Google announced conversational BigQuery analysis in Connected Sheets on 19 October 2026. The announcement explicitly says Gemini can query the full BigQuery dataset, not just the rows visible in the sheet. That difference can explain a total that initially looks inconsistent.
State the intended population
Before asking for a total, name the connected dataset, period and filters. Distinguish a sheet preview from the full authorized data source. A question such as “count records in the currently visible sample” is different from “count all records in this connected table”.
Say the preview shows ten rows from a fictional orders table. To check October totals, agree on the date field, the start and end of the period, and whether canceled orders count. Ask which filters were applied. Compare a small independently checked slice under the same conditions before using the full total.
Review the plan and refresh state
For a generated dashboard, inspect the proposed plan and the connected objects it creates. Check which filters and refresh settings apply before interpreting a change as a real-world trend. Availability requires eligible Gemini access and the underlying BigQuery permissions; a spreadsheet shared with you does not prove access to every source row.
Google listed Business Standard and Plus, Enterprise Standard and Plus, Google AI Pro and Ultra, Google AI Pro for Education, and AI Expanded Access. The gradual rollout began on 19 October in both release tracks. In your account, confirm the side panel is available in the connected sheet and that you have permission to query its data.