Server error (5xx)
When Googlebot asked for the page, the server answered with an error instead. Sometimes the site was down. Often the page works fine in your browser, because the server treats a crawler differently from a person: a rate limit, a firewall, a page that holds up for one visitor and fails under a crawl.
What Google is telling you
The status sits under Why pages aren’t indexed in the Page indexing report. Google’s help page says the server returned a 500-level error when Google requested the page.
It is more than one failed visit. Google’s page on status codes says server errors make its crawlers slow down on the whole site; pages already in the index are kept for a while, “but eventually dropped” if the errors persist. Once the server answers normally again, Google raises the crawl rate gradually.
Is it still happening?
| What you see | What it usually means |
|---|---|
| URL Inspection’s Test live URL works now | A past error: an outage, a deploy, a busy hour. Fix what caused it, then validate. |
| The live test fails too | It is still happening, to Google’s crawler. Look in the server logs for Googlebot’s requests. |
| Many URLs on the same dates | The whole site was failing. The Crawl stats report, under Settings, shows when the server did not respond. |
| Always the same kind of page | One template is slow or crashes: search results, filters, a heavy report, a page that runs a big query. |
“But the page works in my browser”
The server answered Google, not you, and there are a few reasons it can answer a crawler differently:
- A rate limit. A rule written for people, so many requests in so many seconds, is easy for a crawler to hit. The answer is 429 Too Many Requests, and Google’s status-code page says a 429 “is considered a server error”.
- A firewall or bot protection that answers unknown bots with 503, or with a challenge page. A challenge is a different status: Cloudflare’s “Just a moment…” page comes back as 403, which Search Console lists as Blocked due to access forbidden (403).
- Load. A page that is fine on its own can fail when a crawler asks for hundreds of pages in a row: slow database queries, search and filter pages with no limit, a server with too little memory.
- A CDN between Google and your server. Cloudflare answers with its own 520–526 errors when it can’t get a good answer from your server: 521 when the server is down, 522 when the connection times out, and 524 when the server doesn’t answer within 125 seconds by default. Google sees them like any other 5xx.
To see what a crawler gets, use the live test in URL Inspection, which fetches as Google. In your logs, check that the requests really came from Google: its verification guide describes a reverse DNS lookup and lists its crawlers’ IP ranges. Anything can call itself Googlebot.
Find the failing pages yourself
- Export the URLs from the report and look for a pattern: one folder, one template, one parameter.
- Search your server or hosting error log for those addresses at the dates shown. The error message there is the cause.
- Trakis requests every page it crawls and lists the ones that answer with an error, with the status code each one gave (CRAWL-002). Pages that time out or refuse the connection altogether are reported separately (CRAWL-007).
How to fix it
- Fix the page type that fails: the slow query, the plugin that crashes, the memory limit. The error log names it.
- Let verified crawlers through your rate limits and bot rules. Cloudflare, for one, recognises Googlebot as a verified bot; exempt verified crawlers rather than raising the limit for everyone.
- Make removed pages answer 404 or 410, not crash. A deleted product whose page now throws an error stays in this report.
- Don’t use 401 or 403 to slow Google down. Google says so; a short spell of 5xx or 429 slows its crawl, an error you leave in place gets pages dropped.
Then run the live test on a few of the URLs and press Validate fix in the report. Watch Crawl stats: the share of server errors should fall, and Google brings its crawl rate back up over the following days.
Common questions
Will 5xx errors hurt my rankings? A short outage costs little: indexed pages are kept for a while. Errors that last get the pages dropped, and they slow Google’s crawling of the whole site.
Is a 429 a 5xx? Not by its number, but Google treats it as a server error.
Why does Search Console show 5xx for pages I deleted? The removed page throws an error instead of answering “not found”. Make it answer 404 or 410.
What about “Just a moment…” pages? That is a bot challenge, answered with 403. The guide to Blocked due to access forbidden (403) shows how to let Google past it.
What if robots.txt itself answers with a 5xx? Then it is the whole site: Google stops crawling for up to 12 hours and then works from the last copy of the file it had. More in the guide to Blocked by robots.txt.
How long until the report clears? Each URL is rechecked when Google crawls it again, so the count falls over days or weeks after the fix.
Check the whole site
The Trakis audit fetches every page as a crawler does and reports the ones that fail, counted once per problem with the pages it covers. When a site’s bot protection refuses our crawler, the report says so instead of scoring those pages as broken. Create a free account and add your site, or see every Search Console status, explained.
Last changed 13 October 2026.