Security basics
"This site can't be reached" on your own domain: read the code
Your own site says "This site can't be reached". The code under the headline tells you whether the name, the host or the network failed, and what to fix.

In short
- Chrome prints "This site can't be reached" for more than ten different network errors, and the code at the bottom says which one. DNS_PROBE_FINISHED_NXDOMAIN means your domain name gave Chrome no address at all.
- Flushing your DNS cache, switching to 8.8.8.8 and restarting the router are fixes for a visitor. As the owner they change only your own computer, so look your domain up from outside instead.
- The servers that look names up for your visitors remember a "no such name" answer as well as a real one. Somebody who tried while your record was missing can keep seeing the error after you have fixed it.
A customer sends you a screenshot. Where your app should be, Chrome shows a
grey page headed "This site can't be reached", and at the bottom a line of
capitals you have never seen: DNS_PROBE_FINISHED_NXDOMAIN. The app loaded for
you this morning. It may still load for you now.
The guides that rank for that code are mostly written for the person looking at the error: flush your DNS cache, switch to Google's 8.8.8.8, restart the router. When the site is yours, every one of those changes your own computer and leaves your visitors exactly where they were. Flushing your cache can even make the site look fixed to you while customers still see the error, for a reason that comes further down.
It helps to picture what Chrome does when somebody types your address. It sends
a courier. The courier first looks your name up in the directory, which is
DNS: the internet's address book, turning a name like yourapp.com into the
numeric address of the server holding your app. Then it travels to that
address, knocks, and waits for the page. "This site can't be reached" means the
courier came back with nothing. The code is its note about where it gave up.
What does "This site can't be reached" mean?
It means Chrome got no page from your address, and on its own it says nothing about why.
In Chrome's source that one headline is attached to more than ten different network errors, from a name with no address behind it to a server that cut the connection halfway. The headline is the courier saying it failed. The line under it, and the code at the very bottom, say where.
That is also why a different headline means a different problem. A site stuck
redirecting to itself gets "This page isn't working" instead, with
ERR_TOO_MANY_REDIRECTS, because the courier did reach your server and was sent
round in a circle. Nothing below applies to that one.
Which code is on your screen?
Read the code at the bottom of the page first. It is the one part that is the same in every language, and it says how far the courier got.
| The code | The line Chrome prints above it | How far it got | Where to look |
|---|---|---|---|
DNS_PROBE_FINISHED_NXDOMAIN | "Check if there is a typo in yourapp.com." | The directory had no address for your name | Your DNS records and your registrar |
ERR_NAME_NOT_RESOLVED | "yourapp.com's server IP address could not be found." | The same, without Chrome's own test lookup to confirm it | Your DNS records and your registrar |
ERR_CONNECTION_REFUSED | "yourapp.com refused to connect." | It found an address, and whatever answered turned it away | The address your records point to |
ERR_CONNECTION_TIMED_OUT | "yourapp.com took too long to respond." | It found an address, and nobody answered in time | The address your records point to |
The first two rows are about your name and the last two are about the place it leads. Each half has its own section below.
What does DNS_PROBE_FINISHED_NXDOMAIN mean on my own domain?
It means your domain name gave Chrome no address, and Chrome made sure the visitor's own connection was not to blame before saying so.
When a lookup fails, Chrome runs a small test lookup of its own, which is the "probe" in the name. If the test works, it swaps the code for this one. The comment beside it in Chrome's source puts the conclusion plainly: "The DNS servers are working fine, so the domain must not exist." NXDOMAIN is DNS shorthand for "no such domain".
On a domain you connected to Lovable, Bolt, Vercel or Netlify yourself, that leaves the records, the nameservers or the registration.
The records are not there yet. Connecting a custom domain means adding
records at whoever runs your DNS. Lovable's setup asks for an A record
pointing at the address it shows you and a TXT record named _lovable that
proves the domain is yours. Its documentation says changes "may take up to 72
hours to propagate, though most updates are live within a few hours."
The records are somewhere nobody reads. A domain's nameservers say which provider holds its records. Move them, to Cloudflare or away from the registrar, and every record has to be recreated at the new provider. For a domain bought through Lovable and moved to custom nameservers, its documentation says "any connected project and its email stay offline until you recreate those records there."
The address people type is not the one you connected. yourapp.com and
www.yourapp.com are two names, and each needs a record of its own. Lovable
connects the www address by default and redirects it to the bare domain; turn
that option off and only the bare domain is connected, so a visitor who types
www reaches a name your directory may have nothing for.
The registration lapsed. For .com and the other generic domains,
ICANN's Expired Registration Recovery Policy
requires the registrar to interrupt your DNS for at least the last eight days
the name can still be renewed, or to send visitors to a page saying the
registration has expired. Your app stops loading well before the name is gone,
and the clock you are on from that moment
has three stages.
Why does my site work for me and not for my visitors?
Because your computer and theirs ask at different moments, and the servers answering both of them remember the last answer they gave.
Your visitors do not read the directory directly. Their phone or their internet provider asks a resolver, a server that looks names up and keeps each answer for a set time so it does not have to ask again. The standard for this, RFC 2308, applies the same rule to a "no such name" answer: it is remembered too, for a time taken from your domain's own DNS settings, and the standard suggests one to three hours as a sensible default.
So the courier's note can outlive the problem. Somebody who tried your address while the record was missing can keep being told it does not exist after you added it. It works the other way as well: your own laptop can hold on to the old, working address after the record broke, and show you a healthy site that nobody else can reach.
This is why flushing your cache is a trap when the site is yours. You clear your own memory, look again, the site appears, and the resolvers your customers use are still holding the old "no".
