Understanding 404 errors and how Bye404 works

This page answers the three questions people ask most often before running their first scan: what a 404 actually is, why it is worth repairing, and what the tool does when you give it an address.

What is a 404 error?

When a browser asks a server for a page, the server answers with a three digit code before sending anything else. A 404 means the server looked for the requested address and found nothing there. The address is syntactically valid, the server is running normally, but no content corresponds to it.

The name comes from the numbering scheme of the HTTP protocol, where the 4xx family covers errors attributed to the request rather than to the server. It is by far the most common error on the web, which is why it has become the one code that non technical visitors recognise.

Where 404 errors come from

A page was deleted or renamed without redirecting the old address to the new one. Everything that linked to it now points at nothing.
A link contains a mistake. A missing letter, a wrong extension, an extra character copied by accident. The destination exists, the link does not reach it.
The capitalisation does not match. Most web servers run on Linux, where Photo.JPG and photo.jpg are two different files. A site developed on Windows and moved to a Linux host can lose many addresses at once.
A site was restructured and the old address scheme was abandoned without redirections.
An external site links to you incorrectly, or a visitor typed the address by hand. You cannot prevent these, but you can make sure they land somewhere useful.
A relative path broke. An address written as ../images/logo.png resolves differently depending on how deep the page containing it sits.

A 404 is not a malfunction. It is the correct answer to a request for something that does not exist. The problem is never the code itself, it is a link on your own site that should not be pointing there, or an error page that leaves the visitor with nowhere to go.

Why is it important to fix 404 errors?

Crawling effort is wasted

Search engines allocate a finite amount of crawling to each site, sometimes called the crawl budget. Every request spent on an address that leads nowhere is a request not spent discovering new content or refreshing existing pages. On a small site this is negligible. On a site with thousands of pages it measurably slows down how quickly your new content appears in search results.

Your internal structure develops holes

Internal links are how ranking signals travel between your pages and how visitors move from one subject to a related one. A broken link is a dead end. The page it was meant to lead to loses the signals that link was passing along, and if it had no other inbound link it may become unreachable to crawlers entirely.

Visitors leave

Someone who clicks a link and hits an error rarely tries a second route. They return to the search results and open a competing site. On a commerce site the effect is immediate; on an editorial site it shows up as a bounce rate nobody can account for.

The site looks abandoned

Visible broken links tell a reader that nobody is maintaining the site, which changes how much they trust what it says. No algorithm measures this directly, but every human visitor registers it.

What a 404 does not do

It does not trigger a penalty. A page returning 404 is simply dropped from the index over time, and the presence of 404 errors is not in itself a ranking factor. The damage described above is indirect, which is why the useful priority is repairing broken internal links, the ones inside your own pages, rather than chasing every error in your logs.

How does Bye404.org work?

The tool behaves like a search engine robot restricted to your domain.

You give it a starting address. Usually the homepage, but any page works if you want to explore one section.
It requests that page and records the status code the server returns.
It reads the HTML and extracts every link the page contains.
It keeps the internal links, those pointing at the same domain, and adds them to a queue.
It repeats on each queued address, in small paced batches so the site being tested is never flooded.
It stops when the queue is empty or when it has covered 2500 pages.

Any address answering with a code in the 4xx or 5xx families is recorded as broken, together with the page the link was found on. Two codes are deliberately excluded: 403 Forbidden, which usually means the server is blocking automated requests rather than that the page is missing, and 429 Too Many Requests, which simply means the server asked the crawler to slow down.

The report is built around the source page

Knowing that an address returns a 404 is only half of what you need. The other half is the list of pages containing the link, because those are the files you have to open and edit. Every entry in the report gives you both.

What the tool cannot see

A crawler only reaches what is linked. Pages nobody links to, known as orphan pages, are invisible to it even though the address works perfectly. Content behind a login cannot be reached either, and addresses generated by JavaScript after the page loads are not followed. If a section of your site is missing from the report, the first thing to check is whether anything actually links to it.

Three tools, three jobs

The link checker follows the links between your pages and reports pages that do not answer. The image checker requests the images, media and documents your pages reference and reports the files that are missing. The redirect checker reports the addresses that answer with a 301, 302, 307 or 308, along with chains, loops and redirects that end on a dead page.

If your site has many pages, run the link checker first, since a missing page takes its images with it. Run the redirect checker last: it is the one that finds the problems the other two cannot see, because a redirection is not an error and never shows up in a broken link report.

Data and privacy

The state of a scan exists only while the scan is running and is deleted when it ends or when you press stop. Results are displayed in your browser and are not stored afterwards, so copy anything you want to keep before leaving the page. Each visitor is allowed 20 scans per 24 hours so that the service stays responsive and does not become a nuisance to the sites being tested. See our cookie information and terms of use for the full picture.

Going further

Once you have a report in front of you, the step by step repair guide walks through the decision for each entry. The status code reference explains every response you might see, and the FAQ answers the questions that come up most often.