A special ad category is a label that an advertiser on Meta's platforms must attach to a campaign when the ad promotes housing, employment, financial products and services, or social issues, elections or politics. Declaring a campaign in one of the first three categories switches off the targeting tools most likely to exclude people on the basis of protected characteristics: age, gender, postcode and lookalike audiences among them. The political label works differently, triggering identity checks and disclosure rather than targeting limits. The concept exists because US civil rights law forbids discrimination in advertising for homes, jobs and credit, and because Facebook's targeting system was shown to make such discrimination easy. Google, TikTok and Pinterest run comparable rules, but the term is Meta's.

How the declaration works

The label is set at campaign level, by ticking a box at the start of campaign creation in Ads Manager. Through the Marketing API, the same choice travels in the special_ad_categories field, an array that accepts HOUSING, EMPLOYMENT, CREDIT, ISSUES_ELECTIONS_POLITICS or NONE. A companion field, special_ad_category_country, takes two-letter country codes and is required whenever a category is set. According to Meta's developer documentation, every campaign created through the API must carry the field, even if only to say NONE, a rule in force since March 31, 2020. The live reference could not be read for this article; an archived copy was used.

Once HOUSING, EMPLOYMENT or the credit category is selected, the system enforces the restrictions with a hard error rather than a warning. According to the documentation, the age range is fixed at 18 to 65+, with an exception for credit ads in Europe. Gender defaults to all. Lookalike audiences, saved audiences, targeting expansion, bid multipliers, location exclusions and detailed targeting exclusions are unavailable, and some detailed targeting interests are removed altogether. Location targeting must cover at least a 15-mile radius around any chosen point, which rules out ZIP code targeting. Lead forms for financial and employment ads cannot ask for age, gender or location, according to Jon Loomer, a Meta ads consultant. Custom audiences from customer lists remain permitted.

Enforcement combines self-declaration with automated and human review. Undeclared ads that belong in a category can be rejected, according to Meta's documentation.

The political category carries no targeting limits by itself. Instead, an advertiser must complete identity authorisation, reside in the country targeted and run a "paid for by" disclaimer, and the ads are archived in Meta's Ad Library. Since January 9, 2024, an ad-level authorisation value, POLITICAL_WITH_DIGITALLY_CREATED_MEDIA, flags political ads containing digitally created or altered content.

Which categories, and where

Meta's four categories as of October 2026 are housing; employment; financial products and services; and social issues, elections or politics. The financial category replaced the older credit label. Meta announced it in October 2024 and introduced it on January 14, 2025, making it mandatory for campaigns based in, or targeting, the US. It covers everything credit did, including cards, loans and debt recovery, and adds insurance, bank accounts, investment services and payment services, according to Loomer's summary of Meta's announcement. The API value remains CREDIT in the documentation reviewed.

Geography has widened in stages. The requirement started with US advertisers on December 4, 2019. Canada was added on December 3, 2020, and parts of Europe on December 7, 2021, according to the developer documentation.

The rules trace to three US statutes. The Fair Housing Act of 1968 prohibits advertising that indicates a preference based on race, colour, religion, sex, disability, familial status or national origin. Title VII of the Civil Rights Act of 1964 covers job advertising, and the Equal Credit Opportunity Act of 1974 covers lending.

In October 2016, ProPublica reported that Facebook let advertisers exclude users by "ethnic affinity" from housing ads. Lawsuits and employment discrimination charges followed. On August 13, 2018, the Department of Housing and Urban Development (HUD) opened a Secretary-initiated complaint.

On March 19, 2019, Facebook settled four matters, including a suit by the National Fair Housing Alliance and Mobley v. Facebook. According to the American Civil Liberties Union (ACLU), Facebook agreed to build a separate system for housing, employment and credit ads with no targeting by age, gender or ZIP code, a 15-mile minimum radius, and lookalike audiences that could not consider those attributes. "It's crucial that micro-targeting not be used to exclude groups that already face discrimination," said Galen Sherwin, senior staff attorney at the ACLU. The modified lookalike became Special Ad Audiences.

