Hypothetical
Hub page internal links: a worked example, invented
Every figure on this page is made up for illustration. The method is real: find the inlet, count the outlets, move the links, wait for a refresh.
The site below does not exist. Its numbers were chosen to be easy to follow, and they are not benchmarks or results from any customer.
The setup
Imagine a documentation site. It has one hub page, called Guides. It has 31 referring domains, because a conference talk linked to it in 2025. Our inlet method is how you find such a hub on a real site.
The hub lists 120 articles in a single alphabetical block. Every article links back to the hub. Nothing links sideways.
That is a hub that receives and does not distribute. Equity arrives at one door and goes into 120 rooms in equal slices, regardless of which room matters.
Before and after, invented
| Measure | Before | After |
|---|---|---|
| Referring domains to the hub | 31 | 31 (unchanged) |
| Outgoing links from the hub | 120, one block | 18 in six labelled groups |
| Articles linked from the hub directly | 120 | 18, plus group pages |
| Clicks from hub to the 5 priority articles | 1 | 1 |
| Clicks from hub to the long tail | 1 | 2 |
| Sideways links between related articles | 0 | 3 per article |
Why fewer links can mean more flow
Look at the table again. The hub lost 102 outlets, yet the five priority articles stayed one click away.
What changed is the slicing. Each of the 18 remaining links now carries a larger share of whatever the hub receives, and the group pages pass part of that on to the long tail.
No measurement settles this exactly, because the weights inside Google's systems are not public. We would treat it as a hypothesis, and test it on one cluster before touching the others.
A caution on scale. Moving from 120 links to 18 is an extreme case chosen for the example, and a hub with 40 well-grouped links may need no change at all. Look at the depth myths before you flatten anything.
Also keep the audience in mind. A reader of the Guides page in 2025 still has to find things, and a short list of groups is easier to scan than a wall of 120 titles. Page experience and equity point the same way here, which is rare enough to enjoy. Ahrefs and Semrush would both flag a 120-link page in a site audit, and Google's Search Console would show how few of those articles get impressions.
The procedure behind the example
- 1
Find the hub in the inlet list
It should show up in
toppageswith more referring domains than most of its neighbours. - 2
Count what it passes on
List its outgoing internal links from your crawl export and group them by purpose.
- 3
Promote the few
Keep direct links for the pages you most want to rank. Move the rest under group pages.
- 4
Add sideways links
Related articles should point at each other, so equity can circulate without returning to the hub.
- 5
Re-measure after one refresh
The index updates roughly monthly, so give it a full cycle before drawing conclusions.
What we would do
Start with the hub that has the most domains
Change one cluster, keep a copy of the old link list, and compare snapshots. A single clean test teaches more than a site-wide rewrite.
- Do not delete links you cannot explain.
- Keep titles and URLs stable while you test.
- Write the date of each change in a log.