Internal Linking as a System
Internal links are the cheapest ranking lever most sites never pull on purpose. This guide turns them from a chore into a system: three link roles, a hub-and-spoke map you can draw in an afternoon, anchor rules, a four-check audit, and a routine that keeps it alive after launch.
Ask most teams how a page gets internal links and the honest answer is: whenever a writer remembers. The navigation links everything to everything, the footer repeats it, and the body of each article links to whatever the author had open. Crawlers read that as a site with no shape. Nothing says which pages matter, which pages belong together, or where a new page sits.
A linking system fixes that with a small number of rules applied every time. It costs no tools, no plugins and no new content. On most sites we audit it is the single change with the best ratio of ranking movement to hours spent, and it is also the one that makes answer engines quote the right page instead of a stray one.
Key takeaways
- A link has one of three roles: navigation, hub to spoke, or contextual in-body. Each role has its own rule, and a page that only gets navigation links is effectively orphaned.
- Draw the hub-and-spoke map before writing anything new. Every spoke links up to its hub; every hub links down to all its spokes; spokes cross-link only within the cluster.
- Audit four things: orphans, click depth over three, link concentration on pages that do not sell, and links that only exist in JavaScript. Then put a 20-minute check into every publish.
In this article
Why it moves rankings when nothing else has
Search engines discover pages by following links, decide what a page is about partly from the words in the links that point to it, and weigh a page partly by how many of your own pages treat it as important. Internal links are the only one of those signals you fully control. External links are slow and expensive. Content is slow. Linking is a Tuesday afternoon.
The same structure is what answer engines lean on. When a crawler finds the same page linked from a hub, from three related articles and from the navigation, all with consistent anchor text, it has a clear answer to the question 'which page on this site is about X'. When it finds five pages all weakly pointing at each other, it picks one at random, and it is often not the one you would have chosen.
The three link roles
Every internal link does one of three jobs. Naming the job tells you the rule for it.
| Role | Where it lives | Rule |
|---|---|---|
| Navigation | Header, footer, sidebars, breadcrumbs | Appears on every page, so it carries almost no information about any one page. Keep it short. Never rely on it to make a page matter. |
| Hub to spoke | A hub page listing its cluster; a spoke's intro or sidebar pointing back | Every spoke links to its hub with the hub's name. Every hub links to every spoke with a descriptive anchor. This is the skeleton. |
| Contextual | A sentence in the body of a page | Links where the reader would want to go next. Descriptive anchor, one link per target per page, placed where the idea is introduced, not in a 'related posts' box. |
The common failure is a site with only the first role. Breadcrumbs and a mega menu reach every page, so nothing is technically orphaned, and yet no page has any link that says what it is for.
Draw the hub-and-spoke map
A hub is a page that owns a topic and introduces its parts. A spoke is one part, in depth. On a SaaS site the hub might be the integrations overview and the spokes the individual integration pages. On a services firm it is the service page and the articles that answer buyer questions about it. The map is the list of hubs, the spokes under each, and the links between them.
List the pages that make money
Service pages, product pages, pricing, the top category pages. These are your hubs, or they sit directly under one. If the list is longer than twelve, you have more than one business and need more than one map.
Assign every other indexable page to one hub
Export the URL list from your crawler. Give each page a hub. Pages that fit nowhere are candidates for consolidation or removal, and they should not get links until that decision is made.
Write the up-links
Every spoke links to its hub once, early, with the hub's actual name as the anchor. If the spoke's intro cannot mention the hub naturally, the page is probably in the wrong cluster.
Write the down-links
Every hub links to every spoke, with an anchor that describes the spoke, in a section the reader would actually use. A list of forty links is a sitemap, not a hub. Group them, or split the hub.
Add the cross-links inside the cluster
Spokes link to sibling spokes when the reader would move between them. Across clusters, link only through hubs. That keeps each cluster's signal tight and makes the map legible to a crawler.
Anchor text that helps and anchor text that hurts
- Describe the destination, not the action. 'Technical SEO audit checklist' tells a crawler and a reader what is there. 'Click here' and 'read more' tell them nothing. 'This article' is slightly worse than nothing, because it repeats across the site.
- Vary within reason. Pointing at one page with the same five words from two hundred pages reads as a template, and templates carry less weight. Use the page's name and two or three natural variations.
- Match the page's own heading. If the anchor says one thing and the destination's H1 says another, you have described the page twice in two ways. Pick one.
- One link per target per page. The first link to a destination is the one that counts for most crawlers. A second link to the same place adds noise, not signal.
