A small business pays for twenty blog posts and a handful of backlinks, then opens Search Console six months later. The articles have impressions. The service page, the one that actually brings in work, is sitting at position 11.
Often nothing is wrong with that page. It just never receives the authority the site already owns. The posts link to each other, to category archives and to an author page. The link that should say “emergency boiler repair” never gets written, because nobody had that job.
Internal links are the one ranking input you control completely, and they cost nothing per link. Backlinks bring authority in from outside; internal links decide where it goes after it arrives. Most sites leave that second decision to a related-posts plugin. Fixing it is usually stage three of our custom SEO services for small businesses, and you can do a good part of it yourself with this guide.
You may have seen a claim that 80% of small businesses starve their service pages this way. We could not find a measurement behind any figure like that, so we will not repeat one. The pattern is easy to test on your own site, and the 60-minute audit below shows how.
A note on evidence. Google does not publish how it scores internal links. Where we rely on a patent, a Google statement or a study, it is linked. Where a number is our own working estimate, we say so.
The mechanics of internal PageRank and the Reasonable Surfer patent
PageRank started as a random surfer model. In Brin and Page’s 1998 paper, an imaginary visitor clicks links at random and occasionally gets bored and jumps somewhere else. The damping factor, usually set to 0.85, is the chance they keep clicking. In practice it means authority fades with every step away from a strong page, and a page’s score is split among the links it carries.
Equal splitting was the weak part of that model, since nobody clicks a privacy link as readily as a link in the sentence they are reading. US patent 7,716,225, filed in 2004 by Jeffrey Dean, Corin Anderson and Alexis Battle and granted in 2010, replaces the equal split with a weight based on how likely a person is to follow each link. This is the Reasonable Surfer. Its feature list covers anchor font size, position (running text, above or below the first screenful, side, footer, sidebar), the number and meaning of anchor words, whether the anchor is commercial, the words just before and after the link, and how closely the anchor’s topic matches the page. It names “Terms of Service” links and banner ads as examples of links people are unlikely to follow.
Two cautions that most write-ups skip. A patent proves Google owned an idea; it does not prove the system runs that way today. And the patent contains no published weights. Its only numbers come from a worked example. Anyone quoting “a footer link passes X%” from this document made that number up, and the chart below is our estimate, labelled as one.
Google’s public comments point the same way, though. John Mueller said in 2019: “We mostly differentiate between the primary content & boilerplate (things like header, sidebar, menu, etc)”. That is the practical reason a contextual link in paragraph two beats a footer link repeated across 500 pages. Those 500 links are one template decision, copied. Zyppy’s study of 23 million internal links across 1,800 sites saw traffic start to fall at around 45 to 50 links per URL, and its author blames mainly sitewide navigation links that point at low-traffic pages. It is correlation, but it fits.
The Reasonable Surfer weighting model: where internal links likely pass the most ranking power
Estimated relative value of one internal link, by where it sits on the page.
Method: these are our editorial estimates on a relative scale, built from the link features listed in patent US7716225 and from Google’s statement that it separates primary content from boilerplate. They are not measurements. The patent publishes no weights and Google has confirmed none, so read the order of the bars and ignore the precision of the numbers.
On DOM position, be careful with the top-of-page advice. The patent lists position as a feature, yet Mueller said in 2021 “I don’t think we keep track of order of links on a page” and called link order an implementation detail. So write the important link early because readers meet it sooner and any click-probability model would rate it higher. Rearranging source order with CSS to win something is wasted effort.
Click depth, crawl budget and the orphan page trap
Two URLs can look completely different and behave the same. example.com/blog/category/post has three folders and can sit one click from the homepage. example.com/post has none and can sit five clicks deep. Mueller put it plainly in 2018: “we don’t count slashes in the URLs”; what counts is “how many links you have to click through to actually get to that content.” Directory depth is a filing system. Click depth is a distance.
The three-click limit is a convention and not a Google rule. It survives because the arithmetic is unkind: damping alone removes about 28% by the second hop (0.85 squared is 0.72), and that is before each page splits what is left among its other links. A service page reached through Home, Services, Category, Page gets the remains.
