Sensitive categories are named labels that Google's advertising systems attach to web pages, apps and ads whose subject matter some brands want to avoid: gambling, alcohol, politics, religion, tragedy, profanity and similar themes. On the buy side, an advertiser in Display & Video 360 (DV360) could tick a box to keep its ads off pages classified under any of them. On the sell side, a publisher in Google Ad Manager or AdSense could tick a box to keep ads about dating or weight loss off its own pages. The feature exists because lawful content is not suitable for every brand, and because a checklist of named topics is easier to buy against than a probability score. As of October 1, 2026, DV360 no longer offers most of these exclusions for programmatic inventory, which makes the term partly historical on the buying side.

How the buy-side controls worked

In DV360, sensitive categories sat under brand safety targeting at advertiser or line item level, for display, video and audio line items. According to the Display & Video 360 Help Center, the list ran to 17 categories: sexual, derogatory, downloads and sharing, weapons, gambling, violence, suggestive, profanity, alcohol, drugs, tobacco, politics, religion, tragedy, transportation accidents, shocking, and sensitive social issues, defined as topics that "spark strong, opposing views", such as abortion. Weapons was later merged into sensitive social issues. In the DV360 API, selecting either one now selects the other.

Classification is page-level. Google crawls pages and reclassifies them often, prioritising those that change frequently or draw heavy traffic. When a bid request arrives, the DSP checks the URL against its classification and declines to bid if the page carries an excluded label. This is pre-bid blocking: the impression is never bought, rather than bought and reported afterwards. Google's documentation lists three ways it fails: inaccurate URLs passed in the request, unsupported languages, and pages that change between the bid and the render.

For developers, the settings travelled through the targeting type TARGETING_TYPE_SENSITIVE_CATEGORY_EXCLUSION in the DV360 API and through the "Brand Safety Sensitivity Setting" and "Brand Safety Custom Settings" columns of Structured Data Files (SDF), the spreadsheet format agencies use for bulk edits. Sensitive categories sat alongside digital content labels (DCLs), an age-style rating scale from DL-G (general audiences) to DL-MA (mature audiences only). Labels set maturity; categories named topics.

Campaign Manager 360 used the same taxonomy for reporting. According to its Help Center, a single impression, including a blocked one, can count under several categories at once, and impressions that cannot be classified appear as "(not set)".

The sell-side version

Publishers face the mirror problem. A news site may not want ads for get-rich-quick schemes beside its coverage, regardless of which advertiser bids most. Google Ad Manager's protections let a publisher block 15 standard sensitive categories, according to its Help Center: astrology and esoteric, consumer loans (Japanese and Korean only), cosmetic procedures and body modification, dating, downloadable utilities, drug supplements, get rich quick, politics, references to sex, religion, sensationalism, sexual and reproductive health, significant skin exposure (now deprecated), social casino games, and weight loss.

These blocks classify the ad, not the page, and apply to open auctions, private auctions and First Look. Google classifies creatives automatically rather than relying on advertiser declarations, and describes the result as a "best attempt" with no guarantee. Outside Google's stack, the equivalent travels in OpenRTB as the bcat field, a list of blocked advertiser categories coded against the IAB Tech Lab Content Taxonomy.

Origin and evolution

The sell-side feature came first. On April 22, 2009, AdSense product manager Atul Bhandari announced a limited beta of "category filtering" in the Ad Review Center, naming "dating, religion, and politics" as examples. By August 23, 2010, sensitive category blocking had launched in 11 languages with up to 11 blockable categories, according to the Inside AdSense blog.

On the buy side, Google launched DoubleClick Verification in 2012 with predefined categories including adult, violence, alcohol and politics, and in December 2014 brought digital content labels into DoubleClick Bid Manager, the product renamed Display & Video 360 in 2018. By January 2019, PPC Land's overview of how Google brand safety works counted 16 DV360 and Campaign Manager classifiers, nine separate sensitive topic classifiers in Google Ads and six digital content labels: three overlapping vocabularies across one company's products.

Industry frameworks grew alongside. The 4A's published a 13-category framework in 2018. The World Federation of Advertisers' Global Alliance for Responsible Media (GARM) published its Brand Safety Floor and Suitability Framework in September 2020, with 11 categories and three risk tiers, and IAB Tech Lab's Content Taxonomy 2.2 placed those categories under a Sensitive Topics branch the following month.

Google then began consolidating. In September 2024, DV360 phased out digital content labels for YouTube campaigns, steering buyers to inventory modes. The Q2 2026 roadmap scheduled the deprecation of sensitive category exclusions for display, video, connected TV (CTV) and audio for August 2026, warning that failure to move to the new controls "will result in unrestricted serving". On August 27, 2026, the DV360 API team announced the removal of both controls for October 1, as unversioned changes affecting every API version at once.

Why it matters for marketers

Sensitive categories were the most legible suitability control in programmatic buying. A brand with a contractual ban on gambling adjacency, a pharmaceutical advertiser avoiding tragedy, or a public body avoiding politics could translate its policy directly into a setting and show an auditor the checkbox. The replacement system does not offer that one-to-one translation. DV360 now provides three inventory modes, Maximum, Moderate and Limited, and ten content themes, among them politics, religion, health (sensitive) and news (sensitive). No theme covers gambling, alcohol, tobacco, weapons or drugs, as PPC Land's analysis of the switch documented today. Google has published no mapping and no migration tool.

The change also shifts work to verification vendors. Named-category exclusions remain available through DoubleVerify, Integral Ad Science (IAS) and others on display, video and CTV. Audio buyers have no third-party brand safety integration in DV360, according to the same report, leaving them with modes, themes, keywords and exclusion lists. Microsoft Advertising, by contrast, still exposes category-style content exclusions powered by IAS, eight of them blocked by default.

