A LinkedIn 999 status code means LinkedIn refused to show the page to your link checker. It doesn't tell you whether the profile, post or company page exists. LinkedIn sends 999 to requests it decides aren't coming from a real browser session, which is why link checkers so often flag working LinkedIn links as broken. To find out whether a link works, open it yourself. To stop the false alarms, exclude LinkedIn URLs in your checker's config and check that handful of links by hand every few months.
What the 999 status code means
999 isn't a valid HTTP status. RFC 9110 says "All valid status codes are within the range of 100 to 599, inclusive" and tells clients to treat anything outside that range as a 5xx server error (RFC 9110, section 15). Link checkers follow suit and count it as an error.
LinkedIn uses it as a refusal. When a person gets the same treatment in a browser, LinkedIn shows a "Request denied" page. Its help article says that happens when it has "detected an abnormally high number of requests coming from your network connection", that the block lifts by itself once traffic drops, and that support can't lift it by hand (LinkedIn Help).
Here's what a plain curl request for a real, public profile got in October 2026 (headers trimmed):
$ curl -s -o /dev/null -D - https://www.linkedin.com/in/<a-real-profile>
HTTP/2 999
content-type: text/html
content-length: 1530
cache-control: no-cache, no-store, no-transform
x-li-fabric: prod-lor1
server: cloudflare
The 1,530-byte body is a short script that sends a browser on to linkedin.com/authwall, LinkedIn's sign-in wall. A profile address we made up got the same response: same status, same length. As far as the status code goes, a live profile and a missing one look identical.
Why your checker gets 999 and your browser doesn't
LinkedIn's help page doesn't say what counts as abnormal. http.dev lists the usual triggers: a non-browser user agent such as curl, python-requests or wget, a high request rate from one IP address, and addresses in data center or cloud ranges (http.dev). A checker running in GitHub Actions, on a web server or in a hosted crawler matches most of that list before it sends a single request.
LinkedIn also says it outright. Its robots.txt opens with "The use of robots or other automated means to access LinkedIn without the express permission of LinkedIn is strictly prohibited" and its rules for all other bots are User-agent: * followed by Disallow: / (linkedin.com/robots.txt). A checker that honors robots.txt for outbound links won't request LinkedIn pages at all.
The answers aren't even consistent. In the same round of tests, curl -I (a HEAD request) for a made-up post URL under /posts/ came back 200. So a 200 from LinkedIn doesn't prove a link works either, and a checker that mixes HEAD and GET requests can flip a link between "broken" and "fine" from one run to the next.
You'll see advice to send a browser user agent. It sometimes gets a 200 for a while, it stops working whenever LinkedIn changes its filters, and LinkedIn's robots.txt notice prohibits that kind of automated access without permission. Checking a few links by hand takes less time.
How to check a LinkedIn link by hand
- Copy the link from your page's HTML or the live page, not from the editor, so you test the exact href.
- Open it in a private or incognito window, which shows you what a reader who isn't signed in to LinkedIn sees.
- Read the result:
- The profile, post or company page loads. The link works.
- You get a sign-in or join prompt. The link probably works: LinkedIn says people who aren't members "may be prompted to sign up or sign in", and members can switch off their profile's public visibility (LinkedIn Help). Sign in and open it again to be sure.
- You get a not-found page while signed in. The link is dead.
- You're redirected to a generic LinkedIn page instead of the post or profile you linked to. Treat it as dead. That's a soft 404: a missing page dressed up as a working one.
- If the href itself contains
/authwall?, someone copied the address bar after LinkedIn's sign-in redirect. The real address is URL-encoded in thesessionRedirectparameter. Decode it and use that as the link.
