Try it first: the link galaxy. Enter a domain, press “Map the galaxy”, and after 20 to 60 seconds your website spins as a galaxy. No domain at hand? The demo button fills in daniele-michelin.com.
How to read the galaxy
- Core = homepage: the workflow counts click depth from here. The farther a star is from the core, the more clicks a visitor needs, and so does a crawler.
- Star = page: the size shows the inbound internal links. Big stars are the pages you link to most yourself. Google reads that as a signal: this page matters.
- Colour = section: every section, meaning the first path segment such as
/blogor/tools, forms its own spiral arm. Language prefixes like/en/are skipped, so/en/bloglands in the blog. - Filaments = internal links: hover a star and its outgoing links light up in cyan, the links pointing at it in magenta. A click opens the page in a new tab.
- Black hole = 4xx or 5xx: an internal link leads nowhere. The red accretion disk is hard to miss, and that is the point.
- Comet = redirect: the link points to an old address and the server redirects. The tail points away from the target.
- Planet at the edge = orphan page: it is in the sitemap, but no internal link leads there. These planets drift slowly around the galaxy on their own orbits.
Next to it sits the “Findings” panel with every list: broken links with the page they are on, orphan pages, deep pages more than three clicks away and the top pages by inbound links. Below is a table of every crawled page. It is the text alternative to the graphic and works with a screen reader too.
What matters for SEO
- Black holes first: a broken internal link wastes link equity and sends visitors to an error page. For every error the list shows the page the link is on, so you know exactly where to fix it.
- Link the orphans: Google does find a page that is only in the sitemap, but treats it as unimportant. One link from an overview or a related article is often enough.
- Reduce depth: pages more than three clicks from the homepage get crawled less often. Overview pages, categories or a menu link bring them closer to the core.
- Catch the comets: a redirect works but costs an extra round trip. Link internally straight to the target address.
- Check the big stars: the top pages by inbound links should be the pages you want to be found with. If it is only the legal notice, the weighting is off.
From these findings the workflow computes a link health from 0 to 100. Broken links weigh most, then orphans, deep pages and redirects. The AI astronomer sums it up in three or four sentences and names the most important fix.
What the workflow does
- Webhook with a pre-check: before n8n even starts, the server checks the domain: public host names only, no IP addresses, no private networks. Add a pause between two galaxies and a limit per hour.
- Observatory, a light crawler: an HTTP node calls a small PHP crawler. It reads
robots.txtfirst and respects it, then the sitemap. Then it crawls at most 150 pages with four requests at a time and short timeouts. It stays on the checked host, and every request goes to the public address checked before. External links are only counted, never fetched. - Compute the star map: a Code node computes click depth by breadth-first search from the homepage, plus links in and out, sections, errors, redirects, orphans and deep pages. Links are stored as pairs of indices, which keeps the result below 150 KB.
- Clear sky? An IF node checks whether the crawl worked. If not, the workflow skips the AI call and the page shows a clear message.
- AI with rules: the astronomer is gpt-oss-120b at IONOS in Germany. She gets only numbers and paths and must not invent anything. If she does not answer, or answers too briefly, a Code node writes the log from a template.
- Sharing with a picture: the server stores the galaxy with a number and a link and draws a preview image. WhatsApp and LinkedIn show the galaxy when you share, not just a link.
Build it yourself
The workflow is ready as JSON to import. Placeholders are marked: the crawler address, the token and the page that receives the result. You need the crawler as your own endpoint. The important parts: allow public addresses only, stay on the host, respect robots.txt and set a hard page limit. Any OpenAI-compatible API works for the AI.
The graphic needs no library: one canvas, a hand-written 3D projection, at most 60 frames per second and a still view for everyone who prefers reduced motion. More workflows to click and download are in the n8n overview, and if you would rather see your website as a city, build it here: your website as a city.