Internal linking for SEO: internal links, site structure, and the orphan pages Google never finds

What makes an internal link crawlable

Google's requirement is one sentence long: it can only crawl your link if it is an a element with an href attribute. Everything in the table above follows from that. The href also has to resolve to something a crawler can send a request to, which rules out fragments that only mean something to your JavaScript router.

This matters more now than it did five years ago because of how modern front ends are built. A component library ships a clickable card, the card is a div with a click handler that calls the router, and to every human being on earth it is a link. To Googlebot it is a div. Sites lose entire sections this way: the product category pages are reachable from the homepage by clicking, and reachable to a crawler from nowhere.

The test is boring and reliable. Load the page, view the rendered source, and search for the URL you expect to find inside an href. If it is not there, no amount of internal linking strategy helps, because there is no link. Do this on your primary navigation, your pagination, your faceted filters and any infinite-scroll listing, which are the four places it goes wrong most often.

Pagination deserves its own check. If page two of a listing is only reachable by clicking a button that fetches more results with JavaScript, then everything past the first twenty items is effectively orphaned. Rendering real anchor elements with real hrefs for each page of a listing costs almost nothing and is often the single change that gets a large catalog crawled.

How many internal links should a page have?

There is no limit, and the number people still quote comes from a guideline Google retired more than a decade ago. The old advice to keep a page under 100 links existed for a mechanical reason that no longer applies: Google's indexing system used to truncate documents at roughly 100 kilobytes, so about one link per kilobyte was a sensible ceiling. When the size limit went, the guideline went with it. Matt Cutts of Google later said publicly that pages with two, three or four hundred links are entirely normal now, as long as the page is long and has real substance behind it.

So the practical answer is not a number, it is a ratio. Links should be proportional to content. A 400 word post with 40 internal links is a link farm; a 4,000 word guide with 40 internal links is a well-connected reference page. Google did keep one caveat, which is that a page stuffed with links purely to game the system can still be actioned as spam. That is a judgment about intent, not a count.

For a normal editorial page, three to ten in-content links is a comfortable working range, on top of whatever your navigation and footer already carry. What matters far more than the count is the distribution. Ten links spread across ten different pages that need help beats ten links all pointing at your homepage, which already has every link on the site pointing at it.

• No official limit exists. The 100-link guideline was dropped years ago
• The original 100 came from a 100KB indexing truncation limit, not from link theory
• Google reserves the right to act on pages that are obviously link-stuffed
• Judge links per page against content length, not against a fixed number
• Three to ten in-content links suits most editorial pages
• Where the links point matters more than how many there are

Orphan pages: the URLs nothing on your site points at

An orphan page is a page on your site that no other page on your site links to. It exists, it loads, it may even be in your sitemap, and there is no path to it by clicking. Orphan pages are almost never created deliberately. They are a side effect of ordinary site maintenance.

The usual causes are worth knowing because they tell you where to look. A category or tag archive gets removed and every post that was only reachable through it is stranded. A site is redesigned and the new navigation lists eight sections where the old one listed fourteen. Paginated listings are shortened, so posts past page three fall off the end. Landing pages are built for an ad campaign, the campaign ends, and nothing was ever supposed to link to them. Products go out of stock and get pulled from listings while their URLs stay live.

Finding them takes two lists and a subtraction. List every URL your site can produce, usually from your sitemap or your CMS. Then list every URL that is reachable by following links from the homepage, which is what a crawler sees. The difference is your orphan set. If you do not have a crawler handy, the cheaper proxy is to compare your sitemap against what is actually indexed, since a page that has been live for months and has never been indexed is a strong orphan candidate.

Then decide, page by page, which of three things is true. The page matters, in which case link it from somewhere relevant that already gets crawled. The page does not matter but should stay live, in which case leave it and stop worrying. Or the page should not exist, in which case redirect it to the closest live equivalent or let it return a 404, which Google treats as a strong signal not to crawl that URL again.

• Removed category, tag or archive pages strand everything under them
• Redesigns that shrink the navigation orphan whole sections
• Shortened pagination cuts off older posts
• Campaign landing pages that were never linked from the site
• Out-of-stock products pulled from listings while the URL stays live
• Pages reachable only through a search box, which crawlers do not use

Anchor text: the part most sites waste

Google asks for anchor text that is descriptive, reasonably concise, and relevant both to the page it sits on and to the page it points at. That is a low bar and most sites still fail it, because the two most common internal anchors on the web are read more and click here, which describe nothing.

Anchor text is doing real work. It is one of the few direct statements you make about what another page is about, and unlike a title tag you get to make it repeatedly, from different pages, in different contexts. A page linked forty times as read more has been described forty times as nothing.

