As of 2026-11-12, Google is rolling out Gemini's "Help me write" in Gmail to every Workspace customer and every personal account aged 18 or over, and nothing Google has published lets a recipient confirm that a given email came from it. You can only judge the writing. GeminiDetect does that for messages of 100 words or more and reports how closely the style matches Gemini, which is a signal, not proof.
What Google announced
On 2026-11-10, Google's Workspace Updates blog announced that two Gemini features in Gmail are expanding:
- Help me write (non-personalized). It generates a complete email draft or refines existing text from a short prompt.
- AI Overview. It summarizes long email threads at the top of the conversation.
They reach Workspace customers, Workspace Individual subscribers and personal accounts, for users aged 18 or over with smart features in Gmail turned on. Before, they were limited to consumer accounts and some Workspace editions. The features are on by default, end users can turn them off, and admins can disable them. Rollout is gradual from 2026-11-10 and can take up to 15 days to appear. TalkAndroid reported the same details on 2026-11-11.
What is confirmed and what is not
| Question | Status | Source |
|---|---|---|
| Is Help me write reaching more Gmail users? | Yes, from 2026-11-10 | Google Workspace Updates |
| Is it on by default? | Yes; users and admins can switch it off | Google Workspace Updates |
| Is a draft labelled for recipients? | Not mentioned in the post | Google Workspace Updates |
| Does Google watermark Gemini text? | It says SynthID covers text from the Gemini app and web experience; Gmail is not named | Google DeepMind |
| Can the public check text for that watermark? | No public text checker found; the detector covers images, video and audio | Google blog, 2026-10-07 |
The last two rows are why this site never claims to read a watermark. See what Google's tools can and cannot check for the full status.
Why a single email is hard to judge
Help me write can also refine text a person already wrote, so one message can be partly human and partly Gemini. Business email is formulaic to begin with: greetings, thanks, follow-ups and confirmations read alike whether a person or a model wrote them. Many emails are also too short. Below 100 words there is not enough text to read, and the right answer is that the message cannot be judged.
Longer messages are different. Proposals, cover letters, complaints, vendor pitches and newsletter drafts give a detector enough to work with. The patterns that tend to separate Gemini from human writing are described in how to tell if Gemini wrote it.
What the EU rules add
Article 50 of the EU AI Act applies from 2 August 2026. The European Commission's FAQ says providers must mark AI-generated text, among other content, in a machine-readable format, and that providers of systems already on the market before that date must comply only from 2 December 2026. That is an obligation on providers. It does not give readers a checker, and it does not say how any particular Gmail draft is marked.
How to check an email
- Paste the full message, without the quoted thread below it, into GeminiDetect. You need at least 100 words.
- Read the highlighted passages, not only the verdict. A flag on the greeting and sign-off means little. A flag on the substantive paragraphs means more.
- Compare with the sender's earlier emails. A sudden change in register is a clue that no score can replace.
- If it matters, ask the sender. A person who wrote it can usually explain it.
On human-written text in seven languages, 0.5% was wrongly flagged, and 94.9% of text from AI models never used in development was detected; the facts page lists both. A result is a reason to look closer, and heavily edited text carries less signal.
Check a text
Have a message you doubt? Paste it into GeminiDetect for a free check.