- Never link a keyword to a page that is not about it. It is the fastest way to teach a search engine that two of your pages compete for the same query.
The four-check audit
Run this quarterly, or after any large release. It needs a crawler that reports inlinks per page and click depth, and an hour.
| Check | How to find it | What to do |
|---|---|---|
| Orphans and near-orphans | Pages with zero inlinks, or whose only inlinks are navigation and sitemap. | Assign a hub and add the up-link and down-link. If no hub fits, decide whether the page should exist. |
| Depth over three | Pages that need more than three clicks from the home page. | Link them from their hub, or from a page at depth one or two. Deep pages crawl late and rank weakly. |
| Links concentrated on pages that do not sell | Sort pages by inlinks. If the top of the list is tag archives, author pages and a privacy policy, your signal is pointing at the wrong things. | Reduce template links to those pages. Add contextual links to the money pages and their hubs. |
| Links only in JavaScript | Crawl with rendering off and compare inlinks to a rendered crawl. | Any link that disappears without JavaScript needs a real anchor tag in the server HTML. Menus and tabs built from scripts are the usual culprits. |
Keeping it alive after launch
Linking systems die in the second month, when the person who drew the map is on to the next thing and writers go back to linking whatever is open. Three small rules keep it running.
- Every new page gets a hub before it gets a draft. The brief names the hub. The up-link is in the first draft. The down-link is added to the hub on the day of publication, by the same person.
- Every new page gets two contextual links from older pages. Search your own site for the topic, open the two best matches, add one sentence each. Twenty minutes. This is how new pages are found and ranked in weeks rather than months.
- Every release gets a link crawl. Redirected internal links, lost links from a template change and new orphans all show up in a ten-minute crawl. Put it in the deployment checklist next to the noindex check.
The bottom line
- Give every link a role and a rule. Navigation reaches pages; hubs and spokes give them shape; contextual links tell engines and readers what matters.
- Draw the map first, keep it in a spreadsheet, and make the up-link and down-link a condition of publishing.
- Audit orphans, depth, concentration and JavaScript-only links quarterly, and crawl for redirected internal links after every release.
Frequently asked questions
- How many internal links should a page have?
- As many as a reader would use and no more. On a long guide that is usually five to fifteen contextual links plus the hub link. The number that matters more is inbound: a page that matters should have links from its hub and from several related pages, not from one.
- Do footer and sidebar links count?
- They count for discovery, so nothing linked from them is an orphan. They count very little for meaning, because the same links appear on every page. A page whose only inlinks are template links is treated as unimportant.
- Should I use a plugin that adds links automatically?
- Usually not. Automatic linking matches keywords, not intent, and produces the two things engines discount: identical anchors everywhere and links to pages that are not about the anchor. A map and a publishing rule do the job better, and you can see what was done.
- Does this help with AI answer engines?
- Yes. Answer engines need to decide which of your pages is the authoritative one for a question. A clear hub-and-spoke structure with consistent anchors makes that decision easy and makes the cited page the one you intended. Scattered linking produces scattered citations.
How this guide was put together
- The method is the one we apply on client sites. The example figures are from one anonymised engagement in 2026 and are given as an illustration, not a benchmark.
Sources
Founder of Lucidens. A decade of organic-search practice across English-, French- and Arabic-speaking markets, with a particular interest in measurement and in how answer engines choose their sources.
Related capabilities
- Content & Information Architecture
What you say, how it is organised, and how each part supports the whole.
- Technical SEO
Crawlability, rendering, performance and indexation, engineered so nothing you publish is lost.
Related research
- AI Crawlers: What to Allow, What to Block
Every AI provider now runs at least two bots, and they do different jobs. One feeds training. One feeds the answers your buyers read. Block the wrong one and you disappear from ChatGPT search or Claude without noticing. Here is each bot, what it does, and a robots.txt that says what you mean.
- Core Web Vitals: What to Fix First
Three metrics, three thresholds, and a long list of things people fix that never move them. This guide gives you the order that works: which number to read, which template to start with, the usual cause behind each failing metric, and what to stop spending time on.
- How to Run a Prompt Panel for AI Citations
There is no Search Console for ChatGPT, Perplexity or AI Overviews. A prompt panel is the closest thing to one: a fixed set of buyer questions, asked the same way every month, with every cited source written down. Here is how we build and run ours.
Want the map drawn for your site?
We crawl the site, draw the hub-and-spoke map with your team, write the first round of links and leave you the publishing rule that keeps it alive.