Crawl budget gets blamed for far more than it causes. Google’s own guide is written for sites with a million or more pages changing weekly, sites with 10,000 or more changing daily, and sites with many URLs stuck as “Discovered – currently not indexed”. A 40-page plumbing site has no crawl budget problem. What it may have is a discovery and priority problem. Google says more popular URLs tend to be crawled more often. The guide does not mention internal links, so this next step is our inference: a link from a page Googlebot already visits often is the fastest route for a new URL to be found.
Then there are orphan pages. Google’s link guidance says “Every page you care about should have a link from at least one other page on your site.” A sitemap entry helps discovery, but it passes no internal PageRank. WordPress sites make orphans in predictable ways: landing pages built for ads, service pages dropped from the menu during a redesign, and old posts pushed so far down a paginated archive that they are technically linked and practically lost.
One more discovery issue. The same guidance says Google can generally crawl a link only if it is an <a> element with an href. Links inserted by JavaScript are crawlable if they use that markup, but only once the page has been rendered. HTML link discovery is the safer route, so check mega-menus, tabs and “load more” buttons in the URL Inspection tool’s rendered HTML. A <div onclick> is not a link to Google, however much it looks like one. Our technical WordPress SEO fixes cover the wider rendering side.
The reverse silo and hub-and-spoke blueprint
A traditional silo pushes authority down from category to article. A reverse silo runs the other way. Supporting articles can attract links and long-tail impressions that a sales page never will, so they hand that authority up to the page that earns money. Hub-and-spoke, or the topic cluster, is the same shape drawn as a wheel with the service page at the centre.
- Pick one money page per commercial intent and write down its single primary query. No other page may target that query.
- Plan four cluster posts around what a buyer asks before they search that term: cost, a decision, a diagnosis and timing. Each post owns its own informational query.
- Apply the cannibalization test. If a post could satisfy someone searching “hire a boiler engineer near me”, it is aimed at the wrong intent. Rewrite it around the question.
- Inside the first third of each post, link to the money page with a descriptive anchor. One good link per post is enough.
- From the money page, link down only where a post answers an objection on that page. Two to four links will do. A sales page that sends every visitor to the blog is working against itself.
- Link the money page directly from the homepage, then add one or two sideways links between cluster posts.
The reverse silo blueprint: how 4 supporting cluster posts feed 1 commercial service page
Each arrow shows the direction of the link and the kind of anchor that belongs on it. The example is a heating company.
The direction matters more than the count: authority collected by the articles and the homepage ends up on the page that sells.
Cannibalization usually comes from intent overlap and from anchors that contradict each other. When four posts all link to the service page with its commercial term, and the service page refers to each post with a question, Google gets a consistent answer about which URL does what. The planning work behind those posts is covered in our guide to content marketing for small business SEO.
| Page type | Primary outbound internal target | Minimum incoming contextual links | Common mistake to avoid |
|---|---|---|---|
| Homepage | Core service pages, ahead of the blog | None needed; logo and navigation cover it | Giving the blog archive more prominence than the services |
| Core service page | Cluster posts that answer objections, then the quote page | 3 to 5 from topically related pages | Being reachable only through a navigation dropdown |
| Local landing page | Parent service page and the contact page | 2, including one from the parent service | City pages that link only to each other |
| Blog cluster post | Its money page, in the first third | 2, one from the hub and one from a sibling | Relying on a related-posts widget for every link |
Anchor text engineering and the surrounding text window
Advice about anchor text mostly comes from the backlink world, where a profile heavy with exact-match anchors looks bought. Google’s spam policies specifically flag “links with optimized anchor text” placed in articles and guest posts on other sites. Internal links are the opposite case. You are the editor of your own site, and describing where a link goes is exactly what Google asks for.
Matt Cutts said “internal website links will not cause you any sort of trouble” on a normal site, with thousands of identical anchors on one page as his exception. Asked about an over-optimization penalty in 2019, Gary Illyes answered “you can abuse your internal links as much as you want AFAIK”. In the Zyppy data, pages with at least one exact-match internal anchor had at least five times the traffic of pages with none, and more anchor variations went with more clicks. The study’s author cautions that exact-match pages may simply be better optimized overall, so read it as permission and not proof.