Limitations and disputes

The main criticism of category blocking is overreach. Labels such as tragedy, politics and sensitive social issues catch most serious journalism. A January 2020 study by CHEQ and the University of Baltimore estimated that incorrect keyword blocking cost US news publishers about $2.8 billion in 2019, roughly a quarter of their programmatic revenue, and found only 9 of the year's 100 most-read news articles passed standard blocklists. Those figures come from a vendor selling an alternative. Vodafone reported 10% more news inventory and a 41% cut in block rates after narrowing its settings, again company-reported.

Accuracy is a second problem. Categories are applied by Google's own classifiers and audited by no one outside Google for programmatic inventory. YouTube's inventory modes received Media Rating Council (MRC) brand safety accreditation on June 3, 2026, with a reported error rate below 1%, but the scope excludes CTV devices and non-YouTube inventory. The MRC separately restricted the term "brand safety" in October 2025 to services that analyse images, video and audio, not only text.

A third dispute concerns what exactly survived. Google's Q3 2026 help page says sensitive categories are deprecated for programmatic inventory, while the August developer post said only "the majority" were removed. The DV360 API reference, as of October 2026, marks 17 of 20 enum values deprecated; the two live categories, embedded video and live-streaming video, describe formats rather than topics. The "content not yet rated" label was exempted, though Google has not said whether that carve-out applies to categories or only to labels.

Finally, the taxonomy embedded Google's editorial judgements: derogatory content included academic and journalistic discussion of bias.

Not the same as

Sensitive interest categories are a separate Google Ads policy concept. They govern which ads may use personalised targeting, not where ads appear. Google lists 21, including health, sexual orientation, religious beliefs and negative financial status, and in June 2026 clarified how Demand Gen campaigns treat them.

Special category data is a legal term from Article 9 of the General Data Protection Regulation (GDPR), covering data on health, ethnicity, religion, political opinions, sexual orientation, biometrics and union membership. The Court of Justice of the European Union narrowed the "made public" exception for such data in Schrems v Meta on October 4, 2024. It restricts processing of people's data, not the content beside an ad.

Content themes are DV360's replacement controls: ten broader groupings applied with inventory modes, built on YouTube's model.

Brand safety floor categories, such as terrorism or child exploitation, are content no advertiser should fund. Sensitive categories covered lawful content some advertisers merely preferred to avoid, which is the domain of brand suitability.

Recent developments

The October 1, 2026 removal is the defining recent event. One advertiser-level setting now governs YouTube, display, CTV and audio together, and a separate deal exemption lets buyers bypass all suitability settings for chosen deals on non-reservation line items. PPC Land's brand safety explainer, published on October 6, places the removal within a wider dispute over who sets suitability standards, after the WFA discontinued GARM on August 9, 2024 and agreed in a July 29, 2026 settlement with X not to revive it. Google's pre-bid filtering controls now rely on tiers and themes, while named-category exclusions remain available mainly through third-party verification.

Timeline

  • April 22, 2009: AdSense announces a limited beta of category filtering in the Ad Review Center
  • August 23, 2010: AdSense sensitive category blocking available in 11 languages with up to 11 categories
  • 2012: Google launches DoubleClick Verification with predefined content categories
  • December 2, 2014: Google brings digital content labels into DoubleClick Bid Manager
  • 2018: The 4A's publishes a 13-category brand suitability framework; DoubleClick Bid Manager becomes Display & Video 360
  • January 1, 2019: 16 DV360 and Campaign Manager sensitive classifiers in use alongside six digital content labels
  • January 20, 2020: CHEQ and University of Baltimore estimate $2.8 billion in 2019 US news revenue lost to keyword blocking
  • September 2020: GARM publishes its Brand Safety Floor and Suitability Framework with 11 categories
  • October 2020: IAB Tech Lab Content Taxonomy 2.2 adds a Sensitive Topics branch
  • August 9, 2024: WFA discontinues GARM after X's antitrust lawsuit
  • September 23, 2024: Google announces digital content labels will stop applying to YouTube campaigns from October 2024
  • October 18, 2025: MRC limits "brand safety" to content-level verification services
  • May 22, 2026: DV360 Q2 roadmap schedules sensitive category exclusions for deprecation in August 2026
  • June 3, 2026: YouTube inventory modes receive MRC brand safety accreditation, including Shorts
  • July 29, 2026: WFA settles with X and agrees not to restart GARM
  • August 27, 2026: DV360 API team announces removal of digital content label and most sensitive category exclusions
  • October 1, 2026: Removal takes effect for programmatic inventory in DV360

Summary

Who. Google defined and applied sensitive categories across DV360, Campaign Manager 360, Google Ad Manager and AdSense. Advertisers and agencies used them to exclude content; publishers used them to block ads. Verification vendors such as DoubleVerify and IAS offer comparable named-category controls.

What. Named labels, such as gambling, alcohol, politics and tragedy, assigned by Google's classifiers to pages or ads, used to exclude them before bidding or serving. They sat beside digital content labels, which rated maturity rather than topic.

When. Publisher-side category filtering entered beta in AdSense on April 22, 2009. Buy-side categories developed through DoubleClick Verification from 2012 and DoubleClick Bid Manager from 2014. DV360 removed most of them for programmatic inventory on October 1, 2026.

Where. In DV360 brand safety targeting at advertiser and line item level, in the DV360 API and SDF, in Campaign Manager 360 reporting, and in the protections settings of Google Ad Manager, AdSense and AdMob.

Why. They let brands translate a suitability policy into named, auditable exclusions and let publishers keep unwanted ad topics off their sites. Their removal from DV360 replaces that precision with broader tiers and themes, a shift that critics say leaves some contractual exclusions with no native equivalent.