Free Link Extractor - Extract All URLs from a Web Page

In page mode the browser cannot fetch an arbitrary third-party page itself, because the target site's CORS policy blocks cross-origin reads. So the URL is passed to U2L's stateless fetch-text endpoint, which issues one GET with a browser-like user agent, follows up to five redirects, enforces an 8-second timeout, and returns the first 1 MB of the body together with the final URL after redirects, the HTTP status, and the content type. It refuses private and internal addresses, and it returns HTTP 415 for non-text responses such as a PDF or an image, so the tool can tell you the URL is not a web page instead of showing an empty table.

The HTML is then parsed in your browser with DOMParser using the text/html type. DOMParser builds a full document tree with the browser's own HTML5 parser but never executes scripts, loads subresources, or renders anything, which makes it both safe and fast. The tool checks for a base element with an href, resolves that href against the final URL, and uses the result as the base for every link on the page, mirroring the HTML specification's rule that the first base element with an href sets the document base URL. Each a[href] is then resolved with the URL constructor, which implements RFC 3986 reference resolution: ../pricing becomes an absolute path, //cdn.example.com keeps the page's scheme, and #section stays on the same page.

Anchor text is the element's textContent with whitespace collapsed, falling back to an image's alt text when a link wraps only an image. The rel attribute is split into tokens and the tool records nofollow, sponsored, and ugc, the three values Google introduced or redefined in September 2019 for outbound-link classification. The link's type is decided by comparing registrable domains: the last two labels of the hostname, or three when the suffix is a two-label public suffix such as co.uk or com.au. Links with a mailto:, tel:, or javascript: scheme are kept but marked Other and are never offered for shortening. The canonical link element is collected as its own row so you can confirm the page's declared canonical alongside its links; /tools/canonical-url-checker goes deeper on that specific question.

Text mode does not need a server at all. A single regular expression matches three shapes: URLs with an http or https scheme, addresses that start with www., and bare domains ending in a common top-level domain with an optional path. Trailing punctuation such as a period, comma, or closing bracket is trimmed because sentences end and parentheses close after URLs. Bare domains and www. addresses are normalised to https:// for the URL column so they can be copied, checked, or shortened directly. Because there is no source document, text-mode rows have no anchor text and no internal/external classification. Both modes feed the same table, filters, dedupe, copy, and CSV export, and the CSV is built from the filtered rows with RFC 4180 quoting and a UTF-8 byte order mark so Excel reads non-ASCII characters correctly.