No, do not set up a 404 redirect to homepage as a blanket rule. Sending every missing URL to your homepage is worse than leaving those URLs as 404s, because Google reads a redirect that lands somewhere unrelated as a soft 404 rather than as a real move, and you lose the clean signal that the page is gone.
The precise rule is narrower than most articles make it sound. Redirecting an individual 404 is often exactly right. What breaks things is a catch-all rule that sends everything to one destination regardless of what the original URL was about. The rest of this post is the per-URL decision you should be making instead, and how to work out which 404s are worth your time at all.
Last updated 7 September 2026.
Why blanket homepage redirects hurt
Google’s Martin Splitt addressed this directly in the SEO Office Hours Shorts series. His point was that a 404 is useful information: a crawler that hits one learns the content is gone and moves on. Redirect it to the homepage and you replace a clear answer with a misleading one, and at scale Google starts reading those redirects as soft 404s. That is the strongest guidance available, and it comes from Google rather than inference.
The mechanism matters more than the rule. A 301 is a promise that the content moved to the destination. When the destination has nothing to do with the request, Google catches the mismatch and treats it as an error page in disguise. You get the deindexing a 404 would have given you anyway, plus a Search Console report full of soft 404s hiding the real problems.
For users it is worse. Think of a phone system where dialling a dead extension drops you at the main menu without ever saying the extension is dead. You assume you misdialled and try again, and get the same silent bounce. A 404 page that says the page is gone and offers a search box gets someone to what they wanted faster than a homepage pretending nothing happened.
The decision framework, one URL at a time
Pull your list from Search Console, under Indexing, then Pages, then the “Not found (404)” row. Export it. For every URL on that list, work down these three questions in order and stop at the first yes.
1. Is there a live page that answers the same thing the old URL answered?
Then 301 to that page. This is a genuine move and what redirects are for. The test is whether someone who clicked the old link would consider their question answered by the new page. If they would have to go looking again, it fails, and you move to question two.
2. No direct replacement, but does the URL have real value attached?
Real value means external links, or meaningful traffic before it broke. Then 301 to the closest relevant parent: the category, hub or section the page sat inside. Not the homepage. A visitor landing on the category sees the neighbourhood they were aiming for, and the link equity lands somewhere topically related rather than on a page about your whole company.
3. Neither a replacement nor any value?
Let it 404. This is the correct outcome for most URLs on most lists, and doing nothing is doing the right thing. If the page is permanently gone and you want the intent recorded explicitly, serve a 410 instead.
One correction on 410, because the internet still repeats an outdated claim. Google’s crawling documentation, revised in February 2026, says all 4xx errors except 429 are treated the same way, so 410 does not deindex faster than 404. Its value is being unambiguous for your own team and other crawlers, not speed. Use it for the precision, but do not rebuild anything expecting a faster result.
How to work out which 404s are worth fixing
A mid-sized site can easily report four hundred 404s. Nobody is hand-reviewing four hundred URLs, and most of them do not deserve it. Sort them like this.
- Has external links pointing at it. Highest priority, because that is authority arriving at a dead end. With no backlink tool, GA4 referral traffic to a URL that no longer exists gives you the same signal free.
- Had traffic before it broke. Second priority. Set the Search Console date range to before the URL disappeared and check whether it earned clicks. A URL with a year of clicks behind it is worth ten minutes. One that never earned any is not.
- Still linked from your own site. Crawl with Screaming Frog and check inlinks on 404 URLs. Fix these by editing the link, not by adding a redirect. A redirect papering over your own broken link is a permanent tax on every crawl, and you control the source.
- Everything else. Ignore it. Most of any 404 list is junk: scraper-appended tracking parameters, paths from a plugin you removed in 2021, bots probing for admin files that never existed. Redirecting these is exactly what produces the soft 404 mess.
This usually collapses four hundred URLs into fifteen or twenty that matter, which is an afternoon rather than a project. On a D2C retail client the same cleanup, sorting real product 404s from crawl noise and getting the survivors indexed again, was part of the technical work behind a 45% rise in purchases, written up in my case studies.
What “closest relevant page” means in practice
The advice to redirect to the closest relevant page is useless without examples, so here are three real shapes it takes.
Discontinued e-commerce product
/products/steel-shelf-bracket-v2/ is gone. If a v3 exists, redirect to it, because a buyer wanting the bracket is served. If the line is discontinued entirely, redirect to /products/shelf-brackets/, the category. Someone shopping for brackets sees brackets. If it was a one-off clearance item with no links and no history, 404 it and remove the internal links pointing at it.
Deleted or merged blog post
/blog/seo-tips-2019/ was folded into a newer guide. Redirect to the guide that absorbed it, since the content genuinely moved. If it was thin, earned nothing and nobody links to it, do not redirect it to your blog index out of habit. A blog index answers no specific question, so it is only slightly better than the homepage. Let it 404.
Retired service page
/services/ppc-management/ is retired because you stopped offering it. If the closest thing you still do is paid social, redirect there only if a PPC enquiry would be sensibly answered by that page. Otherwise redirect to the services overview, which at least tells the visitor what you do now. This is the one case where a broad destination is defensible, because the parent is still directly relevant to the request. My own technical SEO and services page plays that role for anything under it.
Frequently asked questions
Is a 404 redirect to homepage bad for SEO?
As a blanket rule, yes. Google’s Martin Splitt has said mass-redirecting 404s to the homepage hurts you, because the redirects get read as soft 404s and you lose the clear signal that the content is gone. Redirecting a single 404 to a genuinely relevant page is fine. The catch-all rule is the problem, not the redirect.
Do 404 errors hurt my rankings?
Not directly. A 404 on a URL that should not exist is normal, and Google expects to find them on every site. What costs you is a 404 on a URL with backlinks, or one people still reach from your own navigation, because that is authority and traffic hitting a dead end. Judge each one by what points at it.
Should I use 404 or 410 for a page I deleted on purpose?
Either works. Google’s documentation says all 4xx codes other than 429 are handled the same, so 410 will not deindex any faster. Use 410 when you want the record to show the removal was deliberate and permanent, and 404 everywhere else, including wherever your platform makes 410 awkward to serve.
How many 404s is too many?
There is no threshold, because the count is the wrong metric. A thousand 404s from bot-generated URLs is fine. Five on pages with backlinks and previous traffic is worth fixing this week. Report the number of 404s with links or history attached, not the raw total, or you will spend the quarter chasing noise.
About the author. Kapil Lodhi is an SEO and AI SEO specialist in Indore, India, with over four years of hands-on experience across technical SEO, on-page and content strategy, link building, local SEO and e-commerce SEO for clients in India, the United States and Canada. He has run this 404 triage on sites reporting several hundred errors, and the answer is nearly always that fifteen of them matter.