Home / Blog / How email verification actually works

How email verification actually works

By Andrej Uzusienis · November 16, 2026 · 9 min read

"Email verification" sounds like a single operation. It is not. It is a short pipeline of independent checks, each answering a different question, each with its own failure modes. Understanding what each stage proves — and, more usefully, what it does not prove — is the difference between trusting a result and over-trusting it.

Here is the whole pipeline, in the order it runs, with the reasoning behind each step.

Stage 1: syntax, against RFC 5322

The first check parses the address into a local part and a domain and tests both against the grammar in RFC 5322. This is cheap, happens locally, and needs no network at all.

It catches the obvious damage that real lists are full of: a trailing comma, a space in the middle, user@@domain.com, user@domain..com, a local part that starts or ends with a dot, an unterminated quoted string.

What makes this harder than it looks is that the real grammar is much more permissive than most people expect. "very.unusual.@.unusual.com"@example.com is a legal address. So is a local part containing !#$%&'*+-/=?^_`{|}~. Validators written as a single regular expression almost always reject some legal addresses, and the ones they reject tend to belong to technically literate people who notice.

The practical compromise most verifiers make, ours included, is to parse strictly enough to catch damage but not so strictly that exotic-but-legal addresses are thrown away. Syntax is also where internationalised addresses matter: a domain in a non-Latin script has to be converted to its Punycode A-label form before any DNS lookup can be attempted.

What this stage proves: the string is a well-formed address. What it does not prove: anything at all about whether the mailbox exists.

Stage 2: MX resolution

Next, the domain is looked up in DNS for MX records. An MX record names the host that accepts mail for a domain, with a preference number that sets the order in which senders should try them.

Three outcomes matter here:

MX failures are the cheapest possible save: a domain with no mail server is a guaranteed hard bounce for every address on it, and in a list that has been collecting for a few years there are usually more dead domains than people expect. MX records are explained in full here.

What this stage proves: the domain can or cannot receive mail. What it does not prove: that a particular mailbox on the domain exists.

Stage 3: disposable-provider matching

Temporary-mail services give out an address that works for a few minutes and then stops. People use them to collect a download or a discount code without handing over a real address, which means the address was never intended to receive your later mail.

Detection is domain matching against a maintained list. The hard part is maintenance, not matching: the larger providers rotate through dozens or hundreds of alias domains specifically so that static blocklists go stale, and new services appear constantly.

Two honest limitations. First, any list is behind reality by some margin, so a disposable address on a brand-new domain will pass. Second, the category is fuzzy at the edges: forwarding and alias services such as a personal masked-address provider are not disposable in the "expires in ten minutes" sense, and blocking them can mean rejecting a real subscriber who simply values their privacy. We treat the clear-cut temporary providers as disposable and leave forwarding services as valid. More on the disposable landscape.

Stage 4: role-account matching

RFC 2142 defines a set of mailbox names that belong to a function rather than a person: postmaster, abuse, hostmaster, webmaster, noc, security, plus business names like info, support, sales and marketing. Most are read by several people, or by a ticketing system, or by nobody.

These addresses are usually perfectly deliverable. The flag is not about deliverability, it is about consequence: nobody in particular consented, so complaint rates on role addresses tend to run higher than on personal addresses, and complaints are the metric that does the most damage. When role addresses are fine to keep is a separate question with a genuinely nuanced answer.

Stage 5: the SMTP RCPT probe

This is the stage that distinguishes a verifier from a regular-expression check. The verifier connects to the mail server named in the MX records, speaks SMTP as far as the RCPT TO command, reads the server's response code, and then issues QUIT without ever sending DATA. No message body is transmitted and nothing arrives in the recipient's mailbox.

A clean 250 at RCPT means the server is willing to accept mail for that address. A 550 5.1.1 typically means the mailbox does not exist. A 4xx means try later, not that the address is bad.

The technique has four well-known complications:

A verifier that silently resolves these ambiguities into "valid" or "invalid" is making the result look cleaner than the evidence supports. The mechanics of the probe are covered in detail here.

Check an address before you send to it

Mailcheq runs syntax, MX, disposable, role-account and live SMTP checks in the browser, free.

Verify an address →

Catch-all detection, and why it deserves its own answer

The standard way to detect a catch-all is to probe a random address that is almost certainly not in use — something like x7k2q9fp3v@thedomain.com. If the server accepts that too, it accepts everything, and the earlier 250 for the real address carries no information.

This matters commercially: catch-all domains are common in B2B, where a company configures its mail gateway to accept anything and filter later. If your verifier reports those as plain "valid" your list looks cleaner than it is, and you discover the truth at send time. We report them as their own category so you can decide. The full catch-all explanation is here.

The latency budget

A real-time check has to fit inside a signup form, which means a few hundred milliseconds, not several seconds. Roughly where the time goes:

StageTypical costWhy
SyntaxUnder a millisecondLocal parsing only.
MX lookupTens of milliseconds, often zeroCached per domain; DNS TTLs make this cheap at scale.
Disposable and role matchingNegligibleIn-memory lookups.
SMTP probeHundreds of millisecondsTCP connect, optional TLS, several command round trips to a server that may be anywhere in the world.

The SMTP stage dominates, and it is the one stage that cannot be cached per address without creating stale answers. Domain-level facts — MX records, whether the domain is catch-all, whether it is disposable — cache well, which is why the second address you check on a known domain comes back faster than the first.

Reading a result honestly

A sensible verifier reports three states rather than two:

What verification never tells you is whether the person agreed to hear from you. An address can be perfectly deliverable and still be one you have no lawful basis to mail; it can also be a spam trap, which by design accepts mail and looks completely healthy. Verification protects your bounce rate. Confirmed opt-in protects your complaint rate. They solve different problems and you need both.

Spotted an error on this page? Email hello@mailcheq.com and we will fix it.