Internal linking – how internal links work for SEO and AI search

Definition

Internal linking is when pages on the same website link to each other. Internal links guide the reader onwards and help search engines and AI systems find the pages and understand how they fit together.

What types of internal links are there?

Internal links are often divided into two kinds, according to where they sit and what they signal.

Structural links are in the menu, the footer and the breadcrumbs. They are the same on every page and show how the website is structured. This is where the information architecture becomes visible to both visitors and search engines.

Contextual links sit in the body text, in the middle of a sentence about the topic. The words around the link explain why two pages belong together. These are the links you control page by page, and they are what most of the rest of this text is about.

Both are needed. The menu makes sure every main page can be reached. The links in the body text tie topics together across the structure, for example a guide with the concepts it explains.

Why do internal links matter for SEO and AI search?

Google uses links as a relevance signal and to find new pages to crawl, according to its document Link best practices for Google. The document also gives a concrete minimum: every page you care about should have a link from at least one other page on the website.

The same document recommends thinking about which other pages could help the reader understand a given page, and linking to them in context. In practice, that is a description of contextual links.

A sitemap is therefore not enough on its own. It says that an address exists, but not where the page belongs or which topics it is connected to.

In topic clusters, links are what tie the subpages to their pillar page. Clear links also show which page owns which topic. That helps prevent cannibalisation, where two of your own pages compete for the same search. We explain how to avoid it in How to write SEO content that ranks.

The way links – internal links and backlinks alike – pass link equity between pages is often called link juice.

The same foundation applies to AI search. Google writes in AI features and your website that the same basic SEO advice applies to AI Overviews and AI Mode as to ordinary search. Indexing is also a requirement: to be shown as a source there, a page must be indexed and eligible to be shown with a snippet. A page that no link points to is harder to find, for Googlebot and other crawlers alike.

How many internal links should a page have?

There is no right number. According to Google's document on links, there is no magic ideal number, and if it looks like too many to you, it probably is. The question is easier to answer if you split it in two: how many links a page should have out, and how many pages should link in to it.

Links out from a page. Add a link when the text mentions something another page explains better and the reader may want to know more. Linking to the same page once, where the topic is first mentioned, is usually enough. A long guide will therefore naturally have more links than a short definition. If you cannot justify a link by saying it helps the reader on their way, it should not be there.

Links in to a page. Every page you want to be visible needs at least one link from another page on the website. Beyond that, your most important pages should have more pages linking to them than the rest. The links show both the reader and Google which pages you are putting forward.

A good benchmark is the median for your own website. When we counted on seosdesign.se in October 2026, a typical term in the glossary had 13 other pages linking to it from the body text, excluding the menu and footer. The spread was wide: SEO had 159 pages linking to it, while Seed keyword had none. That is down to how automatic term links work. They are added where the word appears, so a term mentioned in almost every text gets many links, and one that is rarely mentioned gets few, regardless of how important the page is.

So count how many pages link to your most important pages and compare with the median. If a page that matters to the business is well below it, add links to it by hand from pages where the topic is already mentioned.

What is good anchor text for internal links?

The anchor text, or link text, is the clickable text in a link. It tells both the reader and Google what the page behind the link is about. Google's advice on anchor text:

• Make it descriptive, reasonably concise and relevant to both the page the link is on and the page it leads to.
• Avoid generic anchor text such as "click here" and "read more".
• If the link is an image, the image's alt text works as the anchor text.
• Think about the words around the link, and do not place several links right next to each other.

Two ways to link to the same section:

• "Want to know more about pages without links? Click here." The link is on "click here", which says nothing about the destination.
• "Here is how to find pages that lack internal links." The anchor text describes where you will end up.

If you link to the same page several times, the anchor text does not have to be identical. Let it fit the sentence. Using the same exact keyword as anchor text on every page does not help the reader.

When does automatic internal linking work?

A common piece of advice is to avoid automatic internal linking altogether. That advice holds for widgets such as "latest posts" and "popular articles". They add links outside the text, to whatever is newest or most read, so the same few pages get links from every page and the rest of the content gets almost none.

