After a domain migration, a site opens correctly on office Wi-Fi but still shows the old version on mobile data. Changing the record again feels like the obvious fix. It may make the situation harder to understand. DNS answers are cached by different resolvers, and browsers and content delivery networks can retain separate copies of site content. The first task is to find which layer is serving the outdated information, not to keep editing a configuration at random.
The behavior is frustrating because the domain name appears to be one thing: a single address typed into every browser. Behind that address is a distributed system of nameservers and caches. Different devices may temporarily use different information even when nobody has made a mistake.
Before changing a setting a second time, it helps to understand which part of the system provides the answer.
The domain is a name, not the server
A domain such as example.com is a readable identifier. A browser ultimately needs an internet address and the information necessary to connect to the service. The Domain Name System, or DNS, connects domain names with the records that help clients find resources.
ICANN's explanation of DNS describes it as a mechanism that allows people to use recognizable names instead of remembering numerical internet addresses. The process is not a single lookup performed by one universal computer.
As the Cloudflare Learning Center's DNS overview explains, a device may ask a recursive resolver operated by an internet provider, a company network or another DNS service. That resolver can obtain answers from authoritative nameservers and cache them for later use. Another resolver may still hold an older answer.
This explains why a website can appear to change at different times for different people. The internet does not send a synchronized announcement to every device whenever a domain record is edited.
It also means a working website and a working DNS answer are separate pieces of the experience. DNS can successfully point to a server that is broken, and a healthy server can remain unreachable under an incorrect domain configuration.
What the common records actually do
When diagnosing a website, start with the DNS record type rather than a generic label such as “domain settings.”
An A record points a hostname to an IPv4 address. An AAAA record serves a similar purpose for IPv6. A CNAME record aliases one hostname to another name, subject to DNS provider and record-type constraints. An MX record identifies mail exchangers for a domain. A TXT record holds text used for many purposes, including domain verification and email authentication policies.
The Cloudflare guide to DNS record types explains these basic records and the role of time-to-live, or TTL.
A website that loads correctly at one address but not at another may have mismatched records for the apex domain and the www hostname. A change to the A record does not automatically change an unrelated AAAA record or CNAME. Some users may prefer IPv6 where supported, so an outdated AAAA record can cause failures that seem intermittent across networks.
Email problems after a website migration require separate attention. Changing nameservers without recreating necessary MX and TXT records can interfere with email delivery or authentication even if the website begins working.
The most useful debugging question is: which hostname and record type are involved, and what answer is each resolver returning?
TTL is a caching instruction, not a stopwatch
DNS records include a TTL value that tells resolvers how long an answer may be cached. If a resolver cached the old address shortly before a change, it may continue using that address until the cached record expires under its rules.
A lower TTL can make future changes easier to distribute, but changing it only moments before a migration does not erase a high-TTL answer that is already cached elsewhere. Operational behavior can also depend on resolver implementations, negative caching and the chain of records involved.
This is why the popular phrase “DNS propagation” can be misleading. The underlying record is often already updated on its authoritative server. The remaining inconsistency may come from cached old answers or from different nameservers serving inconsistent configurations.
Changing the record repeatedly can restart confusion. Some people may receive the first new value, while others receive a later one, and a third group remains on the original answer.
For a planned migration, reduce TTL in advance when appropriate, confirm that the new destination works and record the change time. During the transition, keep the old service operating safely if possible. A lower TTL is one component of a plan, not a guarantee of instant uniformity.
Find the authoritative answer before blaming caches
A reliable diagnosis separates two questions: what is the authoritative nameserver currently publishing, and what does a particular recursive resolver return?
If the authoritative answer is wrong, waiting for caches will not fix the fundamental configuration. If authoritative records are consistent and correct while a recursive resolver still returns an old address, caching becomes a plausible explanation.
Look up the domain's authoritative nameservers using a trusted DNS inspection tool or a command-line lookup. Check both the parent delegation and the records served by the nameservers. A recent nameserver change is especially important because some resolvers may still be using a previously delegated provider.
