Engagement Bait API

• 

  Demo site interface

• 

  Demo site interface

Inspiration

With the rise of algorithms, content that prioritizes engagement over facts has become a major issues. We built this API to detect content that's aimed at getting an emotional rise out of someone, instead of reflects the facts.

What it does

Engagement Bait API analyzes text and returns interpretable scores across dimensions such as urgency pressure, evidence density, overconfidence, emotional arousal, and narrative simplification. Instead of labeling content as “good” or “bad,” it quantifies how strongly the content is engineered to drive engagement.

How we built it

We built Engagement Bait API as a FastAPI service that analyzes text with a rule-based layer for a variety of factors regarding engagement bait. We then added an OpenAI embedding-based semantic score, a browser demo, and an internal benchmark workflow to make the system both usable and easy to validate.

Challenges we ran into

One major challenge was defining manipulation in a way that measures structure rather than ideology or truthfulness. We also had to carefully design scoring methods that are meaningful, interpretable, and not overly sensitive to specific writing styles.

Accomplishments that we're proud of

We built a framework for measuring persuasion signals without relying on black-box classification. We’re especially proud of designing metrics that developers can eventually integrate into applications and research workflows, along with producing strong documentation around the API.

What we learned

Persuasion and engagement contians patterns that are distinguible from content validity. By separating the way something is written from whether or not its true, we realized that narrative pressue can be measured.

What's next for Engagement Bait API

We hope to continue to refine the embeddings section of the project. Our goal is to strengthen detection of subtle engagement bait signals while preserving transparency.

Built With

• claude
• codex
• cursor
• fastapi
• openaiapi
• python
• render