Find the broken links a Markdown document proves by itself: references with no definition, definitions nothing uses, and anchors matching no heading.
Waiting for Markdown
Checks what the document can prove about itself. No link is ever fetched.
Links and images appear here
Every reference with its line, plus anything the document itself proves is broken.
Markdown Link Inspector runs entirely on your device — no link is requested.
You’re on the Free plan. Compare plans
Lists every link and image in a Markdown document, and reports the broken ones that the document proves by itself.
There are four of those, and all four are provable without leaving the page. A reference link like [text][ref] whose label has no matching definition renders as literal square brackets. A definition that nothing refers to is dead text. The same label defined twice means most parsers take the first and silently ignore the rest. And an anchor like (#some-heading) that matches no heading in this document is a link to nowhere — checked against the same GitHub anchor rules the Outline tool uses.
What this tool will not do is tell you whether https://example.com/page is still there. Answering that means sending a request for every external link in your document, which means telling someone else's server what you are working on, and in practice means uploading your document's link graph to whoever runs the checker. Every 'broken link checker' that works in a paste box is doing exactly that. This one states the limit instead.
Images are listed separately and their alt text is checked for emptiness. In Markdown the alt text is whatever sits between the square brackets of , and when it is empty a screen reader has nothing to read out — often falling back to the filename, which is rarely useful.
Links inside fenced code blocks and inline code spans are ignored, because [not](a-link) inside backticks is an example, not a link.
No, and that is deliberate. Checking a URL means requesting it, which means sending your document's links to someone else's server. Everything this tool reports is provable from the document alone, in your browser.
Four things: a reference with no definition, a definition nothing uses, a label defined twice, and an anchor that matches no heading in this document. All four are facts about the file rather than about the internet.
Most likely it points at a heading in a different file, which this tool cannot see. Anchors are only checked against the headings in the document you pasted.
No. Something that looks like a link inside a fence or inside backticks is an example, not a link, and counting it would be a misreading.
It flags images whose alt text — the part between the square brackets — is empty. A screen reader then has nothing to announce and often falls back to the filename.
They are dead weight, and they are often the remains of a link that was edited inline or deleted. Harmless, but they accumulate in long documents.