Internal linking strategy: hubs, anchors, orphans and depth

An internal linking strategy decides which of your pages link to which, and with what anchor text, so search engines can find every page, understand how your topics relate, and send the most authority to the pages that matter most. This guide covers the hub and spoke model, anchor text, orphan pages, link depth, and how to score link relevance with embeddings, using a coffee brand’s cold brew content as the running example.

What is an internal linking strategy?

Most sites have internal links. Fewer have a strategy. Without one, links accumulate by accident: the newest posts get linked from the homepage, old posts that still drive revenue drift deeper, and some pages end up with no internal links at all. A strategy replaces that drift with a few rules you apply to every new page and check every quarter.

Why internal links matter for SEO

Internal links do three jobs for search engines and one for readers.

• Discovery. Search engines find most pages by following links. Google’s documentation recommends that every page you care about has a link from at least one other page on your site. A page that is only in your sitemap is discoverable, but it carries little weight.
• Context. Anchor text and the surrounding sentence tell search engines what the target page is about. A link reading “how much coffee to use for cold brew” is a strong hint about the target.
• Priority. Pages with more relevant internal links pointing to them are treated as more important within your site. You decide where that weight goes.
• Reader paths. Good links send readers to the next useful page, which keeps them on your site and moves them toward a product page or signup.

Build hub and spoke clusters

The hub and spoke model groups related pages around one broad hub page that links to every narrower spoke page, while each spoke links back to the hub. It is the most reliable structure for showing search engines that you cover a topic in depth, which is the core of topical authority.

For the coffee brand, the cold brew topic looks like this:

• Hub: The complete guide to cold brew coffee
• Spokes: cold brew ratio, cold brew vs iced coffee, how long cold brew lasts, cold brew concentrate, best cold brew makers, cold brew grind size

Rules that keep the structure clean:

1. The hub links to every spoke, in context where possible, with a short summary of each.
2. Every spoke links back to the hub, early in the article, with descriptive anchor text such as “our complete guide to cold brew”.
3. Spokes link to siblings only when the reader needs it. The ratio article links to the grind size article because grind changes the right ratio. It does not need to link to the cold brew makers roundup.
4. Hubs link to other hubs sparingly, where topics genuinely connect, such as a cold brew hub linking to an iced coffee hub.
5. Commercial pages sit inside the cluster. The cold brew kit product page should receive links from the hub and the most relevant spokes, not only from the shop menu.

Your spokes should come from real keyword clusters, so each one targets a distinct group of queries. If two spokes target keywords that share most of the same search results, they belong on one page. Our guide to keyword clustering explains how to check that with SERP overlap.

How should you write anchor text for internal links?

Anchor text is the clickable text of a link. Google’s link best practices ask for descriptive text that makes sense out of context, and generic anchors like “click here” or “read more” waste the signal.

Three rules cover most cases:

• Descriptive beats exact match. Exact-match anchors are acceptable internally, but a varied set of natural descriptions reads better and covers more ways people search.
• One anchor, one target. If “cold brew recipe” links to the recipe page in one post and to the hub in another, you are telling search engines two pages own the same phrase. That is a common cause of keyword cannibalization.
• The sentence matters too. The words around a link add context. Link from a sentence that is actually about the target page.

How do you find and fix orphan pages?

An orphan page has no internal links pointing to it. Search engines may still find it through your sitemap, but it receives no internal authority and readers cannot reach it by browsing.

Here is how to find them:

1. Build a full URL list from your XML sitemap, your CMS export, the Pages report in Search Console and landing pages in your analytics.
2. Crawl the site from the homepage with a crawler that follows links only, not the sitemap.
3. Compare the two lists. URLs in the full list that the crawl did not reach are orphans.
4. Check for near orphans too: pages linked only from paginated archives, tag pages or the footer of one old post.

Then decide what to do with each one:

• Valuable and current: link it from its hub and one or two related spokes.
• Valuable but outdated: update it, then link it.
• Duplicate of a better page: redirect it to that page.
• No value: remove it or noindex it, and drop it from the sitemap.

