, the portfolio matters more than the masterpiece. You need enough genuinely different angles, different hooks, different customer problems, different proof points, for the system to sort people with, because five polished variations of one idea give the algorithm one signal five times. The distinction that matters is concept versus variation: a new color treatment is a variation, a new customer problem is a concept, and the algorithm can only discover audiences you gave it concepts to discover them with. The best-run accounts I see maintain a deliberate concept map, each ad angle tagged to the customer segment it is meant to catch, so the creative portfolio is doing knowingly what audience settings used to do explicitly.
And read creative reports the way you used to read audience reports, because that is where the targeting insight now lives. Which concept wins is market research the market paid you to run: the hook that outperforms is telling you which problem your customers believe they have, which is frequently not the problem your positioning deck says they have. I have watched a losing product narrative get rewritten because one ad angle beat the house creative three to one, and the rewrite outperformed everywhere, in email, on the site, in the sales deck. The feedback loop runs both directions if anyone is reading it.
Here is the reframe I offer buyers who mourn the old levers: we gave up settings and got a feedback loop, and the loop is worth more. Audience settings told you who you guessed your customer was. Creative-as-targeting tells you who your customer actually is, continuously, priced in media dollars you were spending anyway. The message is the media plan now. Write it like one, staff it like one, and read its results like they are the strategy document, because they are.