That variation is semantic anchor text variance: the same page reached through several honest descriptions of it. Each new wording tells Google about another way people describe the service, so six links with six accurate anchors give you more than six identical ones.
The anchor is also only part of the signal. Google says “the words before and after links matter, so pay attention to the sentence as a whole,” and the patent lists nearby words as a feature. Think of a surrounding text window of roughly one clause on either side. “Our engineers replace failed pumps and diverter valves on the first visit, so same-day boiler repair is realistic for most faults” gives far more co-occurrence than the same anchor dropped into “for more information see”. Avoid chaining links side by side, and remember that with an image link, Google uses the alt text as the anchor.
| Target page type | Exact-match ratio | Partial or semantic ratio | Branded or natural ratio | Example: good vs over-optimised anchor |
|---|---|---|---|---|
| Core service page | 20–30% | 50–60% | 15–20% | Good: “boiler repair call-outs”. Over: “best cheap boiler repair London” |
| Local landing page | 15–25% | 50–60% | 20–30% | Good: “our team in Leeds”. Over: “Leeds boiler repair Leeds engineers” |
| Blog cluster post | 5–15% | 65–75% | 15–25% | Good: “why boilers lose pressure”. Over: “boiler pressure boiler repair tips” |
| Homepage | 0–5% | 5–15% | 80–95% | Good: company name. Over: “affordable boiler repair company” sitewide |
| Contact or quote page | 0–5% | 20–30% | 65–80% | Good: “request a quote”. Over: “boiler repair quote cheap” |
Test your own site architecture
Set the four inputs below to match your most profitable service page. The score and bottlenecks update as you change them.
Interactive PageRank and internal link equity simulator
Plus a site audit calculator. Set the four inputs to match your main service page.
In the model site, this page holds 0.55 times the PageRank of an average page.
Top 2 architectural bottlenecks
- Too few relevant pages link to your service page. Add links from three to five related posts.
- Your service page sits three clicks from the homepage. Link it directly from the homepage or main navigation.
The score is a teaching model we built, weighted 40 for contextual links, 30 for click depth, 20 for anchors and 10 for redirects. It is not a Google metric and cannot predict rankings. The PageRank line runs the original formula with a 0.85 damping factor on an 18-page model site, so anchors and redirects do not change that figure.
Fixing technical link leaks: redirects, nofollows, canonicals and JS links
You will still read that every 301 evaporates 15% of a link’s equity. That number has a history but no source. Search Engine Journal notes that Matt Cutts confirmed a small loss years ago without ever giving a percentage. In 2016 Gary Illyes said “30x redirects don’t lose PageRank anymore”. If you are told a redirect is costing you 15%, ask for the source.
Internal redirects are still worth removing, for less dramatic reasons. Google’s crawl guide says long redirect chains have “a negative effect on crawling”, and its crawlers follow up to 10 hops by default. Each hop is a wasted request and a slower page for a visitor on a phone. Permanent redirects tell Google the destination should be canonical, so an internal link to the old URL argues with your own redirect. Updating the link takes a minute.
Link rot is the blunter version. Delete a page and every internal link to it now ends in a 404, carrying authority nowhere. Update those links first, then redirect the old URL to a true replacement if one exists. The same discipline applies to canonical chains: if a link points at a parameter or non-trailing-slash version whose canonical tag names another URL, link straight to the canonical one.
Never nofollow links to your own useful pages. PageRank sculpting tried to, on the theory that nofollowing some links would send more authority through the rest. Matt Cutts explained in 2009 why that stopped working: with ten points and ten links, five of them nofollowed, the five followed links used to receive two points each and now receive one. The other five points disappear. His advice was to “let PageRank flow freely within your site”, with narrow exceptions for pages such as carts and logins.
