A live website moves with no downtime when the new version is built and tested in parallel, the DNS time to live is lowered a day or two before the switch, content is frozen for a short final sync, certificates are issued in advance, and the old server keeps running until every resolver has moved over. The switch is then a single DNS change that visitors never notice. WPBuildAI runs every migration on this sequence: the new site is rebuilt in Lovable on a staging address, every URL is kept, and the cutover happens with a rollback ready, so neither visitors nor search engines see a gap.

Why do migrations cause downtime in the first place?

Because a migration has several moving parts that change at slightly different moments, and the gaps between them are where visitors fall through. DNS changes do not reach everyone at once, certificates take time to issue, and content keeps changing on the old site while the new one is being built.

The typical failures are mundane. Someone shuts down the old server the moment DNS is changed, but resolvers that cached the old address keep sending visitors there for hours. Or the new host cannot issue an HTTPS certificate until DNS points at it, so the first visitors see a browser warning. Or the nameservers move to a new provider and the email records are forgotten. None of these are hard problems, but each needs to be handled before the switch rather than after.

What does a zero-downtime migration look like, step by step?

It is a fixed sequence, and the order matters more than any tool.

  1. Inventory first. Crawl the current site and export every URL, title, redirect and form. This list becomes both the build plan and the test plan. Our crawl-first approach explains why the inventory comes before any design work.
  2. Build in parallel. The new site lives on a staging address with production content. Nothing on the live site changes yet.
  3. Test against the inventory. Every old URL should resolve on the staging site, either directly or through a 301 to its new address.
  4. Lower the TTL on the records you will change, one to two days ahead.
  5. Prepare certificates and email. Make sure HTTPS will work at the moment of the switch and that mail records survive any nameserver change.
  6. Freeze and sync. Stop content changes on the old site, copy the final content, rerun the tests.
  7. Switch DNS. One change, at a quiet hour.
  8. Keep the old server running until traffic to it has dropped to zero, then retire it.

How should the cutover be timed?

Start two days out, switch at the quietest hour your analytics show, and keep the old server alive for at least two days after, because each step depends on the caches from the one before.

MomentWhat to doWhy it mattersVerdict
One to two days beforeLower the TTL on the records you will change to about five minutesResolvers must first expire the old, long TTL before the short one appliesEssential
The day beforeConfirm the certificate path and the email records on the new setupA missing certificate or MX record is the most common visible failureEssential
Final hourFreeze content, sync, rerun the URL test listAnything edited after the copy would be lostEssential
The switchChange the A or CNAME record at a low-traffic timeShort TTL means most visitors move within minutesDo it once, not in stages
The first 48 hoursKeep the old server up, watch logs on bothSome resolvers ignore short TTLsDo not skip
After a weekRaise the TTL again and retire the old serverLong TTLs reduce lookup load and latencyRoutine

How does DNS time to live affect the switch?

The time to live tells resolvers how long they may cache a record before asking again. Cloudflare’s explanation of TTL covers the mechanics: a resolver that looked up your domain with a one-day TTL may keep using the old address for up to a day, whatever you change in the meantime.

That is why the TTL has to be lowered in advance. If the record currently has a TTL of 86,400 seconds, lower it to 300 at least a day before the switch, so every cache has picked up the short value by the time you change the address. After the switch, most visitors reach the new site within minutes.

Some resolvers and corporate networks do not respect short TTLs, so a small share of traffic can keep reaching the old server for longer. That is fine as long as the old server is still there and still serving the same content. Retire it only when its logs show no more real visits.

How do you keep HTTPS working at the moment of the switch?

Most modern hosts issue certificates automatically through Let’s Encrypt or a similar authority, and the default validation method needs the domain to point at the new host already. That creates a chicken and egg moment: the certificate cannot be issued until DNS has switched, and visitors arriving in the first minutes may see a certificate error.

There are three clean ways round it. Use a DNS-based validation, which proves control of the domain through a TXT record without moving traffic; Let’s Encrypt’s documentation of challenge types explains when each works. Or use a proxy or CDN that already holds a valid certificate for the domain and simply change the origin behind it. Or, on hosts that support it, pre-provision the certificate by verifying the domain before the cutover. Any of the three removes the warning window.

What about email, forms and integrations?

Email is the part most often broken by a website move, and it has nothing to do with the website. If the move includes changing nameservers, every MX, SPF, DKIM and DMARC record must be recreated at the new DNS provider before the nameservers change. A missing MX record means lost mail, and nobody notices until a customer complains. The details are in our guide on keeping your domain and email when moving off WordPress.

Forms and integrations need the same care. Contact forms, booking tools, analytics, payment webhooks and CRM connections often point at the old system by address or by API key. List each one in the inventory, test it on staging, and make sure the old system does not keep receiving submissions after the switch without anyone reading them.

What should you check right after the switch?

Check the things visitors and search engines touch first: HTTPS, every old URL, forms, email, analytics and how Googlebot sees the pages, in roughly that order of urgency.