Nine days later, HUD charged Facebook anyway, alleging that its delivery algorithm discriminated even when advertisers chose broad audiences. "Using a computer to limit a person's housing choices can be just as discriminatory as slamming a door in someone's face," said Ben Carson, then HUD secretary.

The charge was later referred to the Department of Justice (DOJ). On June 21, 2022, the department sued and settled simultaneously in the Southern District of New York, in what it called its first case challenging algorithmic bias under the Fair Housing Act. Meta agreed to stop using Special Ad Audiences for housing by December 31, 2022, and to pay $115,054, the maximum civil penalty available. Meta withdrew Special Ad Audiences for all special ad categories between August 25 and October 12, 2022.

The settlement's second requirement produced the Variance Reduction System (VRS), announced on January 9, 2023. VRS estimates the age, sex and race or ethnicity of people reached by a housing ad, compares the result with the eligible audience and adjusts pacing. The DOJ set targets: by December 31, 2023, a variance of 10% or less for sex on 91.7% of housing ads and for race or ethnicity on 81%. Court oversight ran to June 27, 2026.

Equivalents on other platforms

Google announced on June 11, 2020 that housing, employment and credit advertisers in the US and Canada could no longer target by gender, age, parental status, marital status or ZIP code, effective October 19, 2020. A later change extended the demographic limits to consumer finance ads from February 28, 2024. On June 3, 2026, Google clarified how Demand Gen and Discovery campaigns treat these limits.

TikTok uses the same name. Its special ad categories cover housing, employment and credit for audiences in the US and Canada and, according to its help centre, prohibit gender, ZIP code and lookalike targeting while limiting age to 18 and over. Pinterest bars actalike targeting for credit, employment and housing in certain markets.

In the European Union, the closest analogue is legal rather than contractual. Article 26(3) of the Digital Services Act bans platforms from showing ads based on profiling that uses special category data such as ethnicity or health, from February 17, 2024.

Why it matters for marketers

For any advertiser in property, recruitment, banking or insurance, the category determines what targeting is possible on the largest social platforms. Campaigns become broader by design, and performance depends more on creative, customer lists and the platform's own delivery algorithm. When Meta removed detailed targeting exclusions in January 2025 in favour of account-level controls, those new controls did not apply to special ad categories, which Meta said kept them "accessible to a broad audience".

Political advertisers face a different constraint: availability itself. Meta stopped political, electoral and social issue ads in the EU on October 6, 2025, ahead of the Transparency and Targeting of Political Advertising (TTPA) regulation, and Google restricted EU political ads to official communications.

Limitations and disputes

The central criticism is that restricting what advertisers select does not control what the algorithm delivers. In April 2019, researchers led by Muhammad Ali and Piotr Sapiezynski at Northeastern University ran job ads with identical targeting. An ad for lumber industry jobs reached an audience about 90% male and 72% white; one for taxi drivers reached about 75% Black users. In 2021, Basileal Imana, Aleksandra Korolova and John Heidemann found gender skew in Facebook's job ad delivery that job qualifications could not explain, and none on LinkedIn.

VRS has its own critics. A June 2025 evaluation by Imana and colleagues, presented at the ACM FAccT conference, found that VRS brought racial variance below 10% in all 18 paired tests, but cut reach by 9.82% on average and raised the cost per 1,000 people reached by 12.02%. Meta's voluntary extension of VRS to employment and credit ads did not bring variance below 10% in their tests.

Mis-declaration is a third gap. The system relies on advertisers declaring honestly, backed by classifiers whose accuracy Meta does not publish. Outside the covered regions, the rules do not apply at all.

Not the same as

Sensitive categories are Google's brand suitability labels for content such as gambling or tragedy. They govern where ads appear, not whom they target.

Special category data is a GDPR term under Article 9 for data on health, ethnicity, religion, sexual orientation and similar traits.

