How a Website Move Takes Down the Email, and How to Stop It

Your Website and Your Email Share One Set of Records

DNS decides where every kind of request for your domain goes. Web traffic follows one type of record and mail follows another, so a change made for the website can move the mail with it if the mail records are not deliberately preserved.

Most owners think of the website and the email as two services from two companies that happen to share a name. Technically they share something more important than a name.

Your domain has a single set of DNS records, and different record types point different kinds of traffic at different places. One set sends browsers to your website. A different type, the MX record, tells the world where your mail lives. Google describes it plainly: when someone sends you an email, the sender’s computer looks up the MX records for your email domain to figure out where to deliver it (Google, 2026).

The consequence is the part that catches people. Changing where your website is hosted often means changing which service answers DNS questions for your domain, and that service answers every question, not only the ones about web pages. Move the answering service without bringing the mail records along and the mail records simply cease to exist, as far as the internet is concerned.

This is why a migration can leave a website that loads perfectly while quotes stop arriving. Nothing is broken on the site. The letterbox has been removed from the front of the building.

The Automatic Import Is Helpful and Is Not a Guarantee

Providers offer to scan your existing records and bring them across, which usually works. Cloudflare states plainly that its quick scan is not guaranteed to find all existing records, so a manual review is required rather than optional (Cloudflare, 2026).

Modern providers make this easy, which is exactly why it goes wrong.

When you point a domain at a new DNS provider, it will typically offer to scan your current records and recreate them. It is a good feature and most of the time it does the job. But Cloudflare documents the limit in its own words: the quick scan is not guaranteed to find all existing records, which is why a manual review is required (Cloudflare, 2026).

Records get missed for ordinary reasons. Some are not discoverable by scanning. Some were added years ago by a provider nobody remembers. Verification records for mail authentication, a subdomain used by a booking tool, an old record for a phone system: none of these announce themselves, and none of them are visible on the website.

Cloudflare also notes what happens when a domain is activated without its correct records: visitors may experience DNS_PROBE_FINISHED_NXDOMAIN errors (Cloudflare, 2026). The important word is may, because partial breakage is the common case and it is much harder to spot than a site that is plainly down.

The practical rule is to treat the automatic import as a first draft. Export or photograph the full record list from the old provider before anything changes, then compare line by line afterwards.

MX Is the Record That Costs You Customers

Google states that email might not work correctly if you keep old or incorrect MX records, and instructs that other MX records be removed when setting new ones. It also documents that it can take up to 72 hours for new MX records to be recognized (Google, 2026).

Among all the records, MX deserves separate attention, for two reasons.

The first is that mail failures are silent to the person they affect. A customer who emails you and gets nothing back does not phone to report a technical fault. They assume you are busy, or rude, and they call somebody else. You find out days later, if at all, and you cannot recover the enquiries that bounced.

The second is the clock. Google documents that it can take up to 72 hours for new MX records to be recognized (Google, 2026). That is the number that should govern how this work is planned, because it means an MX mistake is not something you fix in ten minutes on Saturday night. You can correct the record immediately and still be waiting into next week for the correction to be seen everywhere.

Google is also explicit that leftovers cause problems: email might not work correctly if you keep old or incorrect MX records, and its setup guidance instructs that any other MX records be removed (Google, 2026). Two competing sets of mail instructions is not a redundancy. It is an ambiguity, and mail delivery resolves ambiguity in ways nobody enjoys.

The Setting That Takes the Whole Domain Down

Cloudflare documents that DNSSEC must be turned off before changing nameservers, or the domain becomes unreachable (Cloudflare, 2026). This one does not degrade gracefully. It fails completely, and it fails for the website and the email together.

There is one item on this list that does not produce a partial failure, and it is the least known.

DNSSEC is a security feature that signs DNS answers so they cannot be tampered with in transit. It is a good thing to have. It also means that if the signing arrangement and the answering service disagree, the correct behavior is to refuse to answer at all.

Cloudflare documents the requirement directly: turn DNSSEC off before changing nameservers, or the domain becomes unreachable (Cloudflare, 2026). Not slow, not partially broken. Unreachable, for the website and for the mail, until it is resolved.

The reason it catches people is that DNSSEC is often enabled once, years earlier, by somebody who is no longer involved, and it sits there working silently. Nobody thinks to check a setting they did not know was on. It belongs on the pre-flight list precisely because it is invisible until the moment it is catastrophic.

The Order of Operations That Prevents All of It

Inventory every record before touching anything, lower the DNS TTL a week ahead, confirm mail routing separately from the website, and switch only when both have been verified against the old list. None of this is difficult. All of it is skipped under time pressure.

Everything above is preventable, and the prevention is sequencing rather than skill.

Take a full inventory first. Export the complete record list from the current DNS provider, or photograph every page of it, before anything is changed. This is the artefact you check against afterwards, and without it you are comparing the new configuration to your memory of the old one.

Lower the TTL in advance. Google recommends lowering the time to live to a conservative low value such as a few hours, at least a week before a hosting move (Google, 2026). A low TTL means a mistake propagates out of the system quickly instead of being cached for a day.

Check DNSSEC before the nameserver change, not after (Cloudflare, 2026). Turn it off, complete the move, and re-establish it afterwards if you want it.

Verify mail as its own test. Do not assume that a working website means working email, because they are separate records with separate failure modes. Send a message from an outside address and confirm it arrives, and do it before anyone announces the migration is finished.

And schedule with the 72-hour figure in mind (Google, 2026). Migrations belong at the start of a working week, when a mail problem can be found and fixed while the recovery window still lands inside business days, rather than on a Friday afternoon when the same mistake costs a weekend and a Monday.

If you would rather have the current picture measured than assumed, our site migration service handles this end to end, and our free audit reports what is publicly reachable and readable on your domain and costs nothing.