Do 404 errors hurt SEO?

Not directly. Google treats dead URLs as normal and they don't drag down your other pages. What costs you is a 404 that used to earn links or traffic, and internal links pointing at pages that no longer exist. The count itself is not a ranking factor.

Should I redirect all my 404s?

No. Redirect the ones with backlinks or recent organic traffic to their closest equivalent. Redirecting everything else — especially to the homepage — gets treated as a soft 404, so you get the work without the benefit.

What's the difference between 404 and 410?

404 means not found; 410 means deliberately and permanently gone. Google drops 410s slightly faster and re-checks them less often. The practical difference is small enough that it's rarely worth engineering time — use 410 when your platform makes it easy.

What is a soft 404 in Search Console?

A URL that returns a 200 status but whose content reads as missing or empty. Google won't index it and keeps crawling it because the status says it's alive. Fix it by returning a real 404, or by putting real content on the page.

How many 404s is too many?

There's no threshold. A site with 3,000 clean 404s and no internal links pointing at them is fine. What matters is whether the pattern indicates something broke — a whole URL family disappearing at once is worth investigating; a long tail of old campaign URLs isn't.

Does a custom 404 page help SEO?

It helps retention, not rankings. A visitor who finds a search box and three relevant links is a visitor you might keep. Just make sure the page still returns a 404 status — a pretty custom page returning 200 is a soft 404 with better typography.

404 Error: What It Means and How to Handle It

404 error

Definition

A 404 error is the HTTP status a server returns when a URL doesn't exist. It is not a penalty and it doesn't drag down the rest of your site — Google expects dead URLs on any site with history. The only 404s worth acting on are the ones with backlinks or recent organic traffic.

Updated 26 July 2026 · Written by the Last Agency team · See what SEO actually costs

The short version

404s are normal. A site with 3,000 clean 404s can rank perfectly well; the count on its own means nothing.

Triage rule: redirect a 404 only if it has backlinks or had organic traffic. Otherwise leave it.

A soft 404 — a not-found page returning 200 — is the version that actually causes problems.

410 is marginally faster than 404 at getting a URL dropped. Marginally. It isn't worth a project.

A 404 is not a penalty

This misconception is worth killing first, because it drives a lot of pointless work. Google has been explicit for years that 404s are normal and don't harm the ranking of your other pages. Products get discontinued. Campaigns end. Somebody mistypes a URL. A site live for six years reporting zero 404s is either brand new or lying about its status codes.

What a large count sometimes indicates is a structural fault — a migration that dropped a URL pattern, an export that broke every product link. That's worth investigating. The number is a symptom you check, not a score you optimise.

An audit tool flagging "3,000 broken URLs" in red is selling urgency. Open the list, sort by backlinks. Usually eleven of them matter.

404 vs soft 404 vs 410

Three ways to say a page is gone, and only two of them are honest.

The three not-found states and what Google does with each.
State	What the server returns	What Google does
404 Not Found	404 status, with or without a helpful page	Drops the URL from the index over time, re-checks occasionally in case it returns.
410 Gone	410 status — permanently removed, deliberately	Same outcome, marginally faster, and Google re-checks less often.
Soft 404	200 OK on a page that says the content is missing	Flags it as a soft 404, refuses to index it, and keeps crawling it because it looks alive.

Why soft 404s are the genuine problem

A soft 404 is a lie your server tells. The page returns 200 — "here is your content" — and the content is the words "page not found". Google can't tell from the status code that the URL is dead, so it keeps fetching it. On a large site that's real crawl capacity spent on nothing.

Two common causes: single-page apps that route client-side and serve 200 for every path, and CMS setups where a deleted item quietly lands on an empty category listing. Both look fine in a browser and both are wrong at the protocol level. Check with curl -I — the browser won't tell you.

Google also applies the label to redirects landing somewhere irrelevant, which is why mass-redirecting old URLs to the homepage doesn't work. See 301 redirect.

The triage rule

Don't redirect 404s as a policy. Redirect the ones carrying something worth keeping. Everything else stays a 404, and that's the correct outcome rather than a compromise.

What to do with each kind of 404.
The dead URL	Action
Has external backlinks	301 to the closest equivalent. This is why redirect maps exist — you're recovering an asset someone else gave you.
Had organic traffic in the last 12 months	Restore the page if the content still holds. Redirect if something replaced it.
Discontinued product with a successor	301 to the successor, or the category if there isn't one.
Discontinued product with nothing similar	404 or 410. A useful not-found page beats a redirect that doesn't answer the request.
Expired campaign, event or job listing	404, with a link to the current equivalent.
URL that never existed — typos, scraped junk, `/wp-login.php` on a Next.js site	Leave it. Nothing to fix.

Finding the ones that matter

Search Console, Page Indexing, "Not found (404)". Google's own list of dead URLs it has crawled.
Your backlink tool's broken-pages view. Ahrefs, Semrush and Moz all filter dead URLs by referring domains. Sort descending, stop after the first page.
Analytics, last 12 months, filtered to your 404 page title. What people are still trying to reach.
Server logs. The only source catching URLs nobody links to but bots keep requesting. See log file analysis.

Building a 404 page worth landing on

The page must return a 404 status code. That's the non-negotiable part, and the part frameworks get wrong. Everything after it is user experience — a real cost, since a dead end is a lost visitor who had already decided to come to you.

Say what happened in plain language. Offer the search box. Link your three or four most useful pages, not a sitemap dump. Skip the automated "related pages" widget unless it's genuinely accurate; a bad guess is worse than an honest dead end.

And check the status code after every deploy, in the same breath as checking robots.txt. curl -I https://yoursite.com/a-url-that-does-not-exist should return HTTP/2 404. If it returns 200, your framework is serving a shell and every mistyped URL on the internet is now an indexable page on your site.