AI assistants now draft, send and answer email, and the same questions come up again and again: can the AI send for me, why is my signature missing, do I have to tell people, and will it land in spam. This page answers each one in a few sentences and links to the guide that covers it in depth. Legal answers are general information, not legal advice.
The questions are grouped in five sections:
- Sending email with AI agents
- Email signatures on AI-sent email
- Disclosure and the law
- Ethics and trust
- Deliverability
Give every AI agent a real signature
Signatoro gives your team and your AI agents one signature each, fetched over MCP or the API every time they send.
AI agent sending questions
Can AI send emails for you?
Yes. Assistants such as Claude and ChatGPT can send email through Gmail or Outlook connectors, and automation tools such as n8n, Zapier and Make can send email an AI wrote. Most chat assistants ask for your approval before each send by default. The AI agent setup guides cover each tool.
How do I get Claude to send an email on my behalf?
Connect Claude to Gmail or Microsoft 365. Anthropic's help center says Claude "can send, reply to, and forward emails from Gmail, and asks for your approval by default before each of these actions", and on Team and Enterprise plans the owners decide whether members can let those actions run without asking. The Claude guide covers each route.
Can ChatGPT send emails from Gmail or Outlook?
Yes, when Gmail or Outlook is connected. ChatGPT creates the draft in the chat and asks before it sends, and in business workspaces an admin has to turn the write actions on. The ChatGPT guide has the details.
Is there an MCP server for Gmail or Outlook?
Yes. Several MCP servers let an AI assistant read, draft and send Gmail or Outlook email. Most of them send through the Gmail API or Microsoft Graph, so the email goes out without your signature unless the agent adds it. The Gmail and Outlook MCP guide covers setup and signatures.
Is it safe to give AI access to your email?
It is as safe as the permissions you grant and the checks you keep. The MCP specification says there "SHOULD always be a human in the loop with the ability to deny tool invocations", and Claude and ChatGPT ask before sending by default. Grant the narrowest access the task needs, and keep approval on for anything sent to customers.
Should an AI agent have its own email inbox?
It depends on whose voice it speaks in. An agent that acts for you can send from your mailbox with a signature that says it is your assistant; an agent that acts for the company, such as a support or sales agent, often works better from its own address. Either way, give it its own signature. The AI agent email signature hub covers agent inboxes such as AgentMail and Nylas.
What is the difference between AI drafting and AI sending an email?
When AI drafts, a person reads the message and clicks send, so the email is that person's. When AI sends, no one reads it first. The difference matters for trust and, in the EU, for the law: the European Commission treats an agent that writes and sends messages to people as interacting with them directly.
Email signature questions
Why does my AI-drafted email have no signature?
Gmail and Outlook add your signature in their compose window, and AI agents send through the Gmail API or Microsoft Graph, which send only the body they are given. Google's Gmail API reference says the signature is included in "messages composed with this alias in the Gmail web UI." The agent has to append the signature itself. See why AI agent emails have no signature.
Can Claude or ChatGPT use my Gmail signature?
Not automatically. Their Gmail connectors build the email outside Gmail's compose window, so the signature saved in Gmail is not added. Paste your signature as text into the assistant's instructions, or connect it to a signature source over MCP; the Claude and ChatGPT guides show both.
Does the Gmail API add my signature to emails I send?
No. messages.send sends the message you built, and the Gmail signature belongs to the web compose window. Read the signature from users.settings.sendAs or a signature service and add it to the body before sending. The Gmail API guide has code.
Can I get a user's Outlook signature with Microsoft Graph?
No. Microsoft Graph documents no endpoint that returns a user's Outlook signature, and sendMail sends only the body you pass. Store the signature in your app or a signature service and append it before sending. See the Microsoft Graph guide.
How do I add my signature to emails n8n, Zapier or Make send?
Zapier's Gmail actions can pull your Gmail default signature; n8n, Make, and Zapier's Outlook actions need the signature put into the body. A step that fetches the signature on every run keeps it current. The n8n, Zapier and Make guide has the steps for each.
Why does Copilot put my name twice?
Copilot in Outlook drafts in the compose window, where Outlook already adds your signature, and users report that it also writes its own sign-off with your name. Tell Copilot in its draft instructions not to add a sign-off or your name. The Copilot guide shows where.
How do I stop AI from making up my phone number or signing "[Your Name]"?
Keep the signature out of the model's hands. Tell the model to end the email without a sign-off, and have the workflow append your real signature from a fixed field or a fetch. A model cannot invent contact details it never writes.
Can my whole team and our AI agents share one signature design?
Yes. With a company signature tool, each person and each agent gets their own name and details inside one shared layout. On Signatoro, the owner sets the logo, layout and company details once, and every person and agent fetches the current version when they send.
Disclosure and law questions
Do I have to tell people an email was written by AI?
