Brand Safety — Alpha Quantum | MFA Detection, GARM Categories & COPPA Compliance

The Challenge

Ad spend bleeds into unsafe, fraudulent and non-compliant inventory

Programmatic advertising delivers reach, but without granular brand safety controls, impressions land on Made-for-Advertising sites, child-directed content and domains that damage brand reputation. Most blocklist solutions trade safety for reach. The real problem is that binary blocking is too blunt and static lists decay within days.

MFA sites drain budgets silently

Made-for-Advertising domains mimic legitimate publishers with ad-heavy layouts, recycled content and aggressive refresh patterns. They pass basic viewability checks while delivering zero engagement. The Association of National Advertisers estimates that 15% of programmatic spend flows to MFA inventory. Without a dedicated MFA scoring engine that analyzes ad-to-content ratios, clickbait patterns and ad-stack composition, these sites remain invisible to standard brand safety tools.

Binary blocklists destroy reach

Traditional keyword-based blocklists treat entire content categories as unsafe. A news article reporting on crime gets blocked alongside content that glorifies it. The GARM Brand Safety Floor and Suitability Framework introduced tiered risk levels for exactly this reason. Whole-page classification with sentiment analysis distinguishes reporting from endorsement, so advertisers can buy journalism at moderate and low risk tiers instead of blanket-blocking categories that represent 30-40% of available inventory.

COPPA violations carry regulatory penalties

Advertising on child-directed content without proper safeguards violates the Children's Online Privacy Protection Act and equivalent regulations globally. Fines reach $50,000 per violation. Identifying child-directed content requires more than checking if a site self-declares as a kids' property. The COPPA risk score (0-100) analyzes content signals, visual elements, language complexity and topic patterns to surface child-directed pages that do not carry explicit labels.

Static lists miss new unsafe domains

Thousands of new domains launch daily, many purpose-built for ad arbitrage or hosting unsafe content. A blocklist compiled last month misses every new entrant. The Website Categorization API classifies new domains in real time and the 120M-domain offline database refreshes every seven days. Continuous classification ensures that brand safety coverage extends to the long tail of the web, not just the domains that were known when the list was compiled.

One API Call

8+ brand safety signals from a single URL

Every request to the Website Categorization API returns a structured JSON response with safety, quality, compliance and content signals. These fields power pre-bid filtering, post-bid verification and publisher vetting workflows across programmatic advertising supply chains.

FieldTypeExampleDescriptionUse

MFA Score0-10087 (high MFA risk)Made-for-Advertising likelihood based on ad density, content quality, refresh behavior and monetization patternsPre-bid filtering

COPPA Risk Score0-10092 (child-directed)Child-directed content probability from topic analysis, language complexity, visual signals and audience indicatorsRegulatory compliance

IAB v3 CategoriesarrayNews > Politics > Elections703 categories across 4 tiers for granular content classification and GARM suitability mappingSuitability tiers

Web Filtering CategorystringGambling, Adult, Malware58 security-focused categories designed for block and allow policy decisionsHard blocks

Quality Score0-100Quality: 72Composite metric measuring trustworthiness, visual design quality, content depth and editorial standardsPublisher vetting

Sentimentper-entityNegative (brand), Neutral (topic)Entity-level sentiment analysis distinguishing tone of coverage for specific brands, people and topicsTone-sensitive ads

Clickbait Score0-10064 (moderate clickbait)Headline manipulation, sensationalism and engagement-bait pattern detection across page contentContent quality

Ad-Stack ForensicsarrayPrebid.js, Amazon TAM, Google AdXHeader bidding wrappers, ad exchanges, verification tags and monetization infrastructure detected on the pageSupply chain audit

Full brand safety field reference on websitecategorizationapi.com/brand-safety.php. All fields available via real-time API and in the 120M domain offline database.

nytimes.comMFA: 2 · COPPA: 5 · Quality: 94 · Safe

clickfarm.xyzMFA: 96 · Quality: 8 · Clickbait: 91 · Block

kidsgames.netCOPPA: 94 · Child-Directed · Restrict

reuters.comMFA: 1 · Quality: 96 · Sentiment: Neutral · Safe

GARM Suitability Framework

Tiered risk scoring replaces binary blocking

The Global Alliance for Responsible Media defined suitability tiers so advertisers can calibrate risk tolerance instead of blanket-blocking entire content categories. The Website Categorization API maps every classified URL to GARM-aligned risk tiers using full-page classification, entity-level sentiment and content quality scoring.

