How Meta Audiences Work Now
Almost every “Meta targeting explained” article you'll find describes 2021: stack interests, narrow the audience, launch. That playbook is largely dead. Here's how targeting actually works in 2026 — and why it's good news for small businesses.
Targeting in 2026 shifted from picking an audience to signalling one and letting Meta’s AI find it. Only location, language and minimum age are hard limits; interests, custom audiences and lookalikes are suggestions the system can exceed.
01The one shift that changes everything
You used to tell Meta exactly who to show your ad to. Now you suggest, and Meta's AI decides. Its targeting engine processes an enormous number of signals for every single impression — far more than any human could. Against that, a hand-picked list of five interests is a rounding error.
So the job has flipped. It's no longer “which interests should I choose?” It's “what's my goal, how much data can I feed the system, and how strong is my creative?”
Your inputs now split into two groups. Hard controls (location, language, minimum age, exclusions, Special Ad Category) genuinely limit who sees the ad. Suggestions (age, gender, interests, custom audiences, lookalikes) merely steer the AI — it can and will go beyond them to get you results.
02The four audience types
| Type | In one line |
|---|---|
| Core / detailed | Define by location, age, interests and behaviours (next lesson). |
| Custom | People who already know you — your first-party data (Lesson 4.3). |
| Lookalike | New people who resemble your best customers (Lesson 4.4). |
| Advantage+ | Meta's AI finds the audience from your signals (Lesson 4.5). |
03How big should an audience be?
Meta generally recommends 2 to 10 million people for a starting audience. This surprises people who were taught to narrow aggressively — but tiny, over-specified audiences now raise your costs and cause learning-limited delivery. Give the algorithm room to work.
04The 2026 playbook in one sentence
Broad targeting plus a strong conversion signal beats narrow targeting plus a weak one — and the signal that matters most is your pixel/CAPI data and your creative, not your interest stack. Everything in this module builds on that idea.
Most experienced advertisers now run roughly 70–80% on broad Advantage+ campaigns, 10–20% on retargeting warm custom audiences, and 5–10% testing lookalikes or interests. We'll unpack each piece as the module goes on.
If a guide tells you to stack ten interests and narrow until the audience meter turns yellow, it's teaching a game Meta stopped rewarding years ago. Judge every targeting tip by whether it fits the hard-controls-vs-suggestions reality above.
- Targeting shifted from picking an audience to signalling and letting Meta find it.
- Hard controls limit delivery; suggestions only steer it.
- Four types remain: core, custom, lookalike, Advantage+.
- Broad + strong signal beats narrow + weak; aim for 2–10M, not tiny audiences.
Frequently asked questions
How does Meta ad targeting work now?
You provide signals rather than fixed rules. Hard controls like location genuinely limit delivery, while suggestions like interests only steer Meta’s AI, which can deliver beyond them.
What are the four Meta audience types?
Core/detailed (location, age, interests), custom (your first-party data), lookalike (people like your best customers), and Advantage+ (the AI finds the audience from your signals).
How big should my Meta audience be?
Meta generally recommends 2 to 10 million people. Tiny, over-narrowed audiences now raise costs and cause learning-limited delivery, so give the algorithm room.
Return to the Meta Ads course hub for the full curriculum.