There is one more way for the same domain to give two answers. Some DNS
providers add an AAAA record, an IPv6 address, to a new domain by themselves,
and Lovable's documentation warns: "Visitors whose devices prefer IPv6 can reach
that address instead of your Lovable app." Whether the site works then depends
on the visitor's network as well as on your records.
What do ERR_CONNECTION_REFUSED and ERR_CONNECTION_TIMED_OUT mean?
Both mean the name worked. The directory gave the courier an address, and the trip ended at that address.
ERR_CONNECTION_REFUSED is something at that address answering and turning the
connection away. ERR_CONNECTION_TIMED_OUT is nothing answering within the time
Chrome waited. In Chrome's list of network errors those two read, word for word,
"A connection attempt was refused" and "A connection attempt timed out."
If the address in the bar is localhost, the courier was sent to the visitor's
own computer, which is a setting in your app rather than your domain, and
confirmation emails that send people to localhost
covers it.
For an app on a custom domain, the first thing to check is that the address is
the one your host gave you. Lovable's documentation says "An extra A or AAAA
record at your DNS provider can send visitors to the wrong site, even when
Lovable shows the domain as Active." Its advice is to keep only the records it
gave you at that name and leave unrelated ones, such as the MX records your
email depends on, where they are.
If the records match and every visitor gets the same code, the problem is at the host, and its status page or its support is where to look. If instead your page loads and only the data in it fails, you are past this error altogether, and telling a Supabase outage from a fault in your app starts there.
How do I check what everyone else sees?
Ask a lookup service that is not your computer, then compare its answer with what your host asked you to set.
- Look the name up from outside. Google runs a public lookup page at
dns.google. Type your domain into it. Its raw view
reports
"Status": 3for a name that does not exist, which is NXDOMAIN, and lists the address underAnswerfor one that does. Try bothyourapp.comandwww.yourapp.com. - Compare the address with your host's. If the lookup returns an address,
it should be the one in your host's setup window. Any other
AorAAAArecord at the same name is a candidate for removal. - Read your DNS provider's records against your host's list. Every record the host asked for should be there, under the exact name it gave.
- Open your registrar. Read the expiry date, the card behind auto-renew and which nameservers the domain points to. If those nameservers are not the provider whose records you just read, you have been editing a copy nobody uses.
- After a change, give the remembered answers time. Judge the fix by the outside lookup, which asks fresh, and expect some visitors to catch up later.
Once your site answers again, our free scan reads your domain's registration date and your certificate's end date from outside, so you know when the next lapse would land. It takes about 20 seconds and needs no account: scan your app.
Hearing about it before a visitor does
Something outside your own computer has to ask for your site on a schedule, or a customer will be the one who tells you.
Reeve Monitor asks for your site every 60 seconds, from outside, and tells you which of these failed.
- An outage counts once three checks in a row have failed, 15 seconds apart, and a two-minute grace period lets a blip clear on its own. You hear about it about three minutes after the first failure, by email, Telegram, Slack or webhook.
- The alert says what happened in words: your site's address couldn't be found, the connection was refused, or it didn't respond in time.
- If the hourly check has already seen your domain registration running out, the alert says your domain name may have lapsed.
- It warns you before the registration or the certificate expires, and tells you when the site is back and how long it was down.
Reeve Care does all of that for an app on Supabase, and keeps a verified copy of your Supabase database.
- Copies run on a schedule, encrypted, on Reeve's own storage in the EU, outside your Supabase account, and each one is verified before it counts.
What to do right now
What to do
- Read the code at the bottom of the error page before you change anything.
DNS_PROBE_FINISHED_NXDOMAINandERR_NAME_NOT_RESOLVEDpoint at your name;ERR_CONNECTION_REFUSEDandERR_CONNECTION_TIMED_OUTpoint at the address it leads to. - Look your domain up at dns.google instead of trusting your own browser, for both the bare name and the
wwwone. - Compare the records at your DNS provider with the list your host gave you, and remove any extra
AorAAAArecord at the same name. - If you moved your nameservers, recreate every record at the new provider, the email ones included.
- Check the expiry date and the card behind auto-renew at your registrar.
- After a fix, judge it by the outside lookup and give remembered answers time to expire.
If the code pointed at the registration, read what happens after a domain expires today: only the first stage renews at the ordinary price.
FAQ
What does "This site can't be reached" mean?
Chrome got no page from the address at all. It uses that one headline for more than ten different network errors, from a name with no address behind it to a server that refused the connection, so the headline on its own tells you nothing about the cause. The code at the bottom of the page does, and it is the same in every language.
What does DNS_PROBE_FINISHED_NXDOMAIN mean on my own website?
Your domain name gave Chrome no address, and Chrome ran a test lookup of its own to make sure the visitor's connection was not the problem before saying so. On a domain you connected yourself it comes down to the records at your DNS provider, the nameservers that say which provider holds them, or a registration that has lapsed.
Why does my site work for me but not for other people?
Because the servers that look up names remember the last answer they got, the working one and the failing one alike. Your computer can still hold an old working address while your visitors get nothing, or the reverse. A record some DNS providers add by themselves, the IPv6 one, can also send part of your visitors to a different address from the rest.
How long does a new custom domain take to start working?
Lovable's documentation says DNS changes may take up to 72 hours to propagate, and that most are live within a few hours. A visitor who looked your name up before the record existed can be told it does not exist for a while after that, for as long as your domain's DNS settings let a "no" be remembered.
Can an expired domain cause this error?
Yes. For .com and the other generic domains, ICANN's policy requires the registrar to interrupt your DNS for at least the last eight days the name can still be renewed, or to send visitors to a page saying the registration has expired. Either way your app stops loading on that name before anyone else can take it.