← Back to desktop
Meta Ads

Meta Ads Audience Targeting in 2026: Broad, Lookalikes, and When to Use Each

NUVIX · 28 January 2026 · 12 min read
TLDR: Meta’s targeting has shifted massively. Broad targeting (Advantage+ audiences with minimal restrictions) now outperforms detailed interest stacking in most accounts. Lookalike audiences still work but they’re best used as a signal rather than a hard boundary. The algorithm is better at finding buyers than you are at guessing who they are. Feed it good creative, good conversion data, and get out of its way.

How Meta Targeting Used to Work

In the early days of Facebook Ads, targeting was the whole game. You’d build audiences by stacking interests, behaviours, and demographics. You’d target women aged 25-34 who liked yoga, followed certain brands, and had recently moved house. The more specific your audience, the better your performance. Audience research was the skill that separated good media buyers from bad ones.

That world is gone.

iOS 14.5 broke a huge chunk of Meta’s tracking infrastructure. Interest categories became less reliable as Apple restricted the data Meta could collect. At the same time, Meta invested heavily in machine learning that could find your ideal customer without being told exactly who they were.

The result: detailed targeting is less accurate than it used to be, and Meta’s algorithm is better at finding buyers than human media buyers are at guessing who they might be. The platforms have shifted from ‘tell us exactly who to target’ to ‘give us a good signal and we’ll figure it out’.

If you’re still building audiences the way you did in 2019, you’re leaving performance on the table.

The Three Approaches

1. Broad targeting

Broad targeting means giving Meta minimal audience restrictions. You set your country, maybe an age minimum if your product requires it, and let the algorithm find the right people based on your creative and conversion data.

This feels uncomfortable. You’re trusting a machine to spend your money on the right people with almost no guidance. But in most accounts we manage, broad targeting delivers the lowest CPA and the best scale.

Why it works: Meta has billions of data points on its users. When you set a conversion objective and feed it data on who actually buys, the algorithm learns to find patterns you’d never identify manually. Maybe your best customers share browsing behaviours, app usage patterns, and purchase histories that no interest category captures. The algorithm sees these patterns. You can’t.

When broad works best:

When broad struggles:

2. Lookalike audiences

Lookalikes take a source audience (your customer list, website visitors, or people who’ve taken a specific action) and find people who share similar characteristics. You choose a percentage — 1% is the closest match, 10% is the broadest.

Lookalikes used to be the gold standard. In 2026, they’re still useful but the dynamics have changed. Meta now treats lookalikes more as a suggestion than a strict boundary. Even if you set a 1% lookalike, Advantage+ audience expansion can push beyond that boundary if the algorithm thinks it can find conversions elsewhere.

Best practices for lookalikes in 2026:

3. Interest and behaviour targeting

Detailed targeting lets you reach people based on interests, behaviours, job titles, and other demographic data. This is what most people think of as ‘Facebook targeting’.

It still has its place, but that place is shrinking. Interest categories are less reliable post-iOS 14. Someone who liked a fitness page five years ago gets put into a ‘fitness interest’ category even though they haven’t touched a gym since. The data decays and there’s no way to know how fresh it is.

When interest targeting still makes sense:

The mistake most advertisers make is stacking too many interests. Adding 15 interests to an ad set doesn’t make it more targeted — it makes it broader, because Meta uses OR logic (people who match any of the interests). If you want precision, use fewer interests, not more.

The Advantage+ Shift

Meta’s Advantage+ suite is their push toward full automation. Advantage+ shopping campaigns, Advantage+ audience, Advantage+ creative — they’re all designed to give the algorithm more control.

Advantage+ audience is the big one for targeting. When you select it, Meta uses your targeting inputs as suggestions rather than boundaries. You might set a 1% lookalike, but Advantage+ can go beyond that if it thinks it can find conversions elsewhere. You might set interest targeting, but the algorithm can expand past those interests.

This is a fundamental shift. You’re no longer defining who sees your ads — you’re providing a signal that the algorithm uses as a starting point.

For most advertisers, the best approach is to set your audience suggestions (a lookalike or basic demographics) and let Advantage+ expand from there. You get the benefit of a starting signal without the rigidity of a hard audience boundary.

The caveat: Advantage+ requires conversion volume. If you’re getting fewer than 20-30 conversions per week per ad set, the algorithm doesn’t have enough data to expand effectively. In that case, tighter audiences with Advantage+ turned off may still perform better.

Custom Audiences: Your Retargeting Foundation

Custom audiences are the one type of audience that hasn’t lost its power. These are audiences built from your own data — customer lists, website visitors, app users, and people who’ve engaged with your content on Meta.

The main custom audience types:

Custom audiences are essential for retargeting, which typically delivers the lowest CPA of any campaign type. They’re also the best source data for lookalike audiences.

The one rule: keep your custom audiences fresh. A 180-day website visitor audience includes people who’ve long since forgotten about you. For retargeting, 30-60 day windows tend to perform best for most businesses.

Testing Audiences Properly

The testing structure

To test audiences fairly, you need to isolate the variable. That means running the same creative to different audiences at the same budget level. If you test broad targeting with one creative and a lookalike with different creative, you don’t know whether the performance difference is about the audience or the ad.

Set up a testing campaign with:

What to test

Start with the big comparison: broad versus your best performing targeted approach. Run broad (country + age minimum only) against a 1% purchase lookalike against your best interest stack. Give each at least £500-1,000 in spend or two weeks — whichever comes first.

Then test within the winning approach. If broad wins, test broad with different Advantage+ audience suggestions. If lookalikes win, test different source audiences (purchasers versus leads versus website visitors).

What to measure

The only metric that matters for audience testing is cost per acquisition or return on ad spend. Not CPM. Not CTR. Not CPC. Those are diagnostic metrics that can mislead you. A broad audience might have a higher CPM than an interest audience but deliver a lower CPA because the algorithm is finding higher-intent users within that broader pool.

Give tests enough time and budget for statistical significance. If one ad set has 3 conversions and another has 5, you don’t have enough data to declare a winner. Wait until you have at least 20-30 conversions per ad set before drawing conclusions.

The Audience Strategy That Works for Most Businesses

Based on what we see working across dozens of accounts in 2026, here’s the structure we recommend for most businesses:

That’s it. Two layers. Prospecting to find new people, retargeting to convert warm audiences. The prospecting layer generates the volume. The retargeting layer improves the economics.

You don’t need 12 ad sets targeting different interest combinations. You don’t need a complex hierarchy of lookalikes at different percentages. Simplicity wins because it gives the algorithm more data per ad set and reduces audience overlap.

The Uncomfortable Truth

If your Meta Ads aren’t performing, the problem is almost certainly not your targeting. It’s your creative or your offer.

In 2026, the creative is the targeting. When you run broad, the only thing differentiating your ad from every other advertiser trying to reach similar people is the ad itself. If your image is generic, your copy is bland, and your offer is the same as everyone else’s, no amount of audience sophistication will save you.

The best advertisers on Meta are spending 80% of their time on creative development and 20% on campaign management. They test new creative constantly — new angles, new formats, new hooks. They treat audience setup as a 15-minute task and creative development as an ongoing process.

If you find yourself spending hours tweaking audience settings and minutes on creative, you’ve got the ratio backwards. Fix the creative. The algorithm will handle the rest.