Internal linking that moves pages

Why internal links still do real work

Internal linking sounds like 2015 SEO. Does it still matter now that people ask AI?

It matters more than most of the newer tactics, for a boring reason: everything else depends on your pages being found, understood and indexed, and internal links are how that happens on your own site. They are the only links you fully control. They cost nothing. And they are one of the few SEO changes that has been tested with proper control groups.

Google's guidance is short and practical. It can reliably crawl only links that are real <a> elements with an href. Anchor text should be descriptive, concise and relevant, not "click here". Every page you care about should be linked from at least one other page (Google, SEO link best practices, last updated 10 December 2025).

On the AI side we are careful. We know of no study showing that internal links directly raise AI citations. What we do know is that Google's AI features only draw on pages that are indexed and eligible for a snippet (Google, AI features and your website). A page nobody links to is slower to be found, slower to be indexed and easy for Google to treat as unimportant. So internal links are a precondition for AI visibility, not a shortcut to it.

What the tests actually show

Most SEO advice is opinion. Internal linking is one of the rare areas with split tests, where one group of pages gets a change and a matched group does not. SearchPilot runs these for large sites. They are a vendor, the tests are on their clients' sites, and they do not always publish confidence intervals, so we read them as good directional evidence, not law.

Two lessons we take from this. First, links that help a buyer move to a closely related page, like a nearby location or the next step in a decision, tend to show up in traffic. Second, the classic cleanup jobs that audit tools love, like replacing every link that passes through a redirect, are worth doing when convenient but rarely move anything on their own. We spend effort on the first kind.

How our link finder works

Our engine runs a link finder for each client once a month. It is a code step: plain text matching, no model needed. It looks for one specific, high value pattern: a page that already talks about a topic but does not link to the page that should rank for it.

1. Start from the phrases that matter. Each client has a market map: the buying phrases we are trying to win. That list, not a crawl of every word on the site, is the input.
2. Pick one target page per phrase. The page Search Console already shows for that query. If none ranks yet, the page whose address clearly matches the phrase.
3. Find pages that already mention it. Read every page in the live sitemap. Keep the ones whose text contains the phrase but that do not already link to the target.
4. Score the pairs. Prefer source pages that get crawled and visited, and mentions in the body over mentions in passing.
5. Suggest two to five links per target. Written as rows in the client backlog, with the exact sentence and the anchor. A writer applies them; the finder never edits a site by itself.
6. Optionally, a cheap model reviews the shortlist. A small, fast model can read each pair and drop the ones that make no sense in context. It may only remove pairs, never add one, and stays in shadow mode until its judgments match ours.

Two to five is our house limit, so that a person reviews every link by hand. Two to five good ones, from pages that genuinely discuss the topic, is what a careful editor would do anyway.

Anchors: descriptive, varied, honest

Should every link use our exact keyword as the anchor?

No. Use the phrase where it reads naturally, and vary it where it does not. The anchor is a promise to the reader about what is on the other side. Google uses it to understand the target page, and a person uses it to decide whether to click. Both are served by a short, specific description.

For image links, Google reads the image's alt text as the anchor. So a linked logo or card should carry a real alt text, not an empty one.

How many links, and where on the page

How many internal links per page is too many?

There is no magic number, and anyone who quotes one is guessing. The useful question is whether each link earns its place for the reader. A long guide can carry twenty links comfortably if each one points somewhere the reader might want to go next. A short service page with forty links in the body reads like a directory and hides the one action you want.

Where the link sits matters more than the count. A link in the opening paragraph, in a sentence about exactly that topic, is the strongest signal you can give, both to a skimming reader and to a crawler working out what the page is about. Links in long footers and sidebars repeat on every page, so they tell Google less about any one page.

One more practical rule: when you add a new page, add links to it in the same release. A page published without links waits for Google to stumble on it through the sitemap. A page published with links from three crawled pages is usually found much sooner.

Where links should come from and go to

Not every link is equal. A link from a page Google visits often, and that people actually read, carries more weight than one buried on a page nobody opens. That gives us a few structural rules we apply to every site.

