Brand safety and brand suitability

In programmatic advertising, an ad can appear on millions of pages its buyer has never seen. Sometimes those pages are ugly: terrorism, hate speech, piracy, or a news report about a plane crash next to an airline's holiday offer. Brand safety is about avoiding content that is harmful for any advertiser. Brand suitability is about choosing content that fits a particular brand's values and tone.

Brand safety is like a restaurant refusing to serve food that is spoiled; nobody should eat it. Brand suitability is like a diner's personal preferences: one person avoids spicy food, another avoids meat. Both matter, but only one is a universal rule.

The floor and the tiers

The World Federation of Advertisers' initiative GARM (Global Alliance for Responsible Media), launched in 2019, gave the industry a shared vocabulary: a brand safety floor of content no advertiser should fund, such as child sexual abuse material or terrorist propaganda, and a suitability framework with low, medium and high risk tiers across a dozen or so sensitive categories. GARM was discontinued by the WFA in August 2024, but its definitions remain widely used in verification tools and platform policies.

How brand safety tools work

1. Classify the contentVerification vendors crawl pages and apps, read text, and increasingly analyse images, audio and video with machine learning to assign content categories and risk levels.
2. Decide before the bidPre-bid segments inside DSPs let buyers avoid inventory classified as unsafe or unsuitable.
3. Check when the ad servesA post-bid tag can re-check the page when the ad loads and block it if the content fails the rules, reported as a block rate.
4. Report and refineReports show where ads ran, where they were blocked and why, so teams can adjust settings.

When blocking backfires

Early brand safety relied heavily on keyword blocklists. If a page contained "shooting", "attack" or "virus", ads were blocked. That also blocked film reviews, sports reports and health information. During the early months of the COVID-19 pandemic, news publishers around the world saw huge audiences but reduced ad revenue because so many pages were blocked by terms like "coronavirus". Over-blocking quietly defunds quality journalism, including local and regional news in markets such as India, Brazil and across Europe.

Context, not just keywords

Modern tools use contextual analysis to understand what a page is actually about. An article on climate policy and an article praising violence may share words but not meaning. Better classification lets brands move from "block everything risky" to "fund what fits", which is the whole point of suitability.

Language is a major challenge. Many classifiers were trained mostly on English, yet huge audiences read in Hindi, Tamil, Bahasa Indonesia, Arabic, Portuguese and dozens of other languages. Weak coverage leads to two opposite failures: unsafe content slipping through because it is not understood, and safe content being blocked because the tool is unsure. Buyers running campaigns across India, Southeast Asia or the Middle East should ask vendors which languages they classify natively and how accurately. Video and podcasts add another layer, since the meaning sits in speech and images rather than page text.

Where safety meets fraud and security

• Fake and MFA sites are often brand-safe on paper but waste money. Safety tools do not replace invalid traffic detection.
• Malvertising, malicious ads that attack readers, is the reverse problem: it is the publisher's and the reader's safety at risk, not the advertiser's.
• Walled gardens such as social platforms offer their own brand safety controls and increasingly allow third-party ad verification on their feeds.