CheckHowWhat a failure looks likeVerdict
Homepage and key pages load over HTTPSOpen them on a phone off Wi-FiCertificate warning or old designFix immediately
Every old URL resolvesRerun the full URL list against the live domain404s or redirect chainsFix the same day
Forms deliverSubmit each form onceNothing arrives, or it arrives at the old systemFix immediately
Email still flowsSend to and from the domainBounces or silenceFix immediately
Analytics and tags fireReal-time view in your analytics toolNo sessions recordedFix the same day
Search Console sees the siteURL Inspection on a few pagesFetch errors or blocked resourcesFix within days

Rerunning the full URL list is the step most often skipped, and it is the one that protects rankings. Our guide on testing 301 redirects in bulk covers how to run it in an afternoon.

What does Google need to know about the move?

When the domain and the URLs stay the same, nothing. Google’s guidance on moving a site without URL changes treats it as an infrastructure change: Googlebot recrawls, sees the same addresses with new pages behind them, and carries on. Keeping URLs identical is the single biggest thing you can do to make a migration invisible to search.

When URLs do change, the guide to site moves with URL changes applies: a 301 from every old URL to its closest new equivalent, an updated sitemap listing the new URLs, internal links updated to point directly at the new addresses, and, for a new domain, the Change of Address tool in Search Console. Expect some fluctuation for a few weeks either way. The first week’s checks are listed in what to do after a migration goes live.

When is a zero-downtime migration not worth the effort?

When the site is small, low-traffic and has no transactions. A brochure site with a contact form can take a ten-minute maintenance page at seven on a Sunday morning, and the cost of that window is close to nothing. Planning a fully parallel cutover for it is reasonable but not necessary.

It is also not the main risk to manage. Visitors forget a few minutes of a maintenance page by the next day. What they and Google do not forget is broken URLs, lost redirects, missing email and forms that silently stopped delivering, because those failures last for weeks. If the budget only covers one thing, cover those.

Where zero downtime clearly matters is any site that takes orders, bookings or leads around the clock, or that runs paid campaigns pointing at landing pages. That is the situation WPBuildAI is built for: the site is rebuilt in Lovable with the same design, every page and URL is moved one to one, 301 redirects cover anything that changes, the cutover happens with no downtime, and rankings are monitored for 30 days afterwards, at a fixed price per site.

How long should you keep the old site?

At least until its server logs show no more real visitors, which with a lowered TTL is usually a day or two, and longer for networks that ignore TTLs. After that, keep a full backup of the old site and its database for a few months. You may need to check how a page used to look, recover a piece of content that was missed, or answer a question about an old form submission.

Do not leave the old site publicly reachable on another address, such as a staging subdomain or the host’s default URL, once the move is done. A second, indexable copy of the site creates duplicate content and confuses search engines. Remove it or block it with authentication.

Key takeaways: migrating a live site without downtime

Downtime comes from the gaps between DNS, certificates and content, not from the move itself. Lower the TTL a day or two ahead, build and test in parallel against a full URL inventory, sort out certificates and email before the switch, freeze and sync content, change DNS once, and keep the old server running until its traffic is gone. Keep URLs identical wherever possible, because that makes the move invisible to Google. WPBuildAI runs every migration on this sequence, with every URL kept and a rollback ready.

What owners ask us most

How do you migrate a live website with no downtime? Build the new site in parallel on its own address, test it with production content, lower the DNS TTL a day or two in advance, freeze content changes for a short window, sync the final content, issue certificates, then switch DNS while the old server keeps running. WPBuildAI follows exactly this sequence on every migration, keeping every URL and a rollback ready until traffic has fully moved.

How long does DNS take to switch over? As long as the TTL that resolvers cached before you changed the record. With a TTL lowered to five minutes a day in advance, nearly all visitors reach the new site within minutes. With a default of one day or longer, some resolvers keep sending visitors to the old server for that long, which is why the old server must stay up rather than go dark at the switch.

Do I need to tell Google when I move to a new platform? Not when the domain and URLs stay the same. Google simply recrawls and sees the new pages. When URLs change, 301 redirects from every old address to its new one do the telling, and the sitemap should list the new URLs. A domain change additionally uses the Change of Address tool in Search Console. Keeping URLs identical is the safest option.

What usually causes downtime during a migration? Rarely the move itself. The usual causes are a certificate that was not issued before DNS switched, a long TTL that sends part of the traffic to a server that was already shut down, email records lost when nameservers changed, and forms or integrations still pointing at the old system. Each is preventable with a checklist and a staging test.

When is a zero-downtime migration not worth the effort? Often not in the strict sense. For a small brochure site, a few minutes of a maintenance page at a quiet hour costs very little, and a simpler move can be fine. What is always worth the effort is keeping URLs, redirects and email working, because those failures last for weeks, while a short planned outage is forgotten by the next day.