Link depth: keep important pages close to the homepage

Link depth, or click depth, is the minimum number of clicks needed to reach a page from the homepage. Pages buried deep tend to be crawled less often and receive less internal authority, so your most important pages should be reachable in a few clicks.

There is no official maximum, but you can control depth in practical ways:

• Link hubs from the main navigation or the homepage. Each spoke is then one click from a hub.
• Use breadcrumbs so every page links back up its path.
• Add related links inside article bodies, chosen for relevance rather than recency.
• Avoid deep pagination as the only route to old posts. Page 14 of a blog archive is a long way from the homepage.
• Audit with a crawler that reports depth, and look for revenue or high-potential pages sitting deeper than your hubs.

Scoring link relevance with embeddings

Embeddings let you measure how related two pages are by meaning, which makes it practical to find good link candidates across hundreds or thousands of pages. An embedding model turns a piece of text into a list of numbers (a vector) so that texts with similar meaning end up close together. The closeness of two vectors, usually measured with cosine similarity, becomes a relevance score.

The concept works like this:

1. Embed every page. Use the title, headings and main content, without navigation and boilerplate that would make every page look alike.
2. Score pairs. For each page, compute similarity to every other page and keep the top candidates.
3. Filter. Drop pairs that already link to each other, and pairs below a threshold you choose by reviewing a sample. Thresholds depend on the model and your content, so tune rather than copy one.
4. Find the placement. Embed individual paragraphs of the source page and pick the paragraph most similar to the target. That is where the link belongs, and it often suggests the anchor text.
5. Prioritize. Rank candidates by relevance combined with need: orphans, deep pages and commercial pages come first.

# pages: list of objects with .url and .text; embed() returns a vector
vectors = {p.url: embed(p.text) for p in pages}

candidates = []
for source in pages:
    for target in pages:
        if source.url == target.url or links_to(source, target):
            continue
        score = cosine(vectors[source.url], vectors[target.url])
        if score >= THRESHOLD:
            candidates.append((source.url, target.url, score))

Know the limits. High similarity does not always mean a useful link: two near-identical pages score very high, which is often a sign of cannibalization to fix rather than a link to add. And similarity cannot judge whether a reader at that point in the article actually wants to go somewhere else. Use scores to shortlist and keep the final decision with an editor.

While you write, the SupaDraft editor shows suggested internal links as dotted underlines in your draft. Press Tab to accept a suggestion or Esc to dismiss it, so link review happens in the same pass as editing.

An internal linking audit you can run every quarter

Run this checklist every quarter, and every time you publish a new cluster:

1. List all hubs and confirm each one links to all of its spokes.
2. Confirm every spoke links back to its hub.
3. Crawl the site and fix any orphan pages.
4. Check click depth for revenue pages and new articles.
5. Export anchor text and look for the same anchor pointing at different URLs.
6. Fix broken internal links and links that pass through redirects.
7. Add links from older, well-linked pages to the articles you published this quarter.

The last step is the one most teams skip. New articles start with few links pointing to them. Plan inbound links at brief stage; our content brief template has a field for the existing pages that should link to each new article.

Frequently asked questions

How many internal links should a page have?

There is no correct number. Link wherever it helps the reader take the next step, make sure every link is relevant, and avoid long blocks of links that no one will read. A hub naturally has more links than a spoke.

Should internal links use exact-match anchor text?

They can, in moderation. Exact-match anchors are fine internally, but mixing in natural, descriptive variations reads better and avoids the over-optimized pattern of identical anchors everywhere.

Do navigation and footer links count as internal links?

Yes, search engines follow them. But they appear on every page, so they say little about how specific pages relate. Contextual links inside the body carry more topical meaning.

Should I use nofollow on internal links?

Rarely. Nofollow exists for links you do not vouch for, such as paid or user-generated links, and your own pages do not fit that description. If you do not want a page indexed, use noindex; if you do not want readers to reach it, remove the link.

Plan your links before you publish

Pick one topic, map its hub and spokes, and run the quarterly audit on that cluster first. Then draft your next article in the SupaDraft editor and accept relevant internal link suggestions with Tab as you write.