Gemini in Google Sheets can now query your whole BigQuery table. Set a cost limit before you share it

Can Gemini in Google Sheets analyze BigQuery data now?

Google announced this on the Workspace Updates blog on 19 October 2026. You connect BigQuery data with Connected Sheets, and then you can ask Gemini for total sales in one region last quarter. You can follow up to ask how that compares to the year before, and then tell it to build a dashboard. Gemini will show you its plan first, and then it builds a new tab with connected charts, pivot tables and database formulas that refresh on a schedule.

The rollout is gradual and can take up to 15 days from 19 October 2026 for both Rapid Release and Scheduled Release domains. You need a Business or Enterprise plan on the Workspace side, either Standard or Plus. Personal accounts need Google AI Pro or Ultra, and the feature also comes with two add-ons, Google AI Pro for Education and AI Expanded Access. Google says it is on by default with no admin control, so it shows up for anyone with permission to query the connected data.

I have not run this on a real client dataset yet. Everything below comes from Google’s own pages and from how BigQuery billing already works. I will tell you plainly if Google does not state something directly.

What does Gemini actually run when I ask a question?

It runs a query. Google’s Connected Sheets documentation says Gemini translates your question into SQL, and SQL is the language BigQuery understands. That SQL runs in BigQuery against the whole table, just like any other query.

This is the part that is easy to miss. A connected sheet only shows a preview of 500 rows, so it feels small. But Gemini’s answer is not built from those 500 rows. Google says the queries run against the full dataset, which is the whole point of the feature, and it also means a casual question can read a very large table.

The good news is on the safety side. Google’s help page Collaborate with Gemini in Google Sheets says Gemini runs read-only queries, and it does not modify or delete your data in BigQuery. The risk here involves your bill and who sees the results, so you do not need to worry about someone losing data.

Who pays for the queries Gemini runs?

That same help page says you need BigQuery access and a Google Cloud billing project to query with Gemini. You pick a Google Cloud project with billing turned on when you set up Connected Sheets, and your queries from the sheet go there. Google’s post does not explicitly say Gemini’s queries land on that same bill, but I expect they do, and you can check this after the first session in the project’s job history.

You pay for the bytes a query reads on BigQuery’s on-demand pricing. Google’s page on controlling BigQuery costs says each project gets 1 TB of free querying per month, and anything above that is billed. BigQuery stores each column separately. So what a query reads depends on the columns it touches and how much of the table it scans.

I will use an example with made-up numbers. Say your orders table holds 200 GB. A question that reads most of it uses up a fifth of the free terabyte in one go, and five of those questions in a month will clear out your free tier. A business owner asking ten follow-up questions in an afternoon would never know because the sheet does not show the query size.

SQL habits can set a trap here. Adding LIMIT 10 to a query only cuts the number of rows that come back. Google’s cost page says LIMIT does not change the amount of data read on a table that is not clustered. A request to show the top ten can cost the same as the full answer.

How do I stop Gemini from running up a BigQuery bill?

You set the limits in BigQuery. Sheets has no admin control for this feature according to Google. I would take these steps before sharing a connected sheet with anyone who does not already write SQL.

1. 01
   Find out which project pays

   Note which Google Cloud billing project the connection uses. If that project also runs other work, give the sheet a project of its own, so its spend is easy to see on its own.
2. 02
   Set a daily query quota

   BigQuery lets you set a custom quota on the amount of data processed per day, for the whole project or per user. Set the per user one. When someone reaches it, their queries stop for the day instead of the bill growing.
3. 03
   Connect a view, not the raw table

   Ask whoever manages BigQuery for a view with only the columns the team needs. A question can never read a column the view does not have. If the table is partitioned by date, a view limited to recent dates also keeps old data out of every question.
4. 04
   Count the refreshes

   Every chart, pivot table and formula on a Gemini dashboard stays connected. Each refresh sends its queries to BigQuery again, so a tab with eight objects on an hourly schedule runs far more than eight queries a day.
5. 05
   Read the job history after the first week

   The Connected Sheets docs show how to read billed bytes from the INFORMATION_SCHEMA jobs views. Look at it once after a week of real use, per user, before you relax any limit.

Can people without BigQuery access see the answers?

This next part matters more for client work than the bill does. Google’s help page says collaborators with view access to the spreadsheet can view saved results even if they lack BigQuery permissions. Running new queries or refreshing data requires BigQuery access, but the answers already inside the sheet are visible to anyone who opens it.

Think about what this means for your data. Your data team may have locked down who can query the customer table, and that lock holds firm, but someone with access might ask Gemini to pull a list of top customers into the sheet. If that sheet is shared with the whole company or a wide link, everyone who opens it can read the list. The BigQuery permission controls the asking, and the sheet’s share setting controls the reading.

This is the same gap I wrote about with exporting Power BI data from a phone, where the data leaves its protected place and lands somewhere with looser rules. Check the share button on any connected sheet before Gemini puts anything in it. I recommend keeping one sheet for asking questions and a separate one for sharing finished results.

How do I know Gemini's answer is right?

Read what Gemini ran. Google’s help page says you can open the plan card and review it before you click Apply. You can use Analysis steps to see the steps Gemini took, and you can use Show code to see the code behind the result. The same page warns that Gemini can suggest inaccurate information, so check the SQL code, because the summary sentence can be wrong.

You do not need fluency in SQL to catch the common mistakes. I check the date range in the WHERE line, the column it summed, and whether it filtered out items like cancelled or test orders. An AI tool gets a business number wrong mostly because of a wrong period, a wrong measure, or including rows that should have been left out.

The last row of that table is the habit that catches most of the errors. You pick one number you already know, like last month’s revenue from your finance report, and ask Gemini for it first. The rest of the session is on a good footing if it matches, and I go through the same idea for formulas in how to check the numbers an AI put in your spreadsheet and for web lookups in checking the sources Fill with Gemini gives you.

Should I use Gemini with Connected Sheets or a normal dashboard?

Gemini in a connected sheet is quicker than waiting for someone to write SQL for a question you ask once. But for a number people look at every week, I would still build the report properly. You can do this as a tab Gemini drafts and someone checks, or in a tool like a Looker Studio client dashboard, so the logic is fixed and reviewed once, and you avoid asking a fresh question each time.

That last row is worth saying out loud. You do not need BigQuery if your data fits in a normal Google Sheet, which I cover in the Google Sheets cell limit, and the plain AI function in Google Sheets may be enough for your needs.