Technical Specification
Deterministic Forensic Signal Detection
A local-only structural auditor that measures the distance between marketing language and verifiable substance. No AI. No tracking. No data leaves your browser. Pure heuristic pattern analysis executed in real time on the active page DOM.
Unit A
30 / 100 pts
Information Density
Measures the ratio between grounded evidence and linguistic inflation. Each sentence is classified as either substantive or hollow based on the presence of quantifiable data, procedural terminology, and named entities versus marketing adjectives and power-word saturation.
METHOD: Sentence-level classification. Grounding markers (numeric values, currencies, dates, technical units) and procedural language override fluff signals. Cynical override applied: if marketing language co-occurs with hard evidence, the fluff flag is suppressed. Final score derived from the inverse grounding ratio across the full content sample.
Unit B
20 / 100 pts
Semantic Alignment
Detects drift between what the page promises and what it delivers. Extracts canonical entities from the primary heading and tracks their persistence throughout the body content. Pages that announce one thing and discuss another trigger a coherence penalty.
METHOD: Entity extraction from H1 element. Top 3 canonical terms are isolated and their frequency is measured across the main content body. Zero persistence triggers maximum drift penalty. Partial persistence applies proportional reduction. Heading-only pages with no verb structure are flagged as pseudo-substance.
Unit C
20 / 100 pts
Trust & Proof Verification
Distinguishes between genuine proof infrastructure and trust theatre — pages that display trust language without linking to any verifiable external source. Detects unsubstantiated social proof claims and orphaned credibility markers.
METHOD: Dual-density analysis. Trust-signal keywords (review, testimonial, rating, verified) and marketing-trust terms are counted and normalized against total word count. Cross-referenced against outbound link domains for verified proof paths (Google, Trustpilot, Clutch, G2, Yelp). High trust language with zero verification links triggers Trust Theatre flag. Image-to-link ratio detects unlinked logo galleries.
Unit D
15 / 100 pts
Commodity Fingerprint
Identifies content that exhibits statistically uniform structure — a hallmark of template-generated or mass-produced copy. Authentic content naturally varies in rhythm and sentence composition. Commodity content does not.
METHOD: Coefficient of variation computed across sentence lengths. Low structural variance (<20% CV) indicates templated or formulaic generation. Progressive thresholds applied: extreme uniformity receives maximum penalty, natural variation receives none. Operates independently of language — structural entropy is language-agnostic.
Unit E
15 / 100 pts
Identity & Authority
Evaluates whether the page declares a verifiable organizational or personal identity through structured data. Anonymous entities with no linked social proof receive the highest penalty. Established identity with cross-platform anchors stabilizes the score.
METHOD: Structured data inspection (JSON-LD). Checks for Organization or Person schema types. Evaluates presence of sameAs cross-references (LinkedIn, social profiles). Missing schema or absent identity declarations trigger Anonymous Entity signal.
Contextual Neutralization Layer
Substance Overrides Jargon
When marketing language appears in proximity to hard evidence (currencies, percentages, dates, technical units), the fluff penalty is suppressed. Context determines whether a word is inflation or legitimate description.
Entity Exception
Proper nouns and named entities detected within commodity-pattern sentences neutralize the template penalty. Unique identity markers override generic-content classification.
Fully Local Execution
All analysis runs inside your browser. No external API calls, no server communication, no cloud processing. The page never leaves your machine.
Deterministic Heuristics Only
No AI models, no machine learning, no neural networks. Every score is produced by fixed, transparent rules applied to the visible DOM. Same page, same score, every time.
Zero Data Collection
No browsing history, no cookies, no analytics, no telemetry. Nothing is stored, cached, or transmitted. Stateless execution — results exist only while the popup is open.
No Background Activity
No persistent processes, no scheduled tasks, no passive monitoring. The engine activates only when you open the popup. Silent until called.
Beyond Signal Detection
Full Semantic Forensic Analysis — Free
This extension detects surface-level structural signals on a single page. For deep multi-page semantic analysis powered by AI — including industry benchmarking, cross-page consistency auditing, and normalized scoring — use the full forensic engine. Free to run. No account required.
Run FREE Full BS Analysis →
Scope
Single page
Multi-page crawl
Method
Heuristic
AI Semantic
Benchmarks
—
Industry avg
Normalization
Raw signal
Weighted score
Output
3 signals
Full report