Search "404 error" on Google. On October 15, 2026, in Canada, the AI-generated summary at the top of the French results claimed that too many broken links can hurt your SEO. A little further down, Google's own Search Console help says the opposite: 404 errors don't harm a site's indexing or ranking. The help page is right. But the conclusion most people draw from it, "so it doesn't matter," is wrong too. That's why we monitor our clients' sites continuously: a dead page rarely costs you rankings. It costs you visitors who follow your link into a dead end.
This piece isn't about redesigns, when every address changes at once; our article on redesigning without losing your SEO covers that. It's about slow wear: a promotion taken down, a service page renamed, a PDF replaced. Add those up over three years and these ordinary moves leave dozens of dead addresses behind, some of which still get visitors.
Pew Research Center measured the effect: a quarter of the web pages that existed at some point between 2013 and 2023 were no longer accessible by October 2023, most often because the page had been deleted from a site that was still online. For pages from 2013, the share rises to 38%.
The short answer, for the busy
A 404 error is what a server returns when the requested address doesn't exist. According to Google's Search Console help, 404 errors don't harm a site's indexing or ranking: Google simply drops the dead page from its index. The real cost is elsewhere. A dead page loses the visitors who arrive through links you no longer control: an article on another site, a printed QR code, an email sent two years ago, a link in your Google Business Profile. If the content moved, a permanent 301 redirect to the new address recovers that traffic. If it's truly gone, the page should return a 404 or a 410, and not redirect to the home page, which Google treats as a "soft 404." To find them, combine Search Console's Page indexing report, a crawler and your server logs, because each one has blind spots. Then keep watching: links die one at a time.
What Google actually says, and what it doesn't
404 errors "don't harm your site's indexing or ranking," says the Search Console help page, and most of them aren't worth fixing. A 2011 Google post, still online, adds that 404s are a perfectly normal part of the web. (If the index itself is fuzzy, start with SEO explained without jargon.)
Three details matter. First, Google doesn't punish a dead page, it forgets it: an indexed address that starts returning a 4xx error is removed from the index. If that page was your second-best source of quote requests, your site's rankings didn't move, but you lost its share of the traffic. Second, Googlebot comes back less and less often, and never forgets the address for good. Third, Google triages for you: fix only the 404s your own site links to or that appear in your sitemap, and redirect common misspellings.
In other words, the search engine handles your dead pages just fine. It's the humans who handle them badly.
Where the real loss is
A web address travels much further than your site. It's printed on business cards and flyers, often as a QR code. It sits in sent emails and archived newsletters. It appears in your Google Business Profile, in directories, on a partner's site. Each of those links was written once and will never be updated.
These links also bring the most motivated visitors. Someone scanning your QR code already has the flyer in hand; someone clicking from a partner's site arrives with a recommendation. An error page with no way forward sends them off at the exact moment they'd decided to listen. Google's 2011 post gives the fix: a 301 redirect from the moved or misspelled address to the right page captures the traffic that link was meant to bring.
Google Ads, at least, is watched: Google's destination policy disapproves an ad whose landing page returns an HTTP error code to its AdsBot crawler. That's a safeguard and a trap: a campaign can stop serving because someone renamed a landing page, and nobody connects the two. A printed QR code or an email already sent has no such mechanism. When it breaks, the only witness is the customer.
Soft 404s, 410s and redirect chains
A real 404 is a page that says "not found" and returns the 404 code: the right behaviour for content that's gone. A 410 ("Gone") is more explicit, but Google currently treats it exactly like a 404.
The soft 404 is the sneaky one. The page says "not found," or shows nothing at all, but the server answers 200, the code for pages that work. A common variant: the site redirects every unknown address to the home page, which feels helpful. Google advises against both, since they confuse users and search engines alike. To the visitor, the home page isn't an answer: they were looking for one specific page and start over without knowing it no longer exists. And a page that says 404 doesn't necessarily return a 404: only a tool that reads the server's response can tell.
Redirect chains are the disease of sites maintained in good faith. One year, old page A redirects to B. Two years later, B is renamed and redirected to C, and nobody goes back to fix A. Over time you get chains that are too long, even loops, which Search Console files under Redirect error. The fix: every old address points straight to the final destination. The detailed rules are in our redesign article.
Then there's the other direction: your links to suppliers, programs or regulations die too. In a sample of 2,063 news sites, Pew found at least one broken outbound link on 23% of pages. A visitor who hits an error doesn't blame the other site. They conclude your page is out of date.
The 404 page can still save the visit
Visitors will land on dead addresses anyway, so the 404 page deserves better than a funny illustration. Google pointed this out back in 2011: you can return a 404 code while showing whatever content you want. A good 404 page says plainly that the page no longer exists, offers a site search, and lists the three or four pages people actually want, which your analytics will give you in two minutes: usually a flagship service, contact, hours or address.
It can also act as a sensor. If every 404 page view sends an event to your analytics tool with the requested address and the referring page, you get the list of dead addresses that still receive visitors, sorted by volume. A dead address nobody visits costs nothing; the one on a flyer still in circulation costs a little every day.
Four sources, four blind spots
Search Console's Page indexing report. Free, it lists the addresses Google tried to crawl, with the reasons Not found (404), Soft 404 and Redirect error. Its own help page spells out the limits: it only shows 404s from the past month, and it ignores addresses Google never found. And for a site with fewer than 500 pages, Google adds, you probably don't need this report at all.
A crawler. A tool like Screaming Frog SEO Spider, free for up to 500 URLs per crawl, walks your site the way Google does and lists broken links, internal and outbound, plus redirects and chains, with the source page of each link. Its blind spot: it only sees what your pages link to. An address removed from every menu but printed on a flyer is invisible to it.
Server logs, or your 404 page's sensor. The only source that records every real request, including those from a QR code or an old email. Also the least friendly to read, hence the event described above.
Your own lists. No tool knows the addresses you printed or sent. Keep a register of every address published outside the site, and check it before deleting or renaming a page.
Monitor continuously instead of auditing once
A one-off audit is a snapshot: it finds today's dead pages, and the counter starts again. Links die at the pace of ordinary decisions: an offer withdrawn, a plugin that changes the shape of your URLs during an update. Only repeated checking keeps up: an automated crawl every week, an alert when a dead address starts getting visits, a monthly look at Search Console. Our article on website maintenance develops the same logic.
The deadest page of all is one whose entire domain has stopped working: in Pew's study, that was the case for 9% of the sampled pages. An expired domain is enough to get there, and our article on protecting your domain name explains how to avoid it.
Where to start
Start with what no tool will find on its own. Gather the addresses your business has published outside the site over the past three years (QR codes, flyers, email signatures, campaigns) and open each one today. Then check the links in your Google Business Profile, booking link included. Run a crawler over your site once to fix internal links and chains. Then wire the sensor into your 404 page, so the next dead address reports itself before a customer does.
→ Let's talk about monitoring your site
Written byXavier Peich