Sensitive interest categories are Google's 21 topics in its personalised advertising policy, such as health or negative financial status, which limit ad personalisation rather than declaring a regulated sector.

Restricted goods policies, covering alcohol, gambling or pharmaceuticals, require authorisation or age gating on Meta but are not special ad categories.

Recent developments

The court oversight period in the DOJ settlement was scheduled to end on June 27, 2026; no public statement from the DOJ or Meta on its conclusion or extension could be found as of October 2026. And on October 6, 2026, Meta told developers it would block new US political ads from October 27 to November 3, 2026, reminding API users to select ISSUES_ELECTIONS_POLITICS to obtain the required disclaimer.

Timeline

  • 1964: Civil Rights Act, Title VII, prohibits employment discrimination
  • 1968: Fair Housing Act prohibits discriminatory housing advertising
  • 1974: Equal Credit Opportunity Act covers lending
  • October 2016: ProPublica reports Facebook allows "ethnic affinity" exclusions on housing ads
  • August 13, 2018: HUD files a Secretary-initiated complaint against Facebook
  • March 19, 2019: Facebook settles with the ACLU, NFHA, CWA and others, agreeing separate housing, employment and credit targeting rules
  • March 28, 2019: HUD charges Facebook with violating the Fair Housing Act
  • April 3, 2019: Ali et al. publish "Discrimination through optimization"
  • December 4, 2019: Meta requires US advertisers to declare housing, employment and credit campaigns
  • March 31, 2020: API campaigns must specify a category or NONE
  • June 11, 2020: Google announces housing, employment and credit targeting limits for the US and Canada
  • October 19, 2020: Google's limits take effect
  • December 3, 2020: Meta extends requirements to Canada
  • April 2021: Imana, Korolova and Heidemann publish their Facebook and LinkedIn job ad audit
  • December 7, 2021: Meta extends requirements to parts of Europe
  • June 21, 2022: DOJ sues and settles with Meta under the Fair Housing Act
  • August 25 to October 12, 2022: Meta withdraws Special Ad Audiences
  • January 9, 2023: DOJ and Meta announce VRS compliance targets
  • January 9, 2024: Meta adds a digitally created media authorisation value for political ads
  • February 17, 2024: DSA Article 26(3) applies to all online platforms
  • February 28, 2024: Google extends demographic limits to consumer finance ads
  • October 2024: Meta announces the financial products and services category
  • January 14, 2025: Financial products and services category introduced for US campaigns
  • June 2025: Imana and colleagues publish an external evaluation of VRS
  • October 6, 2025: Meta stops political, electoral and social issue ads in the EU
  • June 3, 2026: Google clarifies Demand Gen treatment of housing, employment and credit limits
  • June 27, 2026: Scheduled end of court oversight of the DOJ settlement
  • October 6, 2026: Meta announces a US political ads blackout from October 27 to November 3

Summary

Who. Meta defines and enforces special ad categories across Facebook, Instagram and Messenger. Advertisers in housing, employment, finance and politics must declare them. The ACLU, the National Fair Housing Alliance, HUD and the Department of Justice drove their creation, and Google, TikTok and Pinterest apply similar rules.

What. A mandatory campaign label that, for housing, employment and financial ads, removes age, gender, postcode and lookalike targeting, and for political ads requires authorisation and disclosure.

When. Agreed in the March 19, 2019 settlement, required in the US from December 4, 2019, extended to Canada in 2020 and parts of Europe in 2021, and broadened to financial products and services in January 2025.

Where. In Meta Ads Manager at campaign creation and in the Marketing API's special_ad_categories field, applying to US, Canadian and some European audiences.

Why. US civil rights law bars discriminatory advertising for homes, jobs and credit, and Facebook's targeting tools were shown to enable it. Research since 2019 shows that delivery algorithms can still skew audiences after targeting is restricted, which led to the Variance Reduction System and continuing scrutiny.