---
title: "Checks before you send"
description: "Your draft scored for spam, phishing and AI writing as you write it, the way a receiving server would read it."
url: "https://openemail.uk/docs/knowledge/writing/draft-checks"
area: "Knowledge base"
category: "Writing & sending"
status: "live"
---

# Checks before you send

Your draft scored for spam, phishing and AI writing as you write it, the way a receiving server would read it.

## Details

- Three rings sit beside Send in the composer: spam, phishing and AI writing, each scored out of 100 and redrawn a moment after you stop typing. Hover one to read its verdict and what it flagged. The assistant shows the same three on a message before you approve sending it.
- They are the checks incoming mail is scored by, run on your subject and body. No model is called and nothing is kept: the draft is scored on our server and dropped. Spam reads as unlikely under 35 and likely from 60, phishing as worth a second look from 30 and a scam from 60, and AI writing as possible from 35 and likely from 60.
- Your sender checks are assumed to pass, since the receiving server is what decides them, so the rings judge the content alone. Under 40 words is too short for an AI score. In the composer the signature and the attachments are not part of it, and a send to an audience or from a template is not checked.
- The same check is on the REST API as POST /emails/check, on the SDK as emails.check, on the command line as openemail emails check and on the MCP server as checkEmail, so your code or an agent can test a message before sending it.