HTTP 404 error

Long definition

Status 404 is defined in RFC 9110 as "Not Found" — the server cannot match the request to a resource right now and offers no opinion on whether the resource ever existed or might come back. That ambiguity is intentional. It lets a server return 404 for typos, deleted content, never-existed URLs, and temporary configuration mistakes without distinguishing between them.

Google's handling is forgiving. A single 404 doesn't hurt the rest of your site. Googlebot retries 404s several times over weeks before dropping the URL from the index, in case the page comes back. The Search Console "Pages" report shows 404s under "Not found (404)" — a healthy report has some 404s; sites with zero are usually misconfigured (returning 200 or 301 for things that should 404).

The 404 patterns that hurt:

• Soft 404s — pages that return 200 OK but show "Not Found" content. Google detects these heuristically and treats them worse than honest 404s. See the dedicated entry.
• Mass 404s after a migration — an unfixed redirect map drops thousands of URLs into 404 simultaneously. Recover with proper 301s where there's an obvious target, accept 404 (or upgrade to 410) where there isn't.
• 404 on critical infrastructure — robots.txt, sitemap, key category pages. These cascade into broader crawl problems.
• Internal links pointing to 404s — wastes crawl budget on every recrawl. Audit and fix.

Don't reach for 301-everything-to-the-homepage. Google calls that pattern "soft 404" because the destination is irrelevant to the requested URL. The right move: 301 to a topical equivalent if one exists, otherwise honest 404, or 410 Gone if you intentionally removed the page and never want it back.

For UX, a custom 404 page with site search and links to popular sections recovers some lost sessions. Just make sure the page itself returns the 404 status, not a 200 with "Not Found" text.

Common misconceptions

• "404s hurt site rankings." Individual 404s don't damage anything. Google has been explicit: 404s are normal and expected, and the algorithm doesn't penalize sites for having them.
• "Redirect every 404 to the homepage to capture lost traffic." This creates soft 404s and gives Google a worse signal than the honest 404. Worse for SEO, worse for users who land on an unrelated page.
• "A 404 means Google forgets the URL immediately." It doesn't. Google retries the URL multiple times across weeks before dropping it. If you actually want fast removal, return 410 Gone instead.
• "Custom 404 pages with HTTP 200 are fine because Google understands." Google detects most soft 404s but not all. Always return the actual 404 status code at the HTTP layer — don't rely on heuristic detection.