Advantage+ optimizes relentlessly. The question is: for whom?

Meta's automation is genuinely good at its job. The nuance is that its job is maximizing conversions attributable to Meta — which overlaps with maximizing your profit, but isn't identical to it. The gaps between those two objectives are where operators earn their keep. Surface by surface:

Advantage+ audience: use it (you barely have a choice)

Broad delivery with audience "suggestions" Meta may expand past. In head-to-heads we run, broad + strong creative beats interest stacks for prospecting — Meta's purchase-intent signal is simply better than your guesses about people who "like fitness." Telling detail from inside the API: Meta now requires every new ad set to explicitly declare the Advantage-audience flag — ad sets without it are rejected outright. That's not a feature launch; that's a roadmap. Manual targeting is being deprecated in slow motion. The strategic consequence: creative is the targeting now. Your hook selects the customer; the ad set just sets the budget.

ASC: powerful, with a mirage setting

Advantage+ Shopping collapses structure and works well at ~30+ purchases/month. The catch: left on defaults, ASC happily spends on people who already know you — past visitors, past buyers — and books their easy conversions as campaign performance. Blended ROAS looks brilliant; your actual new-customer acquisition is thinner than reported. The fix is one setting everyone skips: cap "existing customer budget" (0–5% for genuine prospecting), define existing customers via your lists, and judge ASC on new-customer metrics. Automation isn't cheating; it's doing exactly what "maximize conversions" says — harvesting the easiest ones first.

Auto-enhancements: the ones editing your ads without asking

Several ship on by default and re-edit creative at delivery time: auto-crops, brightness/contrast "improvements", generated overlays, animated stills, template frames. For designed brand creative, these frequently make ads worse — and you'd never know, because you approved a different image than users saw. From working the API directly: the opt-outs are per-enhancement (image animation, product-browsing modules, profile-card frames, auto-translated overlays, native subtitles — each its own switch; there's no honest master toggle, and the umbrella switch that used to exist is gone). Audit them per ad, kill the visual edits on brand creative, A/B the rest.

Where automation genuinely cannot help

Low signal: under ~30 purchases/month or with a shaky pixel, automation amplifies noise confidently. Fix tracking and volume first. Offer economics: no delivery system rescues a breakeven ROAS your margin can't clear. Objective choice: optimizing purchases vs value vs leads is a business decision the machine executes but can't make.