A Knowmler is a creature of curiosity. The name combines know (the pursuit of understanding) and nom nom (the sound of enthusiastic consumption). Feed it content and it digests complexity into clarity.
When you submit content for analysis, Knowmler spins up specialized AI agents. Each agent has a focused prompt engineered for a specific task. The agents run in parallel when possible, then their outputs are aggregated into your final report.
Evaluates content across 5 dimensions: Factual Accuracy, Source Quality, Logical Coherence, Emotional Manipulation, and Bias Indicators. Outputs a trust score with detailed breakdowns.
Scans for 34+ persuasion and propaganda techniques including: appeal to fear, false dichotomy, bandwagon effect, strawman arguments, cherry-picking data, emotional appeals, and more. Each detection includes the specific quote and severity rating.
Identifies factual claims, opinions, predictions, and recommendations. Each claim is categorized by type and verifiability. Great for fact-checking prep work.
Extracts themes, key topics, actionable insights, and generates follow-up questions. Useful for research and deep dives into unfamiliar subjects.
Each agent receives a carefully engineered system prompt that defines its role, output format, and evaluation criteria. Your content is injected into the prompt along with any context (like video metadata or PDF structure).
The prompts are versioned and tested. When we improve a prompt, we can A/B test it against the previous version to ensure quality improvements before rolling out.
Beyond analysis, Knowmler can generate various outputs from your content:
A summary scorecard showing all detected manipulation techniques, severity levels, and specific examples from the content.
Generated prompts to help you investigate claims further, design experiments, or explore related topics.
Prompts designed to challenge the claims made in the content, helping you think critically about both sides.
Image generation prompts, writing prompts, or other creative outputs inspired by the content. Yes, even nanobanana images.
We regularly evaluate new AI models to ensure we are using the best available for nuanced rhetorical analysis, claim detection, and manipulation identification.
As of January 2026, we have selected Claude Opus 4.5 as the best model for this work.
Uses YouTube Data API with youtube.readonly scope. We fetch video metadata and available captions. OAuth tokens are encrypted at rest using Fernet symmetric encryption.
PDFs are uploaded to S3, text extracted, then chunked into 30K word segments for analysis. Large documents are processed chunk-by-chunk with results aggregated.
Analyses run asynchronously. When you submit content, it enters a queue. You can watch progress in real-time or come back later for results.
OAuth 2.0 authentication, encrypted tokens, HTTPS everywhere, CORS restrictions, and rate limiting. Your content is processed and stored securely.
Your data belongs to you. We do not sell your information. We only collect what is necessary to provide the service. You can delete your account and all associated data at any time.
Read our full Privacy Policy for details.
Questions or feedback? Reach out at [email protected]