Risk Tiers

Content Categories

Sentiment Analysis

Floor Risk (Hard Block)

Content universally considered unsafe for all advertisers regardless of brand or campaign. Includes illegal content, terrorism, child exploitation, malware distribution and extreme violence. The web filtering taxonomy flags these categories with zero tolerance. No advertiser should appear adjacent to floor-risk content under any circumstances.

High Risk

Content that most brands will want to avoid but some categories can tolerate with context. Includes graphic violence in news reporting, explicit drug references, hate speech adjacent content and highly polarizing political commentary. Full-page classification distinguishes editorial coverage from endorsement so that news publishers are not uniformly penalized.

Medium and Low Risk

Content with contextual sensitivity that requires brand-specific evaluation. Medium risk includes crime reporting, health crisis coverage, political debate and gambling-adjacent entertainment. Low risk covers general news, opinion content and lifestyle topics with minor sensitivity. Sentiment analysis separates factual reporting from sensationalism, and quality scoring filters low-effort content aggregation from original journalism.

44 Safety Categories

The web filtering taxonomy includes 44 categories specifically designed for brand safety policy decisions. Categories span adult content, gambling, weapons, tobacco, alcohol, controlled substances, hate speech, piracy, malware and violence. Each category maps to a GARM suitability tier, enabling automated policy enforcement without manual URL review across millions of impressions daily.

IAB v3 Mapping

703 IAB Content Taxonomy v3 categories classified across 4 hierarchical tiers provide granular content context. A page classified as News and Politics at the top tier breaks down into subcategories like Elections, International Relations or Domestic Policy. Advertisers set suitability rules at the tier level that matches their risk appetite, from broad category exclusions to precise subcategory targeting.

Dual-Taxonomy Scoring

Every URL receives both an IAB content classification and a web filtering category simultaneously. The IAB taxonomy describes what the content is about for advertising context. The web filtering taxonomy describes whether the content should be blocked for safety. Running both taxonomies in parallel means advertisers get context for suitability decisions and security teams get actionable block-or-allow signals from the same API call.

Entity-Level Sentiment

Sentiment analysis operates at the named entity level, not just the page level. A news article about a brand crisis returns negative sentiment for that specific brand while the overall page sentiment may be neutral. This granularity lets brand safety systems apply rules based on how a specific brand, person or organization is discussed rather than the general topic of the page.

Tone-Sensitive Placement

Luxury brands avoid negative-tone content even when the topic is safe. Pharmaceutical companies avoid fear-based content. Insurance advertisers may actually perform better next to risk-related coverage. Entity-level sentiment enables these nuanced placement strategies without sacrificing reach. The API returns sentiment per detected entity so campaign managers can write rules matching their brand positioning and voice.

Reporting vs. Glorification

The combination of full-page classification and entity sentiment distinguishes factual reporting on sensitive topics from content that promotes or glorifies harmful behavior. A news article about drug policy returns different signals than a page promoting drug use. This distinction is what allows advertisers to keep buying quality journalism at moderate risk tiers instead of keyword-blocking entire verticals that contain valuable reach.

Full brand safety documentation

MFA Detection Engine

Six signals that identify Made-for-Advertising sites

MFA sites are engineered to extract ad revenue through high ad density, recycled content and aggressive user engagement tactics. The MFA detection engine scores each URL across six dimensions to produce a composite risk score from 0 to 100. Legitimate publishers score below 15. MFA sites consistently score above 70.

Ad-to-Content Ratio Analysis

Measures the proportion of visible page area occupied by advertising units versus editorial content. MFA sites typically show ad-to-content ratios exceeding 60%, with multiple sticky, interstitial and in-content ad placements competing for attention. The ratio calculation accounts for lazy-loaded ads that render as the user scrolls, capturing the full monetization footprint of the page.

MFA scoring methodology

Clickbait Pattern Detection

Analyzes headline structures, thumbnail imagery and content framing for engagement-bait patterns. Common signals include curiosity-gap headlines, emotional manipulation, listicle structures with paginated layouts designed to inflate page views and sensationalized language patterns. The clickbait score (0-100) operates independently of the MFA score, providing a separate content quality dimension for advertisers.

Clickbait detection

Ad-Stack Forensics

Identifies the advertising technology infrastructure running on each page. Detects header bidding wrappers (Prebid.js, Amazon TAM, Index Exchange), ad exchanges (Google AdX, OpenX, Magnite), verification tags (IAS, DoubleVerify, MOAT) and ad refresh mechanisms. MFA sites often run multiple header bidding configurations simultaneously, maximizing competition for each impression slot.

