# AI Overviews click gap: methodology

Updated: 8 August 2026

## Scope

This package is a secondary analysis of Pew Research Center's published browsing-panel results about Google searches with and without an AI summary. BGraph did not receive or process participant-level browsing data.

## Source input

Pew analyzed browsing activity from 900 U.S. adults during 1 to 31 March 2025. The published analysis covered 68,879 unique Google searches, including 12,593 searches that produced an AI summary when results were collected during 7 to 17 April 2025.

## Transformations

Published percentages were transcribed into a flat table. Two rows are explicitly marked as BGraph calculations:

- Percentage point difference: 8 minus 15 equals -7 points.
- Relative difference: 8 divided by 15 minus 1 equals approximately -46.7%.

## Reproducibility

Use rows `click-001` through `click-005` for the next-action chart. Use `click-006` through `click-008` for AI summary presence. Calculated rows use rounded published inputs and should be reported as approximate.

## Limitations

- The sample covers U.S. adults and Google searches during one month.
- AI summaries were reconstructed after the browsing period and can change over time.
- Presence of a summary was not randomly assigned, so the click gap is not a causal estimate.
- Pew's next-action categories depend on the next timestamped URL and a defined session-ending rule.
- Percentages are rounded and the underlying participant-level data is not included.

## Preferred citation

BGraph Research. (2026). AI summaries change what happens after a Google search. Updated 8 August 2026. https://bgraph.io/research/ai-overviews-click-gap.