Google’s brand-safety controls just got simpler to describe and harder to audit. On October 1, 2026, most of DV360’s legacy sensitive-category exclusions switched off. In their place: ten content themes that advertisers tick to keep campaigns away from specific subjects.
If you run programmatic YouTube, display, CTV or audio, this is now your main topic-level lever.
What changed on October 1
For years, buyers could exclude topics such as gambling, alcohol, drugs, derogatory content, transportation accidents, and downloads and sharing. Google says those older signals have been consolidated into ten themes. The Q3 2026 roadmap states that inventory modes plus content themes now cover all YouTube and programmatic DV360 inventory.
The ten themes are:
- Content Suitable for Families
- Games (fighting)
- Games (mature)
- Health (sensitive)
- Health (source undetermined)
- News (recent)
- News (sensitive)
- News (source not featured)
- Politics
- Religion
Three of the ten concern news alone, which quickly becomes a reach problem for advertisers and a demand problem for publishers.
How themes stack with inventory modes
Google frames this as brand suitability rather than brand safety. Inventory mode sets how much profanity, violence or suggestive content a campaign tolerates across Maximum, Moderate or Limited tiers. A content theme is different: it removes an entire subject regardless of tier. The two controls are designed to be stacked.
In Google Ads, the settings sit in the Content Suitability Center under “Advanced settings” as “Excluded content themes” and apply to YouTube, the Google Display Network and Google video partners. In DV360, the same control lives under brand safety in targeting rules and applies at advertiser level. The API exposes a narrower set of nine enum values, where “sensitive” news appears as “unpleasant” news.
Why the shift is contentious
The menu shrank dramatically. Older sensitive categories had named direct counterparts; themes do not. An advertiser that used to block gambling with one switch now has to rely on an inventory tier it cannot inspect by category. Google does not publish theme-level error rates or how much inventory each theme removes.
Classification accuracy is the louder debate. A January 2026 arXiv study of 3,219 news articles found IAS rated 53.72% unsafe, DoubleVerify 35.27% and Oracle 22.22%. Agreement between IAS and DoubleVerify was 0.19 by Krippendorff’s alpha, far below the 0.67 the researchers treat as minimal. Earlier Adalytics research found about 30.3% of sampled New York Times and 43% of sampled Wall Street Journal articles were labelled unsafe by verification vendors. An over-broad theme exclusion quietly strips demand from lawful reporting.
YouTube did earn MRC brand safety accreditation in June 2026 with an advertiser safety error rate under 1%, but that covers in-stream and Shorts bought through Google Ads and DV360. It does not restore category-level control on third-party programmatic supply.
What to do now
Audit your current exclusions before they silently pass into unrestricted serving. DV360 warned that workflows not moved to the new controls would result in unrestricted serving, and the deadline slipped from August to October. Set a deliberate inventory mode as your baseline, apply only themes you can defend, and layer third-party verification from IAS, DoubleVerify or Zefr for an independent check. Watch publisher KPIs if you run news-adjacent campaigns, because exclusions give no per-impression score.
Brands have always needed to avoid unsuitable placements. The risk now is false precision: advertisers think they are excluding a subject, when they are really handing the topic decision to a classifier no one can audit per impression.
Source: PPC Land