Ad-stack analysis

Refresh and Redirect Monitoring

Detects automatic page refresh timers, forced redirects, infinite scroll pagination and other mechanisms designed to inflate impression counts without genuine user engagement. Some MFA sites refresh ad slots every 15-30 seconds, generating multiple billable impressions from a single user session. The detection engine identifies JavaScript-based refresh timers, meta refresh tags and redirect chains.

Content Quality Assessment

Evaluates editorial quality through multiple signals: original versus syndicated content detection, reading level analysis, content depth measurement, source attribution and publication frequency patterns. MFA sites typically feature thin, recycled or AI-generated content with minimal editorial oversight. The quality score (0-100) separates legitimate publishers investing in original journalism from content farms.

Revenue Attribution Analysis

Maps the relationship between traffic acquisition cost and advertising revenue potential for each domain. MFA operations typically spend heavily on paid social, content recommendation widgets and search arbitrage to acquire cheap traffic, then monetize it through dense ad placements at margins that only work with low-quality inventory. Detecting this revenue model pattern strengthens MFA classification accuracy.

Use Cases

How brand safety data protects advertising investment

From programmatic DSP integration to compliance reporting, the same classification engine supports different workflows depending on whether the team is buying media, vetting supply or demonstrating regulatory compliance.

01

Pre-Bid Brand Safety Filtering

Classify bid request URLs in real time before the auction. The API responds in under 100ms, enabling pre-bid decisioning at programmatic scale. Block URLs exceeding your MFA threshold, restrict child-directed content and apply GARM suitability tiers to every impression opportunity. Feed the classification into your DSP as custom contextual segments for inclusion and exclusion targeting.

02

Post-Bid Verification

Audit delivered impressions against brand safety policy after the campaign runs. Compare placement URLs against MFA scores, content categories and quality metrics to measure policy adherence. Identify supply sources consistently delivering unsafe inventory so the trading team can optimize future buys. Post-bid analysis feeds the feedback loop that improves pre-bid rules over time.

03

Publisher Supply Quality

Evaluate publisher inventory before committing media spend. Score every domain in a proposed supply list for MFA risk, content quality, ad-stack composition and audience demographics. Build inclusion lists of verified quality publishers and exclusion lists of domains that fail safety thresholds. The 120M-domain offline database enables bulk assessment of entire publisher portfolios in minutes.

04

COPPA Compliance Screening

Identify child-directed content across your media plan before serving ads that require age-gating. The COPPA risk score (0-100) surfaces pages targeting children even when the publisher does not self-declare as a kids' property. Integrate the score into your ad server decisioning to automatically suppress non-compliant creatives on high-COPPA-risk inventory. Document compliance for regulatory audits.

05

Contextual Targeting with Safety

Replace third-party cookie targeting with contextual signals that include built-in safety scoring. The API returns 1,667+ interest personas, IAB categories and audience demographics alongside brand safety fields. Build contextual segments that are both relevant and safe. A campaign targeting technology enthusiasts can exclude technology content that also carries high controversy or negative sentiment signals.

06

Ad Fraud Prevention

MFA detection overlaps significantly with ad fraud identification. Sites designed to extract ad revenue through manipulated engagement, inflated impressions and arbitraged traffic are both brand-unsafe and fraudulent. The MFA score, ad-stack forensics, refresh detection and quality metrics together provide a comprehensive signal set for fraud prevention teams working alongside brand safety analysts.

Integration Pipeline

From bid request to safety decision in under 100ms

Real-time API for pre-bid decisioning. Offline database for bulk supply audits. Both return the same brand safety field schema with MFA, COPPA, quality and content signals.

01

Input

URL from bid stream, ad server log or publisher supply list

02

Classify

MFA score, COPPA risk, quality metrics, sentiment and content categories

03

Score

GARM risk tier assignment, suitability mapping, policy rule evaluation

04

Act

Block, allow or flag the impression based on campaign safety thresholds

Platforms

Three platforms powering brand safety infrastructure

The Website Categorization API handles real-time brand safety scoring. The URL Categorization Database delivers the full 120M-domain corpus for offline supply audits. The AI Tools Blocklist governs AI-generated content risks across 20,361+ domains.

Website Categorization API URL Categorization Database AI Tools Blocklist

Scale

Enterprise-grade brand safety infrastructure

19 years of continuous operation. Real data, real customers, real scale.

120M+

Domains Classified

700+