Surfaces our work more for you across Google Search and AI answers.
Your host had a rough night. For a few hours, anyone who typed your address got a tidy page that said "We'll be back shortly." Your uptime monitor stayed green the whole time, so nobody got a text. By morning the site was back, and the only thing that changed is that Google's copy of your homepage now says "We'll be back shortly."
Quick read: Every response a web server sends begins with a three-digit Hypertext Transfer Protocol (HTTP) status code. A 200 means "success, here is the page"; a 503 means "temporarily unavailable, try again later." Google's documentation says content returned with a 2xx code "may be considered for indexing," while 5xx and 429 codes "prompt Google's crawlers to temporarily slow down with crawling" and already indexed URLs "are preserved in the index, but eventually dropped" if the errors persist (Google Search Central). On November 17, 2026, Search Engine Roundtable reported a LinkedIn reply from Google's John Mueller to a question about serving a "back soon" page with a 200 code during an outage lasting a few hours. Paraphrasing his two points: the temporary text can be indexed in place of the real page, which corrects itself once the content returns; and if other sites show the same text, Google can treat the pages as duplicates of one another, which takes longer to undo because duplicates are crawled less often. Google's advice for planned downtime, unchanged since a 2011 Search Central post, is to return a 503 with a Retry-After header, and not to keep it up for longer than one to two days.
The three digits a visitor never sees
When a browser asks your server for a page, the server answers with a number before it sends anything else. Browsers don't show it. They render whatever text comes after, which is why a "back soon" page looks the same to a human whether the number in front of it was 200 or 503.
Crawlers read the number first. The short version of what each one tells Google:
- 200 (OK). Here is the page. Google "considers the content for processing," which for Search means indexing. Whatever text arrived is now a candidate to replace what Google had on file for that address.
- 503 (Service Unavailable). Not right now, come back later. Google slows its crawling and holds on to what it already indexed. You can add a
Retry-Afterheader saying when to come back, in seconds or as a date. - 500 (Internal Server Error). Something broke. Treated like a 503, minus the courtesy of a timing hint.
- 404 and 410. Gone. Keep returning these for a page that exists and Google will take you at your word and remove it.
In October, Google's Search Central account reminded site owners taking servers down for a PHP patch to serve a 503, and a reader on Bluesky ranked the options from best to worst as 503, 500, 403, 404, then 200. Mueller called that ordering "a good model," per Search Engine Journal. The polished "we'll be right back" page with a success code comes dead last. Below a plain 404.
Why a success code is the worst way to be down
The question that reached Mueller in November was a fair one: does a 200 "back soon" page matter if the outage only lasts a few hours? His answer, as Search Engine Roundtable reported it, came in two parts, and neither one depended on the length of the outage.
The placeholder gets indexed. Google asked for your homepage, your server said "success," and handed over three sentences about checking back later. That is now the most recent version Google has of the page. Search for your business by name and the snippet under your listing reads like a closed sign. This part fixes itself. Google recrawls, finds the real page, and swaps it back. It fixes itself on Google's schedule, though, and nothing about that schedule is yours to set.
The placeholder is a duplicate. Your host's outage page is usually the same outage page on every other site that host runs. Same words, same layout. Google's canonicalization system exists to collapse identical pages into one, and its documentation is blunt about the consequence: "the canonical page will be crawled most regularly," while "duplicates are crawled less frequently" (Google Search Central). So the page Google now thinks is a copy of a stranger's maintenance notice gets visited less often, and the recrawl that would clear things up is the very visit that just got deprioritized. That is the loop Mueller flagged as slow to reverse.
A few hours of downtime is survivable. A few hours of downtime that your server described as a normal, successful day is what leaves a mark.
Why nobody told you
Here's the thing about the standard uptime monitor. UptimeRobot describes its website monitoring as a check on "whether a page returns a successful HTTP response." A 200 is a successful HTTP response. Your maintenance page passed the test with honors, and the dashboard stayed green while customers bounced off a closed sign. UptimeRobot's own page for its keyword feature says it plainly: "A page can still return a 200 OK status while showing an error message, blank content, or a broken checkout flow" (UptimeRobot).
I ran a grooming salon for years, and the back-of-house problems never announced themselves. You found out from the client standing at the counter. A website works the same way, except the client is a crawler that doesn't complain. It records what it was given and leaves.
That recorded copy is the whole game. Google and the AI answer engines work from what the crawler fetched on the day it visited. Being found and quoted rather than just ranked rests on that copy being your real page and not your host's apology template. The same logic applies to a "not found" page that reports success, which is the soft 404 problem in its everyday form.
Find out what your site says when it's down
Nobody expects a salon owner or a plumber to know their status codes. The checks below take an hour once, and then they belong on someone's monthly list.
