It Didn’t
On the surface, Facebook has removed direct racial targeting categories. But as the essay reveals, beneath this apparent progress lies a complex network of submerged currents. Advertisers still manage to reach specific racial groups using a range of proxy labels, effectively continuing race-based ad targeting under the guise of legality. Why does this continue to happen? Are Facebook’s reforms truly capable of eliminating discriminatory practices? What does this say about the tension between social justice and technological ethics?
Since 2016, Facebook has faced increasing criticism for enabling racial targeting. Investigations and government inquiries revealed that housing, employment, and financial ads were being distributed in ways that systematically excluded certain racial groups. In response, Facebook announced the removal of tags like “African-American multicultural affinity.”
However, data collected by the Citizen Browser project revealed that advertisers continued to use interests like “African-American culture,” “BET Awards,” or “Black History Month” to reach specific racial demographics.
In other words, while advertisers could no longer explicitly select tags like “Black” or “Latino,” they could achieve almost the same effect by leveraging proxies: culture, language, geography, and media preferences. Legally, this may be skirting the boundaries of discriminatory practice. Technically, it exposes the limitations of Facebook’s strategy of blocking discrimination through surface-level “technical fixes.”
The article also investigates how Facebook assigns interest categories. The platform infers user interests through behavioral data—both on and off Facebook. These inferred categories go far beyond what users manually enter. Facebook’s algorithm takes into account browsing history, geolocation, and purchase patterns, both online and offline.
This means a person may never declare their race, sexuality, or religion, yet still be assigned interest categories that function as effective racial or cultural markers. Interest labels like “Mexican food,” “K-pop,” “Caribbean Carnival,” or “Black music awards” may appear neutral, but for advertisers, they serve as clear indicators of ethnic or subcultural identity.
U.S. federal law prohibits discrimination in employment, housing, and credit. Years ago, investigative reports revealed that Facebook’s ad tools allowed real estate and job postings to be shown only to certain racial groups—excluding Black and Latino users. These practices directly violated laws like the Fair Housing Act and sparked widespread backlash.
Under pressure, Facebook pledged to impose stricter controls for ads in housing, employment, and financial services. But do these restrictions apply across all ad categories? Can advertisers still bypass the rules using subtler methods? The article suggests that while explicit racial labels have been removed, proxy tags—based on food, language, or culture—still allow for effective targeting of the same demographics.
Facebook has long claimed that advertisers are solely responsible for complying with the law, while the platform merely supplies the tools. In practice, this position is often criticized as a way of evading accountability. The platform could intervene—through stronger data filters or rule enforcement—to block discriminatory patterns.
Experts cited in the article argue that removing obvious labels isn’t enough. As long as advertisers are motivated, they’ll find alternate interest combinations to target specific racial groups. Unless Facebook makes its targeting system more transparent and tightly regulated, these workarounds will continue.
Facebook’s targeting system thrives on massive datasets. But data reflects and amplifies existing social inequalities. For instance, if someone frequently follows African-American public pages, listens to gospel music, or joins Black community events, Facebook may assign them an “African-American culture” tag.
This kind of inferred racial profiling enables precise targeting—and becomes even more powerful when integrated into a platform with billions of users. Compared to traditional media, the scope of influence is exponentially larger.
Without robust legal oversight or internal accountability mechanisms, the notion of “technological neutrality” quickly becomes a smokescreen. At Facebook’s scale, “neutrality” can mask immense discriminatory potential.
Facebook has made several public moves to remove “sensitive” tags—from “ethnic affinity” to “multicultural affinity,” and most recently, the disappearance of labels like “Black” or “Latino.”
At first glance, this looks like progress. But as long as user behavioral data remains intact, ad targeting logic can simply pivot to alternative proxies. It’s like locking a front door while leaving dozens of side windows wide open. Or like banning overtaking in a specific lane, but installing countless bypasses—allowing determined drivers to bend the rule.
Researchers and advocacy groups such as Upturn have urged Facebook to release more data about its ad targeting practices. Independent scholars, civil society organizations, and government agencies need access to evaluate what labels are used and which communities are being reached.
Facebook’s current Ad Library offers only limited insight. It lacks disclosure of actual targeting criteria. This makes third-party projects like Citizen Browser—built on voluntary user data—especially valuable and rare as sources of public accountability.
At first glance, using tags like “Afrocentrism” or “Black Girls Rock!” doesn’t necessarily seem discriminatory. An event organizer promoting a Black cultural music festival may reasonably want to reach those most likely to be interested. Not all forms of targeting are inherently harmful.
But when such targeting is used in housing, credit, or employment ads—areas where equal access is legally and ethically critical—it becomes exclusionary. It denies opportunity to certain groups under the guise of efficiency.
The platform and the broader public need clearer, more nuanced rules: In what contexts is group-specific targeting acceptable, and in what contexts does it violate anti-discrimination principles? The answer requires far more than simply deleting a few tags. It demands legislation, governance, and public deliberation.
From a business standpoint, Facebook and similar platforms generate the bulk of their revenue from highly efficient ad targeting. Too many restrictions could reduce advertiser appeal. Too few, and they risk enabling harm.
This is the classic tension between profit and responsibility. When advertising comprises the majority of a company’s revenue, how motivated is it to overhaul the very targeting systems that fuel its success?
The article mentions clients ranging from MyPillow and Family Dollar to Hennessy and government agencies—some of whom denied placing discriminatory ads, while others refused to respond. This reflects a broader misalignment: advertisers seek optimization, while the public demands fairness. Without concrete legal enforcement, platforms have little incentive to cut off their own revenue streams.
In today’s data-saturated world, sensitive attributes like race, gender, religion, and age no longer need to be manually declared. They can be inferred, assembled, and deployed through behavioral footprints.
Even when platforms claim to have removed “explicit” racial tags, the system continues to offer ample pathways for precision targeting. This article exposes a paradox of reform: removing labels often just makes discrimination more covert. The reach remains precise; the exclusion continues.
At a societal level, the existence of racial proxies deepens inequality—especially in access to jobs, housing, and financial resources. This calls for intervention: from public policy, from civil society, and from the platforms themselves. Tech companies must show greater courage in managing and disclosing their targeting infrastructure.
I’m reminded of another article I read recently though I can’t recall which, about a social media platform using “AI risk detection” to monitor suspect users. It, too, embedded the same kinds of bias and discriminatory logic. As long as massive datasets drive ad targeting, all “label removals” risk becoming mere cosmetic gestures.
What we need is a genuine multi-sector, interdisciplinary, and cross-industry system of oversight. Only then can we find a sustainable balance between user privacy and anti-discrimination enforcement. As one expert in the article puts it, stopping discriminatory targeting requires more than technical tweaks. It requires transparency and external review.
The good news is that as long as projects like Citizen Browser continue to uncover these hidden practices, we can keep asking the right questions. In the face of persistent racial proxy targeting, how far must Facebook and the entire industry go before real advertising equity and user rights are achieved?