Gmail's AI Overviews Went Global. Here Is What It Reads to Answer You

Type a question into Gmail’s search bar now and you may get an answer instead of a list of emails. A prompt like “what did the vendor say about the renewal price” can return a short summary pulled from several messages, not a set of results you still have to open and read. For a lot of Workspace accounts, this started happening without anyone opting in.

That is because the feature rolled out as a default, not a choice, wherever an organization already had the underlying settings turned on. If you work in Gmail all day, it is worth knowing exactly what changed, what it is reading to build those answers, and where the switch actually lives.

Gmail Search Now Answers Instead Of Listing

Google’s own Workspace Updates post announcing the rollout describes AI Overviews in Gmail search as reaching general availability globally for paid Workspace plans with English language settings, with the rollout beginning September 3, 2026. Ask a natural-language question in the search bar and Gemini reads across matching messages in your inbox to produce a short synthesized answer, with the source emails still available underneath it.

The feature depends on two settings many organizations already had switched on before this rollout: Gemini for Workspace, and a broader setting called Workspace Intelligence. Google’s admin documentation confirms AI Overviews in Gmail is on by default whenever both are already enabled. That is what made this feel sudden. Nobody at most companies made a September decision about Gmail search; a rollout activated a feature an earlier admin decision had already made possible.

This Is A Different Feature From Search’s AI Overviews

Worth separating clearly: this is not the same “AI Overviews” that appears at the top of a Google web search, the one we covered in how AI Overviews change what your title should say. That feature summarizes public web pages for a search engine’s visitors. This one summarizes your own private inbox, using content nobody else can see. Google reused the name for a feature with a completely different privacy shape.

Workspace Intelligence Decides What It Reads, And Admins Can Change That

Workspace Intelligence is the setting that governs which of your organization’s data sources a Gemini feature can search across, and Gmail is one of those sources alongside Drive, Calendar, and Chat. Google’s admin help documentation on controlling Workspace Intelligence walks through managing it from the Admin console under Generative AI, in the Gemini for Workspace section, where a Workspace Intelligence panel lists each data source with its own toggle. Turning off Gmail as a source there stops AI Overviews from drawing on inbox content, though Google’s documentation notes a user can still point Gemini at a specific email by naming it directly in a prompt. Changing the setting needs the Gemini Settings administrator privilege, and Google says it can take up to 48 hours to fully apply.

That is the practical answer to “how do I turn this off”: not a personal Gmail setting, but an admin console change scoped to the organization or a group of users. If your company has not made that decision on purpose, it is worth asking whether anyone has, since the default is that inbox summarization is available the moment the prerequisite settings exist. We wrote about this pattern more generally in when AI features turn on by default: what to check.

The Training Question Has A Specific Answer Here

The question that matters most for anything you would call confidential is whether Gmail content read by this feature ends up training a model. Google’s Generative AI in Google Workspace privacy hub states customer content is not used to train generative AI models outside the customer’s own domain without permission, and is not human reviewed for that purpose. Google also draws a line between this Workspace and Cloud commitment and the terms covering the free consumer version of Gemini, which are different. If your organization runs a Workspace account and also lets employees use consumer AI tools, that distinction is worth stating plainly. We covered how to ask a vendor these questions precisely in questions to ask an AI vendor about data, and what a strict retention promise does and does not cover in what zero data retention actually means.

Check This Before You Trust A Summary Of Your Own Inbox

None of this means the feature is unsafe to use. It means the two decisions, whether it is on and what it can touch, were made for most people rather than by them. Confirm three things this week: whether Workspace Intelligence includes Gmail as a source for your organization, whether IT made that choice deliberately or inherited it from an earlier rollout, and whether anyone handling client or patient correspondence knows a summarization feature now reads across that mail by default.

Flagging this to a team works better when it names the setting instead of gesturing at “the new Gmail AI thing.”

Before:

Heads up, Gmail has some new AI search thing now, might be worth checking.

After:

Gmail’s AI Overviews in search went live for our Workspace plan on September 3. It reads across inbox content when Workspace Intelligence includes Gmail as a source, which ours does. If you handle client correspondence under an NDA, flag it to IT this week so we can decide whether to exclude Gmail as a source for that group.

A Wrivio Context for a settings advisory like this one could say:

Rewrite this as a clear internal notice. Keep every setting name, date, and toggle exactly as written. State the specific action each reader should take. Do not soften a decision that needs to happen into a vague suggestion to “check settings.”

Press Ctrl+Shift+Space, paste the draft, and check the diff. A rewrite that keeps “Workspace Intelligence” and “September 3” intact is doing its job; one that turns them into “recent updates” has thrown away the two facts that make the message actionable.

Common Questions

Is Gmail’s AI Overviews the same feature as Google Search’s AI Overviews?

No. They share a name but not a purpose. Search’s AI Overviews summarizes public web pages for search visitors. Gmail’s AI Overviews summarizes your own private inbox content to answer a question you typed into Gmail search, and only you can see the result.

Is this feature on by default?

Yes, for organizations that already had Gemini for Workspace and Workspace Intelligence enabled with Gmail included as a data source. Google’s admin documentation describes it as active automatically once those prerequisites are in place, rather than something an individual user turns on.

How do I turn it off?

It is not a personal setting inside Gmail. An administrator has to open the Admin console, go to Generative AI, then Gemini for Workspace, and turn off Gmail as a data source in the Workspace Intelligence panel. Google says the change can take up to 48 hours to fully apply.

Does Google use my email content to train its models?

Google’s Workspace privacy documentation states that customer content is not used to train generative AI models outside the customer’s own domain without permission, and is not human reviewed for that purpose. That commitment applies to Workspace and Cloud accounts specifically and is stated to differ from the terms covering the free consumer version of Gemini.

Should I stop using Gmail search because of this?

Not necessarily. It is a reason to confirm the setting is a deliberate choice rather than an inherited default, particularly if your inbox carries client, patient, or legal correspondence with confidentiality obligations attached to it.

Download Wrivio for Windows to draft and rewrite that internal notice, or anything else that depends on getting a setting name and a date exactly right.