Skip to content
pageinspection

How importance moves through your own links

Links pass something, and where you point them decides which of your pages accumulates it. That is the useful part of this idea. Almost everything else written about it is arithmetic from a paper published in 1998 and a set of tactics that stopped working in 2009.

The short answer

Authority moves through your own links, so the pages you link to most from the pages that receive most are the ones your site is promoting, whether you intended it or not. On most sites the distribution is accidental: the privacy policy outranks the flagship guide because it is in the footer of every page.

ImportantAudit check · Internal authority

Importance moves along your own links

The mechanism, stated without the mysticism: a page linked from many important pages is treated as more important, recursively, until the numbers settle. Your home page usually has the most because it is linked from every page on the site and from most external references, and what it links to inherits a share of that.

Internal link value leaving a home page along three branches of different weights: a thick one to the guides section, a medium one to the shop, and a thin one to the blog. Each hub passes what it received on to the pages beneath it. One product page receives a substantial share and links onward to nothing, so the value collects there and stops.Where your own links send importancehome/guides/shop/blog/guides/grits/guides/steel/shop/kn-2000collects, links nowhere/blog/2019/06Line weight is the share travelling down each branch. The blog gets athin one because the home page links to it once, in a footer.
Three branches out of a home page, carrying different amounts because they are linked with different prominence and frequency. The product page at the end of the thick branch receives a real share and links onward to nothing, so the flow arrives and stops. Whether that is a problem depends entirely on whether that page is a destination or a stepping stone.

Two things follow that are worth more than any of the tactics. First, the pages you link to from your home page and your main navigation are the pages you have nominated, whether or not you meant to: a footer link to your terms of service on every URL is a strong nomination. Second, a page that receives a lot and links to nothing relevant is a terminus, and if it is not a page you want people to end on, the fix is a link out rather than more links in.

Why sculpting stopped working

For a period, the accepted technique was to add nofollow to internal links you considered unimportant, on the theory that the value they would have carried was redistributed among the remaining links. This was called link sculpting, it worked, and Google changed the handling in 2009 so that a nofollowed link still consumes its share and simply discards it.

Which inverts the outcome completely. Sculpting now destroys value instead of redirecting it: a page with ten links, four of them nofollowed, passes six links worth of value and evaporates the rest. The technique is not merely obsolete, it is harmful, and it survives in old documentation and in plugin settings.

The same reasoning applies to the family of tactics built on the same premise. Hiding links from crawlers with script, splitting content across pages to increase the number of internal links, and counting links per page against a fixed limit are all arithmetic on a model that does not describe how this works.

What replaced the tactics is a decision, not a technique: which pages should be prominent.

If a page matters, link to it from places that matter, in prose, with anchor text describing it. If it does not matter, stop linking to it from everywhere, which is a different action from marking those links nofollow. Removing a link from a global footer is real; annotating it is not.

The one legitimate use of nofollow on an internal link is close to nonexistent, which is why the safe rule is not to use it internally at all. For genuinely uncrawlable areas, the tool is a directive or a robots.txt rule, chosen deliberately.

A settled distribution needs the whole graph

Authority is computed by iterating over every link on the site until the values converge. It is a property of the graph, not a number a page carries, so there is no single-page reading of it, even approximately.

This one needs the full crawl

Authority is computed by iterating over the whole graph until the numbers settle, so it is a property of the site rather than a value any page carries.

The instant search on this site audits a single page, so rather than show you a verdict it cannot support, this guide sends you to the place the check actually runs.

The raw counts that feed it, inbound and outbound per URL, are reported separately, and they are the numbers to reach for when you want to act rather than understand.

The three moves that actually change the distribution

Skipping the tactics, here is what genuinely redistributes internal importance.

  • Link from your strongest pages, in prose. Find the handful of pages that receive the most external links and add contextual links from them to the pages you want to promote. This is the highest-leverage internal linking work there is, and it takes an hour.
  • Reduce the size of your global links. A footer with two hundred links dilutes every one of them and nominates nothing. Cutting it to the twenty that matter is a real change in how your site distributes attention.
  • Give the termini somewhere to go. A page that receives a lot and links nowhere relevant is a dead end for both a reader and the distribution. Two or three contextual links out of it is the whole fix.

What all three have in common is that they would be worth doing for a human reader with no search engine involved. That is the test to apply to anything you read about this subject: a technique that only makes sense as arithmetic is a technique that stopped working before most of the people recommending it started.

Why a distribution you cannot see is rated Important

It is Important rather than a Refinement because the consequences are concrete even though the quantity is invisible. Pages that receive nothing from your own site compete with nothing behind them, and on most sites the pages in that position include something that was expensive to produce.

It stays below Critical because nothing is excluded. A poorly supported page is crawled, indexed and served, and it can rank on the strength of what it says. The neighbouring checks that do remove pages, an orphan or a dead-end route, are rated higher and are usually the first thing to fix.

The rating carries a warning as well as a priority, and this guide is where the warning goes. This is the check most likely to be acted on with obsolete tactics, and the obsolete tactics are actively harmful rather than merely useless. A team that reads Important and reaches for nofollow has made their site worse in the name of a finding.

Approximating it with a weighted count

A real computation needs the whole graph and several iterations. A useful approximation is to count inbound links and weight them by whether the linking page is itself well linked, which you can do in two passes.

weighted-inbound.sh
# pass one: the edge list
curl -s https://example.com/sitemap.xml \
  | grep -o '<loc>[^<]*</loc>' | sed 's/<[^>]*>//g' | sort -u > /tmp/p.txt

while read -r page; do
  from=$(printf '%s' "$page" | sed 's|https://example.com||')
  curl -s "$page" | grep -o 'href="/[^"#?]*"' | sed 's/href="//;s/"$//' \
    | sort -u | while read -r to; do printf '%s\t%s\n' "$from" "$to"; done
done < /tmp/p.txt > /tmp/e.tsv

# pass two: score each page by its inbound count, then again weighting
# each inbound link by the source page's own inbound count
awk -F'\t' '{ inbound[$2]++ } END { for (p in inbound) print inbound[p], p }' \
  /tmp/e.tsv | sort -rn > /tmp/round1.txt

awk -F'\t' 'NR==FNR { score[$2] = $1; next }
     { weighted[$2] += (score[$1] > 0 ? score[$1] : 1) }
     END { for (p in weighted) printf "%8.1f  %s\n", weighted[p], p }' \
  <(awk '{ print $1 "\t" $2 }' /tmp/round1.txt) /tmp/e.tsv \
  | sort -rn | head -20

Two rounds is enough to see the shape. Compare the two orderings: pages that rise between round one and round two are supported by pages that are themselves supported, and pages that fall are linked mostly from places nothing points at.

Then look at the bottom rather than the top. The pages with the lowest weighted score are where the three moves in this guide have the most to give, and the list is usually short enough to work through in a morning.

Questions this check raises

Does PageRank sculpting still work?
No. Adding nofollow to internal links to redirect authority stopped working in 2009, when the value on a nofollowed link began being discarded rather than redistributed. The technique is still recommended in old material and it now costs you the link without giving anything to the others.
How do I move internal authority to a specific page?
Link to it from pages that already have authority, in the body rather than in a template, and reduce the number of low-value links competing on those same pages. The three moves that actually change the distribution are adding editorial links from strong pages, building a hub, and removing links nobody needs.
Why does my privacy policy have the most internal links?
Because it is in the footer, and the footer is on every page. Site-wide links accumulate enormous counts without expressing any editorial judgement, which is exactly why they are discounted. It is worth knowing about because it distorts every raw internal link count on the site.