But automatic linking can meet the same standards as a link you add by hand. On seosdesign.se, the terms in the glossary are linked automatically from blog posts, service pages and other terms, according to fixed rules:

• As a rule, a term is only linked the first time it appears on the page.
• The spelling has to match exactly. Inflected forms are not linked, so "landing pages" does not become a link to the term Landing page.
• The longest term wins: "Technical SEO" is linked as a whole, not just the "SEO" inside it.
• A page never links to itself.
• An existing link always takes precedence, and code is not linked.
• Draft terms are not linked. A link to a page that does not exist is worse than no link.

Every automatic link therefore has the term as its anchor text and sits in a sentence where the term is mentioned. What the automation cannot do is prioritise: it simply follows the wording of the texts. Service pages and other sales pages therefore still need links added by hand.

A pitfall for Swedish text. If the linking uses regular expressions in JavaScript, the word boundary \b only counts a–z, A–Z, digits and underscores as word characters. A word that begins or ends with the Swedish letters å, ä or ö therefore gets no word boundary at that letter. A Swedish term such as "ämnesauktoritet" (topical authority) would then not match in ordinary running text. We are aware of the limitation from our own term linking, and none of the 131 terms in our Swedish glossary currently begins or ends with å, ä or ö. If you use a plugin for automatic linking on a website in a language with letters outside a–z, it is worth testing with such a word before you rely on it.

Does it matter if the links are created with JavaScript?

Yes. According to Google's advice on crawlable links, Google can generally only follow a link that is an <a> element with an href attribute. Google cannot reliably extract links that only work through a script, such as a click on a <span>.

Googlebot runs JavaScript when the page is rendered. A link added by a script can therefore be found, but only after rendering, which can happen some time after the crawl itself. Google describes this in Understand the JavaScript SEO basics.

Not all crawlers render. In a log analysis that Vercel published in December 2024, none of the crawlers from OpenAI, Anthropic or Perplexity ran JavaScript. Anything that was not in the HTML the server sent, they did not see. The advice to put the links in the served HTML holds however that changes.

An example: when seosdesign.se ran on Webflow, a script in the browser added the term links as the page loaded. Since the move to Astro at the end of September 2026, they have been added automatically when the page is built rather than when it is displayed. The links are therefore already in the HTML the server sends, so even crawlers that do not run JavaScript see them.

To check your own website, open the page source (Ctrl+U), not Inspect, and search for the link. If it only shows up in Inspect, it is created by JavaScript.

How do you find orphan pages and pages with too few links?

An orphan page is a page that no other page on the website links to. It can be in the sitemap and technically flawless and still lack context. It is a classic crawlability problem. Here is how to find them:

1. Compare a crawl with the sitemap. Have a tool such as Screaming Frog crawl the website from the home page, following links only. URLs that are in the sitemap but that the tool never reached are orphan pages.
2. Count inbound links per page. The same crawl shows how many pages link to each page. Ideally, count only links from the body text; otherwise every page in the menu looks well linked. Sort in ascending order and compare with the website's median.
3. Read the Links report in Google Search Console (GSC). Under internal links, it lists the pages Google has seen the most internal links to. Export the list and sort it in ascending order. An important page at the bottom is worth a closer look. The table holds at most 1,000 rows, so on large websites the least-linked pages may be missing completely.
4. Inspect individual pages. In the URL Inspection tool, "Referring page" shows a page Google may have used to find the address, and "Last crawl" shows when Googlebot was last there. If only a sitemap is listed as the source, it may mean Google has not found any links to the page.

Once you have found the pages, link to them from relevant pages that are already indexed, in a sentence where the topic is actually mentioned. If a page is outdated, a 301 redirect to a better page may be the right solution instead.

Also check where the links point. An internal link should go directly to the address you want to rank, not via a redirect or to a copy of the page. We describe a common example from Shopify stores in Shopify's duplicate product URLs.

What should you do when you publish a new page?

Link to it straight away from existing pages that are already indexed and cover the same topic. That way, indexing does not depend on the sitemap alone, and the page has context from day one.

• Look for pages that already mention the topic, for example with site:yourdomain.com "topic" in Google, and add the link where the word already appears.
• Link from the new page to the pages that go deeper into what it covers.
• If the page is part of one of your topic clusters, link to it from the cluster's pillar page as well.