Broken Link Checker: Find 404s and Dead Links

Give it your homepage (or a list of pages) and it crawls your site, checks every link it finds, and tells you which ones are broken: 404 and 410 pages, server errors, domains that no longer exist, redirect loops and redirect chains. For every link you get the HTTP status, a plain-language reason, the page it sits on and its anchor text, so you know exactly what to fix and where.

It works over plain HTTP like a search engine crawler: no headless browser, no proxy. That keeps it cheap: $0.50 per 1,000 links checked, half the price of the most used broken link checker in the Store. Links it cannot verify (bot walls, timeouts, robots.txt) are free.

It obeys robots.txt (RFC 9309) for every site it touches, honors Crawl-delay, slows down when a site answers 429 or 503, and never reads pages behind a login.

Who it is for

• Site owners and SEO teams: find broken internal links before Google does, after a redesign or a migration.
• Agencies: a link audit for a client site in one run, exported to CSV or Excel.
• Developers and AI agents: call it from the API or an MCP client after a deploy and fail the build on isBroken.

What you get

One result per unique link, problems first (broken, then blocked, unreachable, redirected, skipped, ok):

• url: the link, made absolute (the #fragment is dropped)
• result:
  • broken: 404, 410, any other 4xx, 5xx, a domain that does not exist, a redirect loop, a redirect with no target, or a link that is not a valid URL
  • blocked: 401, 403, 429 or 999. Usually a login or bot protection (big news sites, LinkedIn), not a dead page, so it is listed separately and not charged. Open these in a browser to be sure
  • unreachable: no answer within 15 seconds, or the connection failed (free)
  • redirected: works, but only after one or more redirects. Link straight to finalUrl
  • skipped: the target's robots.txt disallows robots, or it is a private address (free)
  • ok: answered 2xx directly
• isBroken (true or false), statusCode, error (why, in plain words)
• finalUrl, redirectCount and redirectChain (every hop with its status code)
• firstFoundOn and anchorText: the first page that links to it and the link text there (or the image alt text, aria-label or title for image links)
• foundOn (up to 10 pages with their anchor text) and foundOnCount (all pages read in the run that link to it)
• isInternal (same site; www. and the bare domain count as one), isStartUrl, contentType, responseMs, checkedAt

The run's status message sums it up, for example: 8 links on 2 pages: 3 broken, 2 redirected, 3 ok. Billed: 9 links checked. (the 8 links plus the start page). When some links cannot be verified it adds, for example, 32 not verified (blocked, unreachable or robots.txt).

Input

Field	Required	Default	Notes
startUrls	yes		Pages to start from, up to 1,000. Each one and every link on it is checked. An address typed without https:// is read as https
maxCrawlPages	no	100	How many more pages of the same site to open and read for links (0 to 10,000). 0 checks only the links on the pages you list
checkExternalLinks	no	true	Also check links to other websites (one request each, never crawled)

Check a whole site, up to 500 pages:

{
    "startUrls": ["https://www.example.com/"],
    "maxCrawlPages": 499
}

Check only the links on a few pages, internal links only:

{
    "startUrls": ["https://www.example.com/pricing", "https://www.example.com/blog/"],
    "maxCrawlPages": 0,
    "checkExternalLinks": false
}

How the crawl works: it reads the start pages, then follows links to other HTML pages of the same site until it has read maxCrawlPages more pages per site. Every link on every page it reads is checked once, however many pages share it. Links beyond the crawl limit are still checked, just not opened for more links. Files (PDFs, images, archives) are checked with a light HEAD request, with a GET fallback for servers that answer HEAD wrongly.

Output example

A real result from the the-internet.herokuapp.com/status_codes test page:

{
    "url": "https://the-internet.herokuapp.com/status_codes/404",
    "result": "broken",
    "isBroken": true,
    "statusCode": 404,
    "error": "The page answered HTTP 404 (not found).",
    "finalUrl": "https://the-internet.herokuapp.com/status_codes/404",
    "redirectCount": 0,
    "redirectChain": [],
    "isInternal": true,
    "isStartUrl": false,
    "foundOnCount": 1,
    "firstFoundOn": "https://the-internet.herokuapp.com/status_codes",
    "anchorText": "404",
    "foundOn": [{ "pageUrl": "https://the-internet.herokuapp.com/status_codes", "anchorText": "404" }],
    "contentType": "text/html",
    "responseMs": 17,
    "checkedAt": "2026-10-31T18:50:46.817Z"
}

And a working link that goes through a redirect:

{
    "url": "http://www.iana.org/assignments/http-status-codes/http-status-codes.xhtml",
    "result": "redirected",
    "statusCode": 200,
    "error": "Redirects once; link straight to http://www.iana.org/assignments/http-status-codes.",
    "finalUrl": "http://www.iana.org/assignments/http-status-codes",
    "redirectCount": 1,
    "redirectChain": [{ "url": "http://www.iana.org/assignments/http-status-codes/http-status-codes.xhtml", "statusCode": 302 }],
    "isInternal": false,
    "anchorText": "here"
}

Pricing

Pay per event, no subscription:

Event	Price	When
Link checked	$0.0005	Each unique link that gets a clear answer: ok, redirected or broken
Actor start	$0.00005	Once per run (Apify's standard start event)

Free, never charged:

• Start pages that cannot be read (dead, blocked, disallowed by robots.txt or not a valid URL); a start page that loads counts as one checked link
• Links that come back blocked, unreachable or skipped, and links that are not valid URLs
• The same link found on many pages (checked and charged once per run)

Example: a homepage plus pages that link to 2,400 unique URLs, of which 40 are blocked or time out, costs (1 + 2,360) x $0.0005 + $0.00005 = $1.18055.

Set a maximum charge on the run and the checker stops when it is reached: it never checks a link it cannot bill within your limit, and the status message says how many found links were left unchecked.

Limits

• No JavaScript rendering: links that only appear after scripts run are not seen. Most sites put their links in the HTML.
• Hyperlinks only (<a href> and <area href>). Images, scripts and stylesheets are not checked, and #fragment targets are not verified.
• A site counts as one host name without www.: blog.example.com is another site, so its pages are checked as links but not crawled.
• About 3 requests at a time per host (1 at a time with a Crawl-delay, which is honored up to 10 seconds). A crawl of a large docs site checked 3,136 links on 86 pages in 6 minutes in our tests.
• At most 100,000 unique links per run. Near the run's timeout it stops checking and saves what it has; give large sites a longer timeout.
• A server that answers 200 for a missing page (a "soft 404") looks fine to any HTTP checker, this one included.

FAQ

Why is a link "blocked" and not "broken"? Many big sites answer 403 or 429 to any robot while the page works fine in a browser. Calling those broken would bury your real problems under false alarms, so they get their own result and cost nothing.

Does it check links on other sites? Yes, with checkExternalLinks on (the default). Each external link gets one request; external sites are never crawled.

Why was a link skipped? The target site's robots.txt disallows robots on that path (for example x.com and twitter.com disallow everything). The checker respects that and does not charge for it.

Can I check a list of pages without crawling? Put them in startUrls and set maxCrawlPages to 0. Each listed page and every link on it is checked once.

How do I find the page to fix? firstFoundOn and anchorText show where the link sits; foundOn lists up to 10 pages and foundOnCount how many pages carry it.