Broad vs interests on Meta Ads: which should you use in 2026?

For years, a big part of running Facebook Ads was finding the right interests. Surf, beauty, competitors, job titles, software: they were proxies for describing the ideal customer by hand, on a simple logic. The more precisely you describe the audience, the more precisely Meta can reach it.

That logic made sense, and in some situations it still does. But the ad system today behaves differently, and the useful question is no longer whether interests still exist, because they do, but how much manual control over the audience you actually need. Below: what changed, when starting broad is the most rational choice, and when narrowing still has a reason.

Why we built audiences from interests for years

When delivery knew little about an account, the audience was one of the few ways to give it direction. With little data and generic creative, an interest worked as a shortcut for telling Meta where to look. It also suited account management: one ad set per interest let you read what worked, assign budget by segment and keep tidy reports.

The limit was always less visible. An interest describes something a person follows or has engaged with, not the moment they are ready to buy. Someone who follows surf content is not necessarily shopping for a wetsuit, and someone who follows a competitor is not necessarily a customer. The perceived precision was often higher than the real one.

What changed in Meta targeting

Meta now separates controls from suggestions. With Advantage+ audience, controls are the limits delivery will not cross: location, language, minimum age and custom audience exclusions. Everything else, meaning age, gender, detailed targeting, custom audiences and lookalikes, can be added as a suggestion. Meta prioritizes people who match it, but when it expects a better result it will also show ads to others.

Detailed targeting has not disappeared, and neither has the option to narrow. But an interest added as a suggestion no longer works as a fence: it becomes a starting hint. The exact behavior depends on the objective and the ad set configuration, so check it in your own account rather than assuming it.

What working broad really means

Working broad does not mean showing ads to anyone at random. It means not narrowing the audience with interests and letting delivery choose who to reach, inside the limits that have a real business reason: country, language, minimum age, exclusions.

Selection still happens. Meta decides who to show an ad to based on the event you optimize for, interaction history, the content of the ad and context. What changes is who does the selecting: no longer the advertiser through a list of interests, but the system, using more signals than any manual list could describe. Broad is not a way of giving up criteria. It moves them to what you can control better.

Why signals, objective and creative matter more now

If the audience no longer describes the customer, something else has to. The first thing is the optimization event, meaning what you ask Meta to look for. Optimizing for an action that is easy to take but rarely followed by a sale leads the system to find exactly those people, whatever audience you set.

The second is tracking quality. Pixel and Conversions API raise the quality of the inputs delivery can use, provided events arrive complete and without duplicates. How Meta’s models select ads is covered in Meta Andromeda and the other AI models, so the principle is enough here: the cleaner the signals, the less you need to steer delivery by hand.

The third is creative. The message helps decide who responds, because people who recognize themselves engage and people who feel distant do not. That does not make creative a replacement for targeting. It becomes one part of it, alongside objective and signals.

Broad vs interests: strengths and limits

Neither column wins outright. The comparison shows what each choice costs: with broad you give up control in exchange for space and concentrated data, with interests you gain direction and pay in fragmentation and possibly unnecessary restrictions.

When to start broad

Starting broad is the most rational starting point when there is no concrete reason to narrow. Some conditions strengthen the case: an optimization event close to business value and properly tracked, conversion history in the account, a budget that does not force everything to be split, and creative varied enough to speak to several purchase motivations.

These make broad more sensible. They are not prerequisites. There is no universal weekly conversion threshold after which broad starts working, and a new account does not have to start with interests. When signals are scarce, the more useful answer is usually to work on the signals, not to add a list of interests.

If a broad campaign does not perform, the cause is not automatically tracking and not automatically the absence of interests. It can be the offer, the creative, the site or the chosen event. As with advanced targeting in general, first work out where the path breaks, then decide whether narrowing the audience is the right lever.

When interests still make sense

Interests still have real use cases, and none of them is automatic. The first is testing a hypothesis: if you believe a specific segment responds differently from the rest, a suggestion can help verify it, with a clear question to start from. The second is genuinely narrow niches, where the offer addresses an identifiable group of people and an initial direction can help, in some B2B contexts for example.

The third is an account with few signals: with no history and still-poor tracking, a suggestion can give delivery a starting point. It can, not it must, and the choice should still be compared against the broad version. If the constraint is real, such as a geographic area or a minimum age, the right way to impose it is the controls, not an interest Meta may expand.

In every case the same rule applies: the restriction should have a reason you can write in one sentence.

How to test them and what I would pick as the default

A broad versus interests comparison says little if audience, creative, budget, period, offer or optimization event change together. To learn anything you need to keep the objective, optimization event, creative, offer, geography, attribution and period constant, and vary only the audience approach. It is the same discipline as creative testing, applied to a different variable.

When it is available and appropriate, a controlled A/B test beats comparing campaigns that ran at different times. Meta’s tool splits the audience so nobody sees both versions, while switching campaigns on and off by hand can produce unreliable results.

As a default, I start broad. I narrow when there is a concrete reason and a test that puts it under pressure, not because a field is there to be filled.