What happens when an email address is verified: the DNS lookup, a real SMTP conversation line by line, and what each mail server reply means.
Key takeaways
Email verification checks whether an address can receive mail without sending anything to it. A verifier confirms the address is well formed, looks up the domain's mail servers in DNS, then opens an SMTP conversation with one of those servers and asks whether it would accept mail for that exact mailbox. The server's reply, a three-digit code, is the evidence. When the reply proves nothing, which happens more often than most tools admit, the honest verdict is "unknown".
Most explanations stop at that list of steps. This one shows the actual conversation our engine has with a mail server, what each reply code means, and why the same question gets a straight answer from Gmail and a useless one from Outlook.com.
A verifier asks its questions in order of cost, and stops at the first one that settles the matter. There is no point opening a network connection for an address that is missing its @.
@, a local part and a domain within the length limits of RFC 5321. Checking this costs nothing. A near miss of a big provider, such as gmial.com, gets a correction here too. (Typos are common: we counted thousands of them in 1.47 million real addresses.)info and support.Questions 1 to 3 are cheap and deterministic. Every serious verifier gets them right. Accuracy is decided almost entirely by question 4.
Here is the conversation for an address at a company domain, with the names changed and trimmed to the lines that matter. Lines starting with > are what our engine sends; the rest is the receiving server.
220 mx.example-corp.com ESMTP ready
> EHLO verify.bounceshift.com
250-mx.example-corp.com Hello verify.bounceshift.com
250 SIZE 52428800
> MAIL FROM:<probe@verify.bounceshift.com>
250 2.1.0 Sender OK
> RCPT TO:<jane.doe@example-corp.com>
250 2.1.5 Recipient OK
> RCPT TO:<x7q2kd9w1v@example-corp.com>
550 5.1.1 User unknown
> QUIT
221 2.0.0 Bye
What each step is for:
EHLO introduces the sender by hostname. A server checks that name against the connecting IP address. If the hostname, its DNS record and the IP's reverse DNS don't all agree, many servers stop answering honestly from this point on.MAIL FROM names a sender address. It sits on a domain whose SPF record authorises the probing IP, for the same reason.RCPT TO is the actual question: "would you accept mail for this person?" 250 means yes.RCPT TO asks about an address that cannot exist, made of random characters. This is the catch-all test; the engine actually asks about a few of them in the same session, and the transcript shows one. The server rejected it with 550 5.1.1, which proves it really checks recipients, so its 250 for Jane is worth something. Had it accepted the random address too, both answers would be meaningless. That domain is a catch-all.QUIT ends the session. No DATA command is ever sent, so no message exists and the person whose address it is sees nothing.The first digit of the reply carries most of the meaning: 2 is yes, 5 is a permanent no, 4 is "not now". The enhanced code after it (like 5.1.1) says what the no is about. This is how our engine maps the replies it sees.
| Server reply | What it means | Verdict |
|---|---|---|
250, and the random address is rejected | The mailbox exists | valid / smtp_verified |
250 for the random address too | The domain accepts everything, so nothing is proven | catch_all, scored from the domain's delivery history |
550 5.1.1 "user unknown" | No such mailbox. A statement about the recipient. | invalid / mailbox_not_found |
550 5.7.x or 554 with "blocked", "spam", "policy" | The server refuses us. A statement about the sender, not the address. | unknown / smtp_blocked |
450, 451, 452 or "greylisted" | Greylisting: try again later | unknown / greylisted, retried automatically |
4.2.2, or a 4xx that says "over quota" or "mailbox full" | The mailbox exists but is full, so mail would bounce today | unknown / mailbox_full |
| No reply in time | A slow or overloaded server | unknown / timeout_exceeded |
The row that catches tools out is the 5.7 one. A verifier that reads any 5xx as "invalid" will throw away real addresses whenever a server dislikes its IP address. A 5.1.1 is about the mailbox; a 5.7.1 is about the sender.
Mail providers protect their users from address harvesting, and each does it differently. The result is that a verifier's accuracy depends heavily on which providers your list uses.
RCPT TO stage, so SMTP alone proves nothing there. For those domains we ask Microsoft's sign-in service whether an account exists for the exact address instead (how that works). The result comes back as m365_mailbox_verified or m365_mailbox_not_found.unknown / microsoft_accept_all.mx_forward.catch_all with a confidence score, never a bare valid.A probe is a single question asked once. Real sending outcomes are better evidence. Customers who connect Mailgun, SendGrid, Postmark or Amazon SES send us the bounces, deliveries and complaints those services report, and we keep them as a record per address. The record holds a one-way hash of the address, never the address itself.
That history decides cases the probe can't. An address that has hard-bounced repeatedly comes back invalid / reputation_confirmed_invalid, even if its server says 250 to everything. An address on a catch-all domain that has received mail cleanly several times recently comes back graded on that history rather than as a coin toss. And history counts for less as it ages: a bounce from last week weighs more than one from last year.
When a server greylists, times out, blocks the check, or accepts every address, the engine has no proof either way. It's tempting to return valid anyway, because it looks better in a report. We return unknown, with a sub-status saying which of those happened, and the credit is refunded. A false valid is the one result that turns into a hard bounce later, and hard bounces are what damage your sender reputation.
An unknown address isn't a bad address. Greylisted ones are re-checked automatically a few times over the following day, and many turn decisive. The rest are worth keeping in a separate segment and mailing carefully. Cleaning a list covers how to handle each status.
Every result carries the same three fields: a status (the verdict), a sub-status (the reason, like greylisted or mailbox_not_found) and a confidence score from 0 to 100. The status and error code reference lists all of them.
The free email verifier runs the first three questions on any address without an account. The API quickstart shows how to run all four from your own signup form, with the full response. And if you're choosing between verifiers, the email verification services comparison sets ten of them side by side on price, free credits, credit expiry and the checks each one runs.
Hussam Abd is the founder and engineer behind BounceShift. He has been building software since 2013 — full-stack engineering, then founding and leading engineering teams, including CTO and founder roles at MSAAQ and product and front-end work on customer-support tooling at Hsoub. He built BounceShift to make email verification that is accurate and honest about what it cannot know.
Related terms
Upload a CSV and get back the addresses that are safe to send. 100 free credits to start, no card, and no charge for the ones we can't verify.
Have a list to clean? 100 free credits to start. No card.
Clean a list free