Skip to content

Blog Website 8 min read

Website down? The error it shows decides what Google keeps

Website down? The error it shows decides what Google keeps

Author

Is a website down for a day enough to lose its place on Google? Last week the website of VLC, the free video player, answered Google with a joke error code, 418 "I'm a teapot", which Google treats like most other 4xx errors: "basically not found". Google kept the site in its results anyway, and Gary Illyes, who works on Google Search, explained why in a comment the trade site Search Engine Roundtable reported on 5 November: VLC is "arguably one of the most important pages on the internet", so Search "temporarily protected the URL from falling out of the index". Even then, the result lost its description and its little logo: "no snippet, no favicon, no nothing", Illyes wrote, only the link and the last good title. Don't count on that courtesy for a small business site. What happens to your pages depends on how long the outage lasts and on what your site says to Google while it's broken, and the second part is the one you can control.

TL;DR: When a website is down, Google reacts to the error code it receives. "Temporarily unavailable" (a 5xx code such as 503) gets a few days of patience. "Not found" or "forbidden" (404, 403, 410) starts removing pages. An error page that answers "OK" can replace your page. Check which one your site sends.

Website down: what Google does with each error code

Google reads the three-digit status code your server sends with every page, and its documentation gives each code a different outcome. These codes are invisible to visitors, who only see the error screen, so two outages that look identical in a browser can end very differently on Google:

  • 5xx codes, such as 503, or no answer at all. Google slows its visits and keeps your pages for now. Its documentation says pages already in its index, the list of pages Google can show in results, are "preserved in the index, but eventually dropped" if the errors persist. John Mueller, who answers site owners' questions for Google, put the grace period at "the next day or so" of retries, with "no drop in indexing or ranking until a few days have passed", in a 2021 answer on Reddit.
  • 4xx codes (403, 404, 410). Google takes these as "the content doesn't exist", and "URLs that are already indexed and return a 4xx status code are removed from the index". Mueller said a site owner would generally see a visible drop in indexed pages over the first week or so.
  • A 200 "OK" code on an error page. Google treats the page as normal content. If the content looks like an error, Search Console reports a "soft 404". Your services page can end up replaced, in Google's eyes, by a page saying something went wrong.

Why a "not found" costs more than a crash

A "not found" answer is an instruction, and Google follows it. A crashed server says "try later", which Google is built to tolerate. A page that answers 404 or 403 says the page is gone or closed to visitors, and Google acts on that statement address by address.

The VLC case shows the limit of Google's patience. The protection Illyes described is reserved for pages Google considers important, and VLC's homepage still lost everything except its link and title while the error lasted. A plumber's or a clinic's services page has none of VLC's standing, so a week of answering 404 would most likely start taking it out of results. Google's guide to pausing a website is blunt about it: "Don't block the website by returning 403, 404, 410 HTTP status codes, or with a noindex robots meta tag". Each of those, it says, "will remove the website's URLs from Google Search".

Recovery runs page by page as well. Mueller said pages that return within days to weeks are "usually back exactly how they were", with the important ones first and the rest following as Google recrawls them, which for most addresses happens "somewhere in the range of hours to months".

The outages owners cause without noticing

Most of the outages that hurt a small business site never look like outages. The site loads for the owner, so nobody checks what it tells a crawler. These are the cases to rule out, starting with one that catches European businesses because of where Google crawls from:

  • A security setting that blocks visitors from outside Europe. Google says it "generally crawls from the US", and "If you block the US, Google Search won't be able to access your site at all". A country block or bot filter added after a spam attack can answer Google with 403, which reads as "forbidden" and starts removal.
  • A maintenance or "coming soon" page left on during a redesign. Whatever code it sends is applied to every page it covers. If it says 200, Google may index the placeholder; if it says 404, Google drops the real pages.
  • A suspended hosting account or an expired domain. The notice from your host, or from the company your domain is registered with, replaces your site, with their status code rather than yours.
  • A robots.txt file that errors. This small file tells crawlers which parts of the site they may visit, and Google warns not to return 503 for it, because that "blocks all crawling".
  • Search Console's removal tool used to "hide" a site during a closure. Google advises against it, because people then can't find out why the business is unreachable.