The other failure is the opposite one. Linking every instance of a phrase to the same page with the same exact anchor reads as automated, and it usually is, because a plugin did it. Vary the phrasing naturally. If a page is about bulk index checking, linking it once as checking thousands of URLs at once and once as a bulk index checker is better than forty identical anchors, and it reads like a person wrote it because a person did.

One practical rule: write the sentence first, then link the words that were already there. Anchors chosen that way are almost always descriptive, because you wrote them to explain something rather than to carry a keyword.

Click depth and site structure: why level four never gets crawled

Click depth is the number of clicks from your homepage to a given page, following the shortest path. It is not a ranking factor Google publishes, but it correlates hard with what actually gets crawled, for a reason that is more arithmetic than algorithm. Your homepage is the most linked page on your site. Whatever it points at inherits some of that, and whatever those pages point at inherits some of what is left. By the fourth or fifth level there is very little arriving.

Crawl demand works against depth too. Google describes crawl demand as a function of perceived inventory, popularity and staleness. A page five levels down is rarely popular, and Google's picture of your inventory is built from the links it can see. Deep pages are visited less often, so their changes are noticed later, which makes them look staler, which makes them visited even less often.

The fix is structural rather than tactical. Flat is better than deep: most sites can reach every page in three clicks with sensible hub pages, and a site that needs six is usually organizing by internal convention rather than by what readers look for. Hub pages are what make flatness possible. A well-built category page that genuinely lists everything under it turns a five-level tree into a two-level one, and it is exactly the pattern Google uses in its own documentation when describing how it discovers new posts.

Pagination is the common leak. A blog listing that shows ten posts a page and runs to page fifteen puts your older posts fifteen clicks from the homepage even though the tree looks shallow on paper. Category hubs, a decent tag structure, and related-post links between articles all pull those pages back up.

• Aim to reach every page that matters within three clicks of the homepage
• Build hub pages that genuinely list everything beneath them
• Watch deep pagination, which quietly buries older content
• Link between related articles so posts are not reachable only through listings
• Put your most important pages one click from the homepage, not in a footer

An internal linking workflow that survives a real site

Most internal linking advice assumes you are starting from nothing. You are not. You have a site with several hundred pages, a navigation nobody wants to touch, and no appetite for a six week audit. This is the version that fits into an afternoon and repeats monthly.

Start from what is broken rather than from a map. Pull the list of URLs that are not indexed. For each one, ask a single question: what links to this? If the answer is nothing or only the sitemap, you have found an orphan and the fix is one link from a page that already gets crawled regularly. Do that before anything else, because those pages are currently worth zero.

Then work the other direction. Take your pages that already rank and get traffic, since those are the pages Google crawls most often, and check what they link to. This is where you place links to pages you want crawled sooner. A link from a page Google visits weekly is worth many links from pages it visits twice a year.

Make new links part of publishing rather than a separate job. When an article goes live, add two or three links to it from existing related articles the same day. This is the step people skip, and it is the reason a new post can sit in Discovered for six weeks. The link from an already-crawled page is what puts it in the queue.

Finally, re-check after anything structural. A navigation change, a category cleanup, a pagination change or a migration can orphan dozens of pages silently, and nothing on the site will look broken. Comparing what is indexed before and after is the cheapest way to catch it while it is still one deploy old.

• Step 1: list URLs that are not indexed and find which have nothing linking to them
• Step 2: add one link to each orphan from a page that already gets crawled often
• Step 3: audit what your best-crawled pages link to, and place new links there
• Step 4: link every new page from two or three existing pages on publication day
• Step 5: re-check coverage after any navigation, pagination or category change

Internal linking mistakes that quietly cost you indexing

Linking everything to the homepage. The homepage already receives more internal links than any other page on your site and needs none of them. Links pointed there are links not pointed at the pages that need help.

Treating the footer as an internal linking strategy. A block of links repeated on every page is worth very little per link, and an identical link block across a whole site is a pattern rather than an endorsement. Contextual links inside content are worth more, and a footer full of them will not rescue an orphan.

Linking to redirects instead of destinations. Internal links that point at a URL which then redirects still work, but they add a hop for every crawl and they get worse over time as chains accumulate. When you move a page, update the internal links that point at it rather than relying on the redirect forever.

Linking to pages that are noindexed or blocked in robots.txt. This is wasted crawling. Worse, a page blocked in robots.txt that carries a noindex tag will never have that tag read, because the crawl needed to read it is the crawl you blocked.

Automating anchor text with a plugin and never reading the output. Automated interlinking produces exactly identical anchors at scale, links inside sentences where they make no sense, and sometimes links a phrase to the page it is already on. Skim what it produced.

Assuming the sitemap covers it. A sitemap and an internal link do different jobs. Sites that lean on the sitemap alone tend to have large numbers of URLs sitting in Discovered, which is Google saying it knows the URL exists and has not judged it worth fetching.