Ask your host or platform one question. "When my site is unavailable, what HTTP status code does your outage page return?" They should know the number. If they answer with a feature name instead of a number, ask again.
Check your maintenance mode. WordPress core's built-in maintenance mode does this right: its
wp_maintenance()function serves a 503 and sendsRetry-After: 600, per the WordPress developer reference. Coming-soon and maintenance plugins are a different story. Open the plugin's settings and look for a status-code option. If there isn't one, test it.Test it yourself. Turn maintenance mode on for sixty seconds and look at the number. In a browser, open the developer tools, pick the Network tab, reload, and read the Status column on the first row. From a terminal:
curl -I https://yourdomain.com
The first line of the reply is the status code. You want to see 503 while the holding page is up and 200 the moment it comes down.
Move the outage response upstream if you can. If your site sits behind a content delivery network (CDN) or a firewall service, that layer can return the 503 even when the server behind it is completely gone. Mueller made the same point in October: "You can do this outside of PHP / point DNS elsewhere / CDN."
Give your monitor a keyword. Add a check for a phrase that appears on your real homepage and never on a holding page. Your business name in the footer works. Now "the page loaded" and "the right page loaded" are two different alarms.
Mind the clock. A 503 is a short-term tool. Google's guidance is not to run one for "longer than 1-2 days," and warns that "if Googlebot observes these status codes on the same URL for multiple days, the URL may be dropped from Google's index." For a closure longer than that, Google's pause-your-business guide describes a different recipe, built around an indexable placeholder home page.
After an outage, look at what Google kept. Open Search Console, inspect your homepage URL, and read the indexed version. If it shows the holding text, request indexing and move on.
Hosting keeps the server on; it doesn't check what the server is saying. The difference between those two sentences is the entire reason this post exists.
Repairs that dig the hole deeper
- Leaving the 503 up for a week. The temporary signal stops being temporary, and pages start dropping. The 2011 post already warned that "lasting 503 HTTP result codes" can read as a permanently unavailable server.
- Taking the site "properly" offline. Google's pause guide lists the moves to avoid: returning 403, 404, or 410 for pages that still exist, disallowing all crawling in
robots.txt, and using Search Console's temporary removal tool for a closure. Each one asks Google to forget you, and it will. - Serving a 503 for robots.txt. The guide calls this out separately: "Don't return a 503 HTTP response status code for the robots.txt file because this blocks all crawling."
- Rewriting the homepage after the outage to "refresh" it. The copy Google holds is stale, not wrong. A recrawl fixes stale. New copy written in a panic needs its own recrawl and introduces its own problems.
- Trusting the green dashboard. Green means the server answered with a success code. It says nothing about what the answer was.
None of this is a reason to lose sleep over the next outage. Every host has them. The job is making sure your site tells the truth while it's happening.
FAQ
What status code should a website return when it is down?
A 503 (Service Unavailable), ideally with a Retry-After header saying when to check back. Google's documentation says a 503 prompts its crawlers to slow down and preserve already indexed URLs, which is what you want during an outage. A 200 with a "back soon" message tells Google the holding page is your content.
Does a 200 "back soon" page hurt search rankings if the outage only lasts a few hours?
It can. Per Search Engine Roundtable's November 17, 2026 report of John Mueller's LinkedIn reply, the holding text can be indexed in place of your page, and if other sites show the same text, your pages can be grouped with them as duplicates. The first problem clears on the next recrawl. The second is slower, because duplicates are crawled less often.
How long can a site return a 503 before Google drops pages?
Google's crawl-rate guidance says not to serve 500, 503, or 429 responses for longer than one to two days, and that a URL returning those codes for multiple days may be dropped from the index. For a longer closure, Google recommends a different approach built around a placeholder home page served normally.
Does WordPress maintenance mode send a 503?
WordPress core's built-in maintenance mode does. The wp_maintenance() function returns a 503 and a Retry-After: 600 header. Maintenance and coming-soon plugins each make their own choice, so check the plugin's settings or test the response yourself.
Why didn't my uptime monitor alert me?
Most basic website monitors check for a successful HTTP response, and a 200 holding page is a successful response. To catch a page that loads but shows the wrong content, add a keyword check for a phrase that appears only on your real page.
How do I check what status code my site returns?
In a browser, open the developer tools, select the Network tab, reload the page, and read the Status column on the first row. From a terminal, curl -I followed by your address prints the status line first. Do the check once with the site running normally and once with maintenance mode switched on.
This is the kind of check that belongs on a monthly runbook, not on an owner's to-do list. It is one of the quiet checks we run on every site under Care: what the site says when it's up, what it says when it isn't, and whether the monitor would notice the difference.
Want a back office for your site?
The build is free on any Care plan with a 12-month minimum. Care starts at $199/mo, and you can switch tiers anytime.
Apply for a build slot →RELATED READS