We can't find the internet
Attempting to reconnect
Something went wrong!
Attempting to reconnect
How Bouncer Works
Live prompt packs, scoring methodology, and version history. What you see here is exactly what's running in production right now.
Scoring Model
Two Independent Axes
Every video is scored on two independent axes. The combination determines the overall assessment:
Technique Intensity
How much persuasion is used — rhetorical skill, emotional appeals, narrative structure. A passionate, well-argued video scores high here. This is not inherently bad.
Transparency
Does the content's apparent purpose match its actual purpose? Is the persuasion overt or covert? High transparency + high intensity = healthy rhetoric.
High intensity + High transparency
Healthy rhetoric
High intensity + Low transparency
Actual manipulation
Low intensity + High transparency
Straightforward content
Low intensity + Low transparency
Subtle propaganda
Six Dimensions
Each video is scored across six influence dimensions. Every finding requires specific evidence and a mechanism explanation — not just a label.
Call to Action
Every video wants something from you, whether it says so or not. Sometimes the ask is small (subscribe, share), sometimes it's big (change your mind, distrust an institution, buy something). The ask itself isn't the problem — the question is whether you noticed it and decided freely to say yes.
“Did I notice what this video wanted from me, and did I decide freely to say yes?”
Emotional Appeal
Every video makes you feel something — that's normal. Emotional appeal is when the feelings are doing the arguing: music swells, powerful images, and urgent language that make a conclusion feel true before you've thought it through. When you notice your emotions spiking, ask yourself: "If I turn the sound off, does this argument still hold up?"
“If I turn the sound off, does this argument still hold up?”
Engagement Mechanics
Some choices in a video exist to serve the algorithm, not you: sensational thumbnails, artificial cliffhangers, questions designed to bait comments, urgency framing. None of these help you learn or understand anything — they're designed to maximize clicks and watch time. Ask: "Is this structured to help me understand something, or to keep me watching?"
“Is this structured to help me understand something, or to keep me watching?”
Group Characterization
Videos sometimes use shorthand for people and groups — the clueless bureaucrat, the dangerous outsider, the noble underdog. These shortcuts save time but they also train your expectations about what real people are like. Notice who gets to be a full, complicated person in this video and who gets reduced to a type.
“Who gets to be a full, complicated person in this video and who gets reduced to a type?”
Implicit Claims
Some ideas in a video are argued for openly — but the most powerful ones are just assumed. These are things the video treats as obviously true without ever making the case. They slip past your defenses because they were never presented as claims you could agree or disagree with. Ask: "What would I have to already believe for this argument to make sense?"
“What would I have to already believe for this argument to make sense?”
Story Shaping
A story can only include so much, so every video chooses what to show you and what to leave out. Story shaping is about those choices: who gets to speak, what context is given, and what's treated as obvious. When you finish watching, ask: "Whose perspective is missing here, and would the story change if they were included?"
“Whose perspective is missing here, and would the story change if they were included?”
Live Prompt Packs
These are the actual prompt packs running in production right now. Click any pack to see the full system prompt and response schema.
Version History
Every prompt pack version is recorded automatically on deploy. Click a version to see the full prompt at that point in time.
AI Content Detector
Influence Analyzer
Metadata Extractor
Theme Clusterer
Cross-Spectrum Calibration
Influence techniques differ by rhetorical tradition. Our prompt is calibrated to detect manipulation across the political spectrum with equal sensitivity. Read the full audit →
Conservative-coded
- Fear escalation → product/action pitch
- Explicit us-vs-them with named outgroups
- Sensationalist framing
Progressive-coded
- Moral authority as neutrality
- Institutional appeal as cover
- Empathy as compliance
- Consensus manufacturing
- Righteous outrage
Anti-establishment
- Revelation framing
- Systemic enemy (unfalsifiable villain)