Broad vs interest targeting in gym ads: the 2026 answer

In 2026, for a local gym, broad targeting wins almost every time. That's not a trend opinion: it's the consequence of Meta's algorithm, given good conversion signal, finding your next member better than your interest list does. The "fitness and wellness" interests you keep checking are outdated, overlap with each other, and describe half the country. There are three cases where interests still make sense, and we'll get to them. But the starting point has shifted: broad is the default, interests are the exception that needs justification.

Anyone who learned Facebook Ads between 2016 and 2021, when building audiences was the craft, finds this hard to accept. Picking "interested in CrossFit, 25-40, who also follow Nike" felt like control, and sometimes it even worked. Today that same setup usually underperforms leaving the interest field blank, and letting go of a skill that took years hurts. Let it go anyway.

Why interests have lost

Three concrete reasons, not one vague one:

First, the data behind interests is worse than before. Between iOS 14, privacy restrictions, and the gradual removal of targeting categories, the "fitness" interest in 2026 is a blurry label built on fewer signals than in 2019. Meta knows this, which is why they've spent years pushing toward Advantage+ and broad audiences.

Second, fitness interests overlap to an absurd degree. "Fitness", "wellness", "healthy living", "training", "gym": stack them and you're not narrowing, you're tagging the same group of people multiple times under different names. Stacking 15 interests produces an audience nearly identical to using none, but with an artificial constraint that takes options away from the algorithm.

Third, and this is the underlying reason: signal is what matters. If your campaign optimizes for leads and you give it enough conversions, the algorithm learns what profile converts in your area and actively seeks it out. Your interest list can only get in the way. The algorithm sees patterns you can't (what pages that person visits, which ads they skip, what time of day they buy) and your "interested in yoga" checkbox is poor information by comparison. This is the same logic running through the 7 optimization levers: quality signal outweighs any manual adjustment.

What actually targets a local gym

Here's the shift most people don't make: your real targeting was never the interests. It's two things you already control.

The first is the geographic radius. A gym sells to people who live or work within 10-15 minutes. That's 3-5 km in a city, a bit more in car-dependent areas. That radius is the most important targeting decision in your entire account, and it's yours, not the algorithm's. Never hand it to Advantage+. Within that radius live maybe 30,000-150,000 people. With an audience that small, every additional interest restriction cuts where there's no fat to cut.

The second is the creative. A yoga class ad shot in morning light attracts yoga people even if the audience is the whole city. A WOD video of sweating athletes attracts people who want that and actively repels people who don't. The creative is the targeting: each piece pre-selects its audience better than any checkbox, because the user self-selects by stopping their scroll. That's why a gym with broad targeting and 5 different creatives is, in practice, more segmented than one with 5 interest audiences and a single ad for all of them.

The operational consequence: the time you used to spend building audiences, spend it producing creatives that filter. Lead quality vs quantity is decided there far more than in the audience panel.

The 3 cases where interests survive

Broad wins almost every time. The "almost" is these three cases:

Niches where the interest is the tribe

CrossFit, martial arts, weightlifting, pole dancing. Here the interest isn't a vague demographic label, it's an identity: someone who has "Brazilian jiu-jitsu" marked either trains or wants to train BJJ, full stop. For a box or a dojo, a niche-specific interest audience can beat broad, especially if your radius includes enough of that niche population. The litmus test: if the interest describes a specific practice and not a generic desire ("crossfit" yes, "get in shape" no), it's worth testing.

New accounts without signal

A freshly created ad account, with no trained pixel and no conversion history, leaves the algorithm flying blind for the first few weeks. There, a sensible interest works as a training wheel: it narrows the search space while the account builds up its first 50-100 conversions. Past that point, open to broad and compare. It's a phase, not a strategy. If you're starting from scratch, the complete Facebook Ads guide for gyms covers that first month in detail.

Exclusions

The forgotten side of targeting. Excluding your current members, recent leads already in follow-up, or a segment you know doesn't convert, that's still manual work and still valuable. Broad targeting decides who to look for; exclusions decide who to stop paying for. Both coexist without issue.

The test that settles the argument for your account

Don't trust this article or any guru: your account has its own answer, and it shows up in three weeks.

The last row is what invalidates most tests. It's common for interests to show a slightly lower CPL while broad delivers better members: the unconstrained algorithm sometimes finds people who cost more to acquire but are more willing to pay a subscription. If you judge after 5 days by CPL, you'll kill the real winner. Hold the 3 weeks and follow the lead all the way to the visit or signup.

The mistakes we see every week

Stacking 20 interests. Piling "fitness + gym + training + healthy living + Adidas + Nike + running" doesn't narrow anything: it produces a huge, blurry audience with the added downside of constraining the algorithm. If you use interests, use one or two specific ones, not a buckshot approach.

Audiences of 8,000 people. Cross-referencing interests with "and must also match" until you're left with a few thousand users guarantees frequency spiking within days and premium CPMs. Locally, an audience below 50,000 people is a red flag, not precision targeting.

Lookalikes in small cities. A 1% lookalike of a whole country spreads those people across the nation; intersected with your 4 km radius in a city of 60,000 residents, you're left with almost no one. Lookalikes were built for e-commerce with national reach. In a neighborhood business, the radius already does that job, the lookalike only shrinks what shouldn't be shrunk.

Changing audiences every week. Every targeting change restarts part of the learning phase. The owner who rotates audiences looking for "the right one" keeps their campaign in perpetual learning mode and concludes that "Facebook doesn't work." Pick a structure, give it 3-4 weeks, and evaluate with data.

And the inverse mistake, less often discussed: switching to broad without having anything else in order. Broad with a single generic creative and a pixel with no data means giving the algorithm a blank map. If your account converts rarely and your ads are interchangeable with the gym next door, opening the audience fixes nothing, first creatives that filter, then open up.

What to do this week

If you've had active campaigns for more than six months and you're still on interests, set up the 50/50 test today: it's an hour of work and the answer is worth months of budget. If your account is new, start with a sensible interest and plan the move to broad once you have signal. And in both cases, redirect energy from audiences to creatives, that's where this is won: the campaign structure for gyms explains how to organize ad sets and creatives so the ad does the filtering.

One final note of pragmatism: broad targeting demands more from the rest of the system. Without good creatives and clean conversion signal, "broad" just means "lost in a bigger space." Platforms that automate the full cycle (Pilotium among them) work exactly this way: radius + broad + rotating creatives + quality signal, because that's the configuration Meta in 2026 rewards. If you're running it manually, the order is the same.