Every email message should carry a Message-ID header, a unique label that identifies that exact message. Mail clients use it to build conversation threads and to remove duplicates, mail servers write it in logs so that administrators can follow a message through a chain of systems, and spam filters look at it as one of many small signals of whether the sender is a real mail system. Most of the time the ID is generated for you and nobody notices it. It becomes important when you send mail from your own code, when you switch providers, when replies fail to thread, when a message is delivered twice or when you need to find one message among millions in a log. This guide explains the format defined by the standard, how to generate good IDs, what happens when you leave it out, how it relates to In-Reply-To and References for threading and how to use it as a tracing key in practice.
The format
A Message-ID has the form <left@right>. The part on the left is any string of permitted characters that makes the ID unique; the part on the right is normally a domain name or the hostname of the system that generated it. The angle brackets are part of the value. A typical example looks like this:
Message-ID: <[email protected]>
Date: Mon, 26 Oct 2026 10:15:00 +0000
From: Example <[email protected]>
To: [email protected]
Subject: Your receiptThe standard asks for global uniqueness, which you achieve by combining something unique on the left (a timestamp with high resolution plus a random string, or a UUID) with a right-hand side that you control. Because the domain belongs to you, no other system will generate the same ID. The left side must not contain spaces or characters that need quoting, and the total length should stay modest, because some software limits lines to 998 characters and mail clients may truncate long values.
Which domain to use on the right
| Choice | Pros | Cons |
|---|---|---|
| Your sending domain (example.com) | Matches the From domain; looks consistent to filters | Needs care not to collide with IDs created by other systems that use the same domain |
| A mail subdomain (mail.example.com) | Separates IDs by system or stream; easy to trace | Another name to manage |
| The server hostname | Traditional; shows which host created the message | May expose internal names; may not match anything in DNS |
| localhost or an IP literal | None | Looks like a script; commonly penalised by spam filters |
| The provider's domain | The provider generates it for you | The ID says nothing about your system |
Spam filters have long checked for IDs that look machine-made and careless, such as those ending in @localhost, @localhost.localdomain or a name that is clearly not a real domain. A domain that resolves in DNS and relates to the sender is a modest positive signal. No filter will accept or reject a message on this header alone, but it is one of several things that distinguish well-formed mail from sloppy mail. Read how content filters judge a message.
What happens if it is missing
The standard says that a Message-ID should be present. If your application does not add one, many mail servers will add it for you on submission, and some relays will too. If it is added late, you lose the ability to know the ID at the moment you sent the message, which makes tracing harder. If nothing adds it, the message may be treated with more suspicion, some filters add a small penalty for a missing header, and threading by clients that rely on it may fail. For this reason you should either let your sending library generate it or set it yourself. Most libraries do so by default: PHPMailer, Nodemailer, Python's email.utils.make_msgid() and Java Mail all have helpers.
# Python: generate a Message-ID on your own domain
from email.utils import make_msgid
msg['Message-ID'] = make_msgid(domain='mail.example.com')Threading: In-Reply-To and References
When a user replies to a message, the client writes two headers. In-Reply-To holds the Message-ID of the message being answered. References holds the chain: the IDs of earlier messages in the conversation, with the most recent last. Mail clients and webmail use these to place the reply in the right thread. If you send automated follow-ups and want them to appear in the same conversation, copy the original Message-ID into In-Reply-To and append it to References. The subject line should also match, with the usual "Re:" prefix, because some clients thread by subject as a fallback.
Message-ID: <[email protected]>
In-Reply-To: <[email protected]>
References: <[email protected]>
Subject: Re: Your receiptFor support systems and inbound parsing it works in the opposite direction: when a customer replies, your system reads In-Reply-To or References to attach the reply to the right ticket. See inbound email processing. Be careful that a provider may rewrite the Message-ID when it relays; if you rely on the ID to thread replies, store the ID that appears in the delivered message, not only the one you generated.
Using the ID to trace a message
The Message-ID is the key that ties together records in different systems: your application log, the provider's message log, the recipient's headers and a bounce message. When a customer says that an email never arrived, ask for the time and the recipient, find the message in your logs and note its ID. A provider with a message log can then show the delivery events for that ID. In a bounce, the original headers are often included, so the ID lets you match a bounce to the message that caused it. In a complaint feedback report the original message or its headers are included for the same reason. See tracing email with message logs and reading headers.
Common problems
| Problem | Cause | Fix |
|---|---|---|
| Replies do not thread | In-Reply-To or References missing, or the provider replaced the ID | Store the delivered ID and copy it into the follow-up |
| Message appears twice for one recipient | Retry sent a second message with a new ID after an ambiguous failure | Use a stable ID per logical message and an idempotency check |
| Two different messages share an ID | Weak generator, for example a timestamp in seconds only | Add random bytes or use a UUID |
| ID ends with localhost | Library default on a server without a hostname | Set the domain explicitly |
| Cannot find a message in provider logs | Searching by an ID the provider changed | Search by recipient and time, or ask the API for the provider ID |
| Header rejected as malformed | Missing angle brackets, spaces, non-ASCII characters | Follow the format strictly |
Good practice for developers
- Generate the ID in your application, on a domain you control, with at least 64 bits of randomness or a UUID.
- Save it with the record of the send, along with the recipient and the time.
- Keep the same ID when you retry the same logical message, so that duplicates are easy to spot.
- Do not put personal data, such as an email address or user name, in the left part. IDs appear in headers that third parties can read.
- Preserve existing IDs when you forward or redirect, and add new ones only to new messages.
- Return the ID, or the provider's equivalent, to your own system so that support staff can look it up.
The Message-ID is a small header with a large effect on how easily you can run an email system. A correct, unique ID costs nothing to produce and pays back every time you need to find, thread or de-duplicate a message.
Trace every message you send Message logs and event webhooks let you follow a message from API call to delivery.
References and further reading
Frequently asked questions
What is the Message-ID header?
A unique identifier for one email message, written as <unique-part@domain>. Clients use it for threading and servers use it in logs.
Does a missing Message-ID hurt deliverability?
It can. Many systems add one, but a missing or sloppy ID, such as one ending in localhost, is a minor negative signal.
How do I make replies thread correctly?
Copy the original Message-ID into In-Reply-To, append it to References and keep a matching Re: subject.
Which domain should I use in the Message-ID?
A domain or subdomain you control, ideally related to your sending domain.
Can I use the Message-ID to find a message in logs?
Yes. It is the main correlation key between your application, your provider's log, headers and bounces.