Google’s answer to 404 panic is that most of it is misplaced, and the real crawl risk sits with server errors. Gary Illyes and Cherry Prommawin made the case in a new episode of The Non-Canonical Cabinet, Google’s video series, which Barry Schwartz of Search Engine Roundtable covered on October 28. The video runs under 10 minutes.
Schwartz noted that nothing in it is new. The value is in the plain wording, because site owners still open Search Console, see a long list of 404 errors, and assume a penalty is coming. Prommawin described that reaction as expecting a manual action or a domain that “will be cursed and never rank again.”
No penalty, but a business question
Illyes opened with a flat answer: “No, we don’t care at all.” He added: “There is no penalty, no secret Google jail, and no negative impact on your other healthy pages.” Googlebot sees the 404, concludes the page no longer exists, and in time drops it from the index.
Fixing a 404 is a separate decision. In Illyes’s words, “Fixing a 404 is entirely optional, but it can be a smart business move depending on your site.” His example was a store that deletes a popular product page. If links or social media still send visitors there, a redirect to a newer model can save sales. He rejected the shortcut of sending every dead page to the homepage: “That just confuses humans and makes your data very confusing.”
The three-step framework
The hosts laid out an order of operations for deleted URLs:
- Redirect one-to-one to a true replacement, such as a refreshed article or a successor product. Prommawin stressed that the destination must be specific and highly relevant.
- Serve a 404 or 410 for pages removed for good with no successor.
- Make the error page useful, with a welcoming note, site search, and routes to top categories or fresh content.
Illyes deferred the mechanics of 301 and 302 redirects to a later episode and said the next installment will cover redirects.
Crawl budget and server errors
On the myth that accumulated 404s starve the rest of a site, Illyes was direct: “A 404 error does not drain your crawl capacity.” Prommawin then drew the contrast: “A 500 error affects crawl budget because it tells us the server is gasping for air.” Google slows down so it does not make matters worse.
Illyes said a 429 gets the same handling. Although it is formally a client error code, he described it as a server telling Google it is sending too many requests: “We treat that like a 500 because, frankly, it’s the same thing.”
How this fits the series
The robots.txt episode we covered on October 22 told teams that unparseable lines vanish silently, which created an invisible risk. This episode runs the other direction, telling teams that a highly visible report deserves less alarm than it triggers. Both share a limit: Google describes its behavior but publishes no thresholds, so there is no figure for how many 404s or how many 429s change anything. The transcript also skips broken internal links, which send users to dead ends whatever the crawl-budget answer.
What to do with it
Treat the Search Console 404 list as a triage queue. Sort URLs by the traffic and inbound links they still draw and whether a true successor exists, then redirect only where one does. Leave the rest as 404 or 410 and improve the template.
The more consequential check is on the other side. Teams that run bot rate limiting at a CDN or firewall should confirm Googlebot is not being served 429 or 5xx responses, since those are the codes Google says will slow crawling. Prommawin’s recap covers the rest: “if a page is gone and has no replacement, let it 404. If it has a natural successor, redirect it.”
Based on reporting by Barry Schwartz at Search Engine Roundtable, published October 28, 2026, on Google’s Non-Canonical Cabinet video about 404 errors, featuring Gary Illyes and Cherry Prommawin.