It depends on who sends it and where. In the EU, the Commission's guidelines say AI agents that write and send messages to people should disclose that they are AI and on whose behalf they act; a person using AI to help write their own email is generally outside that duty. Most other places have no general rule for AI-drafted email. The AI email disclosure guide compares the US, UK and EU.
What is Article 50 of the EU AI Act?
Article 50 sets transparency duties for certain AI systems: people must be told when they are interacting with an AI unless it is obvious, and providers of generative AI must mark its output as AI-generated in a machine-readable way. It has applied since 2 August 2026. The Article 50 email guide explains what it means for email.
Does the EU AI Act apply to emails sent by AI agents?
Generally, yes, for agents that send on their own. The Commission's guidelines name "managing correspondence" as an agent task in scope, count the ability "to write and send messages to natural persons" as direct interaction, and give as an example "an email generated by an AI agent sent to a natural person that features an AI label at the top".
Does the EU AI Act apply to US companies?
It can. Article 2 covers providers that place AI systems on the EU market wherever they are based, and providers and deployers in other countries "where the output produced by the AI system is used in the Union". A US company whose AI agent emails people in the EU should check how the Act applies to it.
What is a good AI disclosure statement for email?
One line that says what the AI did, names the person or company it acted for, and offers a way to reach a person. For example: "This email was written and sent by an AI assistant on behalf of Maya Okafor, Head of Operations at Northwind Supply." See 20 AI disclosure statement examples.
Should an AI disclaimer go in the email signature?
The signature is the natural place for the "on whose behalf" part: the name, role and company the AI acts for. For an agent that sends on its own, add a short label at the top of the message as well, as in the European Commission's own example.
Is it legal for an AI agent to send cold email?
The usual email marketing rules apply whoever writes the email. In the US, CAN-SPAM requires commercial email to identify itself as an ad, offer an opt-out and include "a valid physical postal address of the sender". In the UK and EU, marketing email to individuals generally needs consent, with a narrow exception for existing customers. See the CAN-SPAM and PECR guides.
Does CAN-SPAM apply to AI-generated emails?
Yes. CAN-SPAM makes no distinction between emails a person writes and emails an AI writes, so a commercial email an agent sends needs the same opt-out and postal address. An agent sending through an API leaves out the footer that usually carries them unless it adds it.
Ethics and trust questions
Is it OK for AI to send email in my name?
It is OK when you read and send each email yourself. When the AI sends on its own, sign as your assistant, for example "Ava, AI assistant to Maya Okafor", so the reader knows who wrote it and who stands behind it. See should your AI assistant sign emails as you.
Is it unethical to use AI to write emails?
Using AI to write emails is widely accepted when you review what you send. The ethical line is deception: an AI presenting itself as a person, or sending in someone's name without their knowledge.
Can my employer use AI to send emails as me?
Employers generally control company mailboxes, and the law here is unsettled and varies by country. An agent sending in your name without your knowledge raises consent and reputation problems; ask for the agent to sign with its own name and for access to what it sends.
Who is responsible if an AI email says something wrong?
The person or company that sent it. FINRA tells firms they "are responsible for their communications, regardless of whether they are generated by a human or AI technology", and Utah law says it is not a defense that generative AI made the statement.
Deliverability questions
Do AI-written emails go to spam?
Gmail's sender guidelines say nothing about AI. They require every sender to authenticate with SPF or DKIM and to keep spam rates reported in Postmaster Tools below 0.3%. An AI email is filtered like any other email: by the sender's authentication, reputation and how recipients react.
Do AI agents sending email need SPF, DKIM and DMARC?
An agent sending from your own Gmail or Microsoft 365 mailbox uses that mailbox's setup. An agent sending from its own domain or service needs SPF or DKIM, and Gmail requires senders of more than 5,000 messages a day to Gmail accounts to have SPF, DKIM and DMARC plus one-click unsubscribe for marketing mail.
Does an image in my email signature hurt deliverability?
Gmail's sender guidelines set no rule on signature images; they ask that HTML follow web standards and not hide content. Some cold email senders still use a short, text-only signature for outreach. Signatoro's signature API returns a short version for that, with name, role and one link.
Sources
Checked on 2026-10-07:
- European Commission: Guidelines on the transparency obligations under Article 50 of the AI Act, C(2026) 5054, 20 July 2026
- EUR-Lex: Regulation (EU) 2024/1689 (AI Act), Articles 2 and 50
- Cornell LII: 15 U.S. Code 7704 (CAN-SPAM)
- FINRA: Advertising Regulation FAQ, D.8 AI Created Communications
- Utah Legislature: S.B. 226 (2025), Utah Code 13-75-102
- Google: REST Resource: users.settings.sendAs (Gmail API)
- Google: Email sender guidelines
- Microsoft: user: sendMail (Microsoft Graph v1.0)
- Anthropic: Use Google Workspace connectors
- Model Context Protocol: Specification 2026-07-28, Tools