Methodology

Data Source

All indices are built from Google Trends "Interest over time" data configured for the United States, Web Search category, from 2004 to present.

Google Trends data is not raw search volume; it is a normalized index (0–100) representing search interest relative to the highest point on the chart for the given region and time.

Composite Calculation

Each index tracks a set of thematically linked search terms (e.g., "ai course" + "learn ai"). The composite index is calculated by summing the normalized interest values of all component terms.

To ensure consistency across views (5-Year vs. All-Time), the final composite series is re-normalized (0–100) based on the all-time maximum sum since 2004.

Year-over-Year (YoY) Metrics

To filter out weekly volatility and provide a stable signal, we use a 4-week trailing average.

Current Value: Mean of the last 4 weeks.
Prior Value: Mean of the equivalent 4-week period one year ago.
YoY Change: The percentage growth or decline between Current and Prior.

Guardrail: If the prior year's baseline is extremely low (normalized index < 5), we display the change in points (pixels) rather than percentage to avoid astronomical but meaningless growth figures (e.g., "+500%").

Caveats & Limitations

These indices are proxies for public attention, not direct economic measures. Spikes can be driven by news cycles, viral media, or changes in platform features (e.g., Google's UI changes).

Furthermore, semantic overlap exists; searching "will AI replace jobs" might signal anxiety, curiosity, or academic research. We assume that in aggregate, the trend direction provides a valid signal of shifting public sentiment.