Then query more than one resolver and record the answers, record types and approximate TTLs. Comparing results is more useful than asking a friend to refresh their browser twenty times. A browser refresh usually reloads a page; it does not compel every shared DNS resolver to abandon cached data.
If the answers match but the websites differ, investigate the next layer: web-server configuration, caching proxies, CDN routing, TLS certificates or application behavior. DNS cannot explain every difference that appears after a domain migration.
A simple table of hostname, resolver, answer and time will often reveal the pattern more clearly than another control-panel screenshot.
The old site might be in a browser or CDN cache
Even when every resolver points to the new server, users may see stale HTML, images or assets because of browser caching or a content delivery network. Those mechanisms are separate from DNS caching.
A browser can reuse an older CSS file. A CDN may serve a cached page. A reverse proxy might route traffic to several application instances that run different versions. Clearing the browser cache can sometimes alter the result, but it does not diagnose the upstream cause.
Check the page's response headers and content version where possible. If the server response includes an identifiable build version, compare it across networks. For a CDN, review its cache rules and purge procedure. For multi-server deployments, confirm that all nodes received the same release.
TLS certificate errors add another dimension. The domain may resolve to the intended server, but the service presents a certificate that does not cover the requested hostname. That can look like a network failure to an end user.
The troubleshooting sequence should advance one layer at a time: DNS answer, network connection, TLS negotiation, HTTP response and rendered page. Mixing them leads to unnecessary changes.
How to plan a cleaner migration
Before moving a website, inventory the domain's DNS records, including email-related records and verification tokens. Verify that the new hosting account is ready to serve the correct hostname securely. Install and test the certificate when the provider's process allows it.
If the change affects an A or AAAA record, check that the new address serves the expected site. If changing nameservers, recreate the full required zone before changing delegation. Document the old settings and a rollback route.
Schedule the change when someone can monitor it. Lower TTLs sufficiently far in advance, subject to the provider's rules. During the overlap, keep the previous environment in a safe state so old cached answers do not immediately strand visitors.
After the change, inspect authoritative records and test from several networks. Monitor the site's actual health, not only whether DNS resolves. A successful lookup is a prerequisite for reaching the site, not evidence that all pages, email services and APIs are functioning.
For a business that processes payments or customer records, a migration also requires attention to session handling, data consistency and external integrations. Those concerns cannot be solved by DNS alone.
A fault-finding example
Suppose a company changes the A record for shop.example to a new server. A mobile carrier's resolver returns the new IPv4 address. The office resolver still returns the previous one. An independent resolver returns the new A record but also an older AAAA address.
There are at least two separate problems to investigate: stale IPv4 caching in one location and inconsistent IPv6 configuration. The website might work for some users while failing for devices that connect via the wrong IPv6 destination.
The sensible repair is to check authoritative A and AAAA values, confirm the intended endpoints, and let known cached records expire or use appropriate resolver management where available. Randomly replacing the A record again would not address the AAAA inconsistency.
The example is hypothetical, but it illustrates why “works for me” is not enough information to diagnose a distributed system.
What to tell customers during the change
When users report inconsistent access, avoid promising that everyone will be fixed at a precise hour unless the cause and recovery window are known. A concise service notice can explain that a domain update is being checked and that some networks may temporarily reach an older destination.
Give users a clear contact route that does not depend entirely on the affected domain. If transactions are involved, tell them whether an order submitted to the old environment will still be processed. That business question matters more than the terminology behind DNS.
Internally, save timestamps, authoritative answers and the versions running on each server. These records make a later incident review more useful and reduce the risk of repeated guesswork.
DNS is reliable enough that most people never notice it. When a domain migration exposes its distributed nature, careful measurements work better than frantic configuration changes.
References and background
Our explainers link to source documents so readers can explore the evidence and interpret it in context. Websites and policies may change after publication.
- ICANN, *The Domain Name System*www.icann.org
- Cloudflare Learning Center, *What Is DNS?*www.cloudflare.com
- Cloudflare Learning Center, *DNS Records*www.cloudflare.com