Detailed Targeting

Detailed targeting does not work as a hard filter the way it can look on the surface. Interests, behaviors, and demographic attributes you add function as a seed the delivery system studies to understand who is likely to respond, and how strictly Meta honors that seed depends heavily on whether Advantage+ audience, previously called detailed targeting expansion, is switched on for the ad set. Within a single targeting row, multiple criteria combine with OR logic, adding a second interest broadens the audience rather than narrowing it, while narrowing only happens when you deliberately add a second row using the Narrow Audience or must also match option, which applies AND logic between rows. As of March 2025, Meta removed the ability to exclude an audience by detailed targeting criteria, interests, behaviors, or demographics, across every ad set; that restriction previously applied only to Special Ad Categories, housing, employment, credit, and social issues, under a 2022 settlement, and Meta extended it account-wide. You can still exclude a Custom Audience directly, existing customers or past purchasers, for instance, that mechanism is untouched, but excluding by an interest or behavior specifically is no longer an available option anywhere in the interface.

The single biggest interaction is with Advantage+ audience. With it enabled, which is the default state on most ad sets created today, your detailed targeting selections act as a starting point Meta can deliver outside of when its own model predicts better results elsewhere, so the practical effect of any given interest shrinks the more purchase or lead signal the ad set accumulates from its optimization event. Special ad category status is the other major interaction: flagging an ad set as housing, employment, credit, or a social-issue campaign strips out most demographic and many interest and behavior options entirely, not just the exclusion mechanism, leaving a much shorter approved list to build from. Detailed targeting also loses relative influence as an ad set matures under a well-fed optimization event like Purchase Optimization, since a model with enough of its own conversion data to lean on needs the manually stacked interests less than a brand-new ad set with no signal history does.

Detailed targeting still earns its place when the delivery system genuinely has no signal to work from, a new account, a new product line with no purchase history, or an audience the pixel and Conversions API data simply have not seen yet, and it remains the only lever for reaching audiences defined by explicit interest or behavior rather than by algorithmic prediction. It matters less on accounts with real optimization-event volume and Advantage+ audience turned on, where broad targeting constrained mostly by age, location, and language guardrails frequently performs comparably or better, because the delivery algorithm's own modeling has more to work with than a manually assembled interest list does.

The most common mistake is treating a saved audience as a fixed asset that performs the same way indefinitely; Meta periodically retires specific interest and behavior options without a loud notification, so an ad set built around a two-year-old audience can silently lose part of its seed with no visible warning beyond the audience definition quietly changing. A second mistake is stacking interests into the same row assuming it narrows the audience, when OR logic within a row does the opposite, broadening reach with every additional interest added; narrowing requires the separate Narrow Audience step. A third mistake, common since the March 2025 change, is spending time hunting for an interest-exclusion option that used to exist, troubleshooting a setup problem when the setting itself has simply been removed platform-wide rather than misconfigured.

Because Advantage+ audience expansion means the people an ad set actually reaches can drift from the interests and behaviors originally selected, check the Breakdowns by age, gender, and placement rather than trusting the saved targeting summary as a report of who the ad set reached. If the goal is to learn whether a specific interest layer is adding value at all, the only reliable read is a controlled split test, one ad set with Advantage+ audience expansion on, one with it off, or one built entirely on Broad Targeting, since cost per result alone can look nearly identical between two setups whose underlying audience composition is meaningfully different.