Methodology

The AI Fear & Greed Index aggregates two complementary signals to provide a balanced view of sentiment:

1. News Sentiment: Daily AI-related articles are analyzed and assigned a sentiment score on a -1 (fear) to +1 (greed) scale. Duplicate URLs and rewrites are deduplicated using host+path logic.
2. Community Sentiment: Visitors vote once per day to indicate fear, neutral, or greed. We compute same-day and rolling multi-day averages to smooth noise.

Source Selection

We prioritize reputable publications and high-signal community sources. Our aim is to capture material changes in AI capability, deployment, safety, regulation, and market impact.

Scoring and Normalization

• Article sentiment is normalized to -1 or +1 for aggregation consistency in the daily index.
• Duplicates across days are filtered so a single story does not over-weight the index.
• Community votes are averaged for today, 3-day, and 5-day windows.

Update Cadence

New data is generated once per day.

Limitations

• Sentiment scoring is an approximation and may not capture nuanced contexts. We are constantly improving and tweaking our algorithm.
• Coverage varies by day; low article volume can reduce precision.
• Community votes reflect our audience and may not be representative of the broader public.

Planned Improvements

• Richer topic tagging and category-level sentiment tracking (safety, regulation, capability, business).
• Longer horizon aggregates (7-day, 30-day) and comparative baselines.
• Additional source expansion and transparency dashboards.