1. Every guide links to money. Each informational page links to the relevant service, product or contact page once near the top and once in the body. A guide that ranks but sends nobody anywhere is a cost centre.
2. Money pages link back to proof. Service pages link to the guides and cases that answer the buyer's doubts. That is where people go before they enquire.
3. No orphans. Every page in the sitemap has at least one link from another page. Pages that exist only in the sitemap tend to be crawled and indexed late, or not at all.
4. Hubs for clusters. A short hub page that links to every page in a topic, like the track page for this guide series, gives Google one clear map of the cluster.
5. Real HTML links. Links built with click handlers instead of <a href> may not be crawled at all. Check menus and cards especially.

Our new guides are live and in the sitemap, but Google has not indexed them. What do we do?

We have seen this often: pages that are published, sitemapped and fine, sitting unindexed for weeks on a site with little authority. In our experience the strongest lever is links from pages Google already crawls regularly, usually the homepage, the main service pages and the most visited guides. Request indexing for the most important few, but links are what make it stick. Our guide to getting indexed covers the rest.

A worked example

An illustrative case, using the office coffee rental business from our fan-out guide. Its market map has the phrase "office coffee machine rental". Search Console shows the rental service page for that query at an average position of 11.

Two links get added, both in sentences that already discuss renting. The blog post is skipped even though it is about coffee machines, because a link there would feel forced. That restraint is the point: two links a reader would actually click are worth more than ten that exist for a crawler.

The decision is logged with a metric, "clicks for the target page", a baseline from the last 28 days, and a check date 4 weeks out. We compare against untouched pages in the same language, because the whole site rises and falls with the season.

What is hype, and what is a waste of time

Mistakes we see, and a checklist

Mistakes

• Menus and cards built with JavaScript click handlers instead of real links.
• Ten links to one page in one week, all with the same anchor.
• Guides that rank and link nowhere commercial.
• Linking to the redirecting old address of a page after a migration, then forgetting to update the menu.
• Measuring a link change in a week that also had a Google update, then crediting the link.

Checklist

1. List your money phrases. One target page each.
2. Find mentions without links. Search your own pages for the phrase, or run a site search for it.
3. Add two to five. In sentences that already discuss the topic, with descriptive anchors.
4. Fix orphans. Every sitemap page linked from somewhere a crawler visits.
5. Check the markup. Real <a href> links, image links with alt text.
6. Measure. Clicks for the target page after 4 weeks, against pages you did not touch.

This is step eight of the Get Found track. Before it: mine your long-tail searches. After it: measure AI visibility without fooling yourself.

Sources and dates

Every external fact in this guide was read on 20 October 2026. Each source shows its publisher and the date it was published or last updated, and we flag what changed in 2026. Studies are their authors' measurements, usually correlations, not promises of results. Our own numbers come from our own Search Console and tools and say so where they appear.

• AI Features and Your WebsiteGoogle Search Central · Last updated 10 December 2025 · Read 20 October 2026.
• SEO link best practices for GoogleGoogle Search Central · Last updated 10 December 2025 · Read 20 October 2026.
• Spam policies for Google web searchGoogle Search Central · Last updated 28 August 2026 · Read 20 October 2026. Changed in 2026: Site reputation policy section updated 2026-08-28 for the EEA; 2026-04-13 back button hijacking added to malicious practices; 2026-05-15 clarification that spam policies apply to generative AI responses.
• Growth Engine scoring, indexing and probe rulesDLVX (Growth Engine operating notes) · Our own measurement, data to 19 October 2026 · Read 20 October 2026.
• Do 301 Internal Links Harm SEO?SearchPilot · Published 4 May 2022 · Read 20 October 2026.
• Internal linking case study: can links to nearby location pages help?SearchPilot · Published 27 January 2021 · Read 20 October 2026.
• The Importance and Impact of Internal Linking for SEOSearchPilot · Published 30 May 2024 · Read 20 October 2026.