How to check what your site tells Google

You can see the status code yourself in under a minute. Search for a free "HTTP status code checker", paste in your address, and it shows the code, such as "200 OK". Developers do the same with the command Google's own guide uses, curl -I https://yourdomain.com/. Run it on your homepage, your main service page and an address that doesn't exist, such as yourdomain.com/test-missing-page, which should answer 404.

If the site has already been down, today's check can't show what it sent then, but Search Console can. In Google's free tool for site owners, Settings > Crawl stats covers the past 90 days: its host status flags any availability problem Google hit, and its crawl responses table shows which codes Google received and roughly when. Its URL Inspection tool then fetches a single page the way Google does. If a page you rely on shows an error there, the next questions go to your developer or host:

  • What code does our maintenance page send?
  • What does the site send if the database or hosting fails?
  • Does any firewall, country block or bot filter stop requests from the US?

Our blog once told Google a live post didn't exist

Many small business sites now pull something from another service: prices, availability, reviews, posts. That handover is where we found this failure on our own site. Our blog posts live in a separate system, and the site refreshes its copy every hour. On 3 September 2026 a new post was live there, but someone opened its address before the site's copy had refreshed. The site answered "not found" and kept serving that answer until the next refresh, so for up to an hour every visitor and crawler was told the post didn't exist, and nothing alerted us.

Two changes came out of it. Publishing a post now refreshes every blog page at once, so there is no gap to fall into. And when the system holding the posts can't be read, or answers with nothing, the site keeps showing the last good copy of each page rather than saving a "not found" over all of them. We would rather show a page that is an hour out of date than tell Google a whole blog is gone. We checked again on 8 November: a missing address on sharphaw.com answers 404 and the homepage answers 200, which is what Google expects from each.

Ask the same of any part of your site that depends on another service: when that service fails, does the page say "try later" or "gone"? It is one of the checks we run on the websites we build and maintain.

How to take a website offline without losing Google

Google's advice is not to take the site offline at all. For a temporary closure, it recommends keeping the site up with limited functionality: switch off the cart or booking form, add a banner explaining the delay, and update opening hours and product availability so people can still read about you.

If you must take the whole site down, Google gives two options by length. For one to two days, show a short notice page that answers 503, ideally with a "Retry-After" note in the code telling crawlers when to return, and keep robots.txt working. For anything longer, publish a simple placeholder homepage that answers 200, so people searching your name still find you and learn what's happening. Google adds that "Completely closing a site even for just a few weeks can have negative consequences" for indexing, and that coming back after a long gap is harder because pages may need to be indexed again.

Frequently asked questions

How long can a website be down before Google removes it?

It depends on the error code. With a 5xx server error, Google retries and nothing changes for a few days, though persistent errors eventually drop pages. A 404 or 403 starts removal as Google revisits each page, and the drop is usually visible within about a week.

Will my Google rankings come back after downtime?

Usually, if the site was down for days rather than months. Mueller said returning pages are "usually back exactly how they were", with important pages recovering first. If traffic is still down a few weeks after everything works again, look for another cause rather than waiting.

Should I take my website offline while it's being redesigned?

No. Keep the current site live until the new one replaces it at the same addresses. A "coming soon" page across the whole site for weeks is the closure Google warns against, and it can send the wrong code to every page you already have on Google.

What to do this week

Run the status check on your homepage, your main service page and a made-up address today; the three codes tell you whether a breakdown would read as "try later" or "gone". Then ask whoever runs your site the three questions above, and get the answers in writing before your next redesign, host move or security change.

If nobody on your side owns those answers today, see how Foundation, our website plan, keeps a site maintained, or compare the terms on our plans page.

The service behind this article: Conversion-First Websites

Read more

The newsletter

New posts, in your inbox.

One email per post, covering websites, ads, content and AI automations for owner-operated businesses in Europe. It goes out when a post is published, which in a quiet month means nothing at all. No drip sequence, no sales cadence, and every email carries a one-click unsubscribe.

We don’t sell or share your details. See the privacy policy (opens in a new tab).

Let’s make your digital work compound.

Choose a time on Gabriel’s calendar. We’ll go through where you are today, what’s slowing growth, and what’s worth doing first, whether or not we work together.

Gabriel Espinheira

Gabriel Espinheira

Senior software engineer · English or Portuguese

A focused 30 minutes, not a sales pitch.