How to stop 999 false alarms in your link checker
Exclude LinkedIn URLs rather than accepting 999 as a success. An excluded link shows up as skipped or excluded in the report, which is accurate. Accepting 999 marks the link as working when nobody checked it.
lychee
Put this in lychee.toml in the directory lychee runs from (it reads that file by default):
exclude = [
'^https?://([a-z]+\.)?linkedin\.com/',
]
Or put the same regex on its own line in a .lycheeignore file at the repository root. lychee's own example config excludes LinkedIn the same way (lychee.example.toml). If you'd rather accept the code, accept = ["100..=103", "200..=299", "999"] works, but accept replaces lychee's default of 100..=103,200..=299, so you have to list the normal codes too (lychee README).
markdown-link-check
In the JSON config file:
{
"ignorePatterns": [
{ "pattern": "^https?://([a-z]+\\.)?linkedin\\.com/" }
]
}
Matching links are skipped (markdown-link-check README).
LinkChecker
In linkcheckerrc, the ignore option makes LinkChecker check only the syntax of matching URLs, without requesting them (linkcheckerrc docs):
[filtering]
ignore=
^https?://([a-z]+\.)?linkedin\.com/
Screaming Frog SEO Spider
Go to Configuration > Exclude and add linkedin\.com. The exclude list uses partial regex matching, and a matching URL "is not crawled at all" (Screaming Frog user guide).
Broken Link Checker for WordPress
With the plugin's local engine, open its settings, go to the "Which Links To Check" tab and add linkedin.com to the "Exclusion list" box. The plugin skips any URL that contains one of the listed words (checked against version 2.4.15). The cloud engine keeps its ignore rules in the WPMU DEV Hub instead, so check WPMU DEV's documentation for the current menu. If false alarms like this are why you're looking elsewhere, see alternatives to the Broken Link Checker plugin.
linkinator
Since version 7.2.0, linkinator reports 999 responses and Cloudflare-challenged 403s as skipped rather than broken (release notes). If that's your checker, you may not need a rule at all.
The "Link Checker Report" issue full of 999s
If your repository uses the workflow from the lychee-action README, it opens a GitHub issue titled "Link Checker Report" whenever lychee finds an error (lychee-action). Every LinkedIn link shows up in those reports as a line like this:
* [999] <https://www.linkedin.com/in/example/> | Rejected status code: 999 Unknown status code
lychee groups errors by file. One LinkedIn link in an author box or footer is repeated once for every generated page that carries it, so one link can produce many lines. Add the exclude rule above, or a .lycheeignore file (the action reads it from the repository root, or from the workingDirectory you set), and the next report drops them. The CI guide has a full workflow with a sensible fail policy.
When LinkedIn links really break
They do break, for a few specific reasons:
- A member changes their custom profile URL. LinkedIn allows up to five changes in six months and keeps the previous URL linked to the member for six months unless they unlink it. Someone who visits a previously claimed URL "may see a page that reads Profile Not Found" (LinkedIn Help).
- The member closes their account, or a company page is taken down.
- The post or article you linked to is deleted.
- The link was copied from the sign-in wall and points at
/authwallinstead of the page.
To fix one, search LinkedIn for the person or company and update the href, or remove the link and keep the name as plain text. If a post was deleted, there's nothing to point the link at, so quote what you need from it or drop the reference. The general options for any dead link are in how to fix broken links.
LinkedIn links are usually few: an author box, a team page, a few citations. Checking them by hand once a quarter is enough for most sites. The rest of your outbound links need a schedule of their own (checking external links).
Treat 999 as unchecked
A 999 answers a different question from the one your checker asked. It says LinkedIn didn't serve the page, so the honest label is "couldn't check". Mixing those into your broken-link count trains people to skim past the report, and the one LinkedIn link that really died gets lost among the ones that didn't.
Linkbruno shows a LinkedIn 999 as "blocked, couldn't check" instead of broken, and never sends an alert for it, so the link stays listed for a manual check without landing in your inbox. The same goes for Cloudflare challenges and 429 rate limits, covered in why link checkers report 403 and 429.
To see how your own site's links look, run the free 25-page scan, or start with the main guide to finding broken links.