Finally, the First Link Priority rule and mega-menus. The theory says that if a page links twice to the same URL, only the first anchor counts. Evidence is mixed. Cutts said in 2009 that Google often uses the first anchor but not always. Mueller rejected the naive version in 2021. Zyppy’s controlled tests found Google recorded at most one text anchor and one image anchor per target, usually the first of each, but they did not test navigation directly. Mueller told one site owner worrying about this that they were “overthinking it”. The cheap hedge is to make navigation labels descriptive. “Boiler repair” in the menu is safe under every version of the theory, and “What we do” is not.
A 60-minute internal link audit using only Google Search Console
- Minutes 0 to 10. Open Performance, then Pages, for the last three months with average position switched on. Use the table filter to keep positions 8 to 15, then sort by impressions. Choose the three commercial pages with the most impressions.
- Minutes 10 to 20. Click each page and open the Queries tab. Confirm the top queries match what the page sells. If they do not, internal links will not fix it, so drop that page from the list.
- Minutes 20 to 30. Open Links, then Internal, then Top linked pages. Note how many links each target has, and click through to see which pages send them. The table caps at 1,000 rows and the report is a sample, so treat it as a guide.
- Minutes 30 to 45. Back in Performance, sort pages by clicks. These are your strongest donors. Open the Queries tab on the top ten and pick three whose queries sit close to each target topic.
- Minutes 45 to 55. Add one contextual link from each donor, in the first third of the copy, with anchors from the matrix above. Replace any link that passes through a redirect while you are in there.
- Minutes 55 to 60. Run URL Inspection on each edited donor and request indexing. Write down the date and the current position, then compare a 28-day window once the pages have been recrawled.
| Symptom in Google Search Console | Root architectural cause | Exact fix | Expected ranking impact in 30–60 days |
|---|---|---|---|
| Service page with high impressions at positions 8–15 | Few contextual links from related, well-performing pages | Add 3 to 5 links from top-clicked related posts | Medium to high if the page matches intent; the first place to look for movement |
| Blog post outranks the service page on a commercial query | Post overlaps commercial intent; service page under-linked | Refocus the post on the question and link up with a commercial anchor | Medium; ranking URLs may swap before settling |
| Several URLs alternate for the same query | No single page owns the query | Choose one owner, retarget the others, align anchors on the owner | Medium; volatility should ease first |
| “Discovered – currently not indexed” on a few new pages | URL known only from a sitemap or a deep archive | Link from two pages that are already crawled often | Indexing first; rankings follow only if the page is strong enough |
| “Page with redirect” for URLs still used in menus or copy | Old URLs left in internal links after a migration | Update links to the final 200 URL | Low on rankings; cleaner crawling and faster clicks |
| Top linked pages led by privacy, terms and author pages | Template links dominate; few body links to services | Trim footer links and add contextual links to services | Low to medium; depends on how under-linked the services were |
If the audit turns up thin authority rather than poor routing, the next job is earning backlinks. Internal links can only redistribute what the site has.
Practical questions about internal linking
Where internal linking fits in a full SEO campaign
Internal linking is the third stage of our roadmap: Google Business Profile, technical SEO, internal linking, topical authority, external links. The order is deliberate. Links are worth little on a site Google cannot crawl cleanly, and new content and backlinks do more once there is a route carrying their authority to the pages that sell. If you would rather check where your site stands first, our small business SEO checklist is a good place to start.
If you want someone to map this on your own site, we will review your click depth, redirects, orphan pages and anchors, then give you a prioritised plan and a price for your actual scope.
Sources
- US patent 7,716,225: ranking documents based on user behavior and/or feature data
- Brin and Page: The Anatomy of a Large-Scale Hypertextual Web Search Engine
- Google Search Central: link best practices
- Google Search Central: managing crawl budget for large sites
- Google Search Central: HTTP status codes and network errors
- Google Search Central: redirects and Google Search
- Google Search Central: spam policies
- Search Console Help: Links report
- Zyppy: study of 23 million internal links
- Zyppy: selective link priority tests
- Matt Cutts: PageRank sculpting
- Search Engine Land: no PageRank dilution with 30x redirects
Sources checked on 8 November 2026.
