How RankForge calculates authority flow

Every internal link is a vote and a pipe. As a vote, it tells search engines a page is worth visiting; as a pipe, it carries a share of the linking page's authority to the target. Backlinks decide how much authority enters your domain, but your internal graph decides where it ends up — and on most sites the answer is 'pooled on the homepage and a handful of hubs, nowhere near the pages that need to rank'.

You can't fix what you can't see. A page-by-page audit shows you that /pricing has a weak position; it doesn't show you that /pricing receives one nofollowed footer link and nothing else, while three dead category pages soak up most of your homepage's authority. Modelling authority flow makes the invisible plumbing visible, which is the precondition for the structural fixes that move rankings.

Forget the deprecated toolbar number. PageRank is an algorithm, and the algorithm is alive and well as a model of how importance propagates through a link graph. The intuition: a page is important if important pages link to it, recursively. The mechanic that makes it computable is simple — each page divides its authority evenly among its outbound links and passes it along; you iterate until the numbers stop changing.

A has authority 1.0 and 2 outbound links:
   A(1.0) --> B        B receives 0.5 from A
   A(1.0) --> C        C receives 0.5 from A

B has authority 0.5 and 1 outbound link:
   B(0.5) --> C        C receives another 0.5 from B

After this pass C holds the most authority because two
pages point to it — one of which (B) was itself fed by A.
Iterate until values converge.

The damping factor, at a high level

Raw link-passing has a problem: authority could circulate forever in loops, or get permanently trapped in pages with no outbound links. PageRank solves this with a damping factor (classically ~0.85). The interpretation is a 'random surfer' who follows a link about 85% of the time and, the other 15%, teleports to a random page. Practically, damping means a small, constant baseline of authority reaches every page — and that authority doesn't propagate infinitely far, so a link's value decays with each hop it travels.

Textbook PageRank treats every edge identically. Real search engines don't, and neither does RankForge — because a link's placement on the page tells you how much it was 'meant'. RankForge classifies each internal link by where it sits in the rendered DOM and weights its contribution accordingly.

LINK TYPE     WHERE              VALUE   WHY
────────────  ─────────────────  ──────  ──────────────────────────
Contextual    in main content    HIGH    editorial, topical, chosen
                                          for THIS page
Navigation    global menu/header  LOW     same on every page; generic
Footer        global footer       LOW     boilerplate; often "junk
                                          drawer" of links
Sidebar       module/widget       MEDIUM  semi-curated; varies by
                                          template, sometimes topical

Contextual links

In-body editorial links are the high-signal edges of your graph. They carry topical context, descriptive anchor text, and the implicit endorsement of being chosen for that specific page. RankForge weights these most heavily — they are where you actually steer authority, and where its recommendations focus.

Navigation and footer links

Global navigation and footer links are structural plumbing: they appear on every page, so their per-link signal is low and search engines discount them. They still matter — a link in the main nav puts a page one click from everywhere, which is great for crawl depth — but you can't fix a buried money page by adding it to the footer and expect it to rank. RankForge counts them, flags when a page's only inbound links are boilerplate, and weights them down so they don't inflate the picture.

Sidebar links

Sidebar and widget links sit in between — 'related posts', 'popular in this category'. They're semi-curated and sometimes topical, so they get a middle weight. The trap is auto-generated sidebar modules that link the same 'recent posts' from every page: that's navigation in disguise, and RankForge treats site-wide repeated modules accordingly.

A model that only handles clean <a href> links on static HTML will badly misread a real site. Here is how RankForge handles the things that actually distort authority graphs in the wild.

Nofollow

A rel="nofollow" (or sponsored/ugc) link is a hint that the linking page doesn't fully endorse the target. RankForge treats nofollowed internal links as carrying little to no authority — so if a page's only inbound internal links are nofollowed, it's effectively orphaned from an authority standpoint even though a crawler can still reach it. Sites nofollow internal links far more often than they realise (login, cart, faceted filters), quietly starving real pages.

Canonicals

A rel="canonical" tag says 'treat this URL as that one'. Authority should consolidate onto the canonical target, not stay split across duplicates. RankForge follows canonicals so that links pointing at /product?color=red and /product both feed the single canonical /product — modelling the consolidation search engines perform rather than counting each variant as its own page. Mismatched or self-conflicting canonicals are flagged because they leak authority into pages that should never have been separate.

Redirects

Links pointing at a URL that 301-redirects pass their authority through to the destination, minus a small loss per hop. RankForge resolves redirect targets so authority lands on the live page — and flags redirect chains (A→B→C→D), where each extra hop bleeds equity and slows crawling. A surprising amount of internal authority is lost to internal links that were never updated after a URL changed.

JavaScript-rendered links

If a link only exists after client-side JavaScript runs, whether it counts depends on whether the crawler rendered the page. RankForge distinguishes links present in the server-rendered HTML from those that require execution, because relying on JS-injected internal links is a real risk — they're crawled less reliably and later. When a site is heavily client-rendered, the authority model says so rather than pretending the rendered graph is the whole graph.

Duplicate URLs

Trailing slashes, uppercase, tracking parameters, session IDs, and faceted filters spawn many URLs for one page. Naively, each looks like a distinct node and the page's authority shatters across them. RankForge normalizes URLs and consolidates duplicates (respecting canonicals) so a page is one node in the graph — otherwise the most-linked page on the site can look weak simply because its votes are scattered across thirty address variants.

Orphan pages

An orphan has no inbound internal links, so in the authority model it receives only the tiny damping baseline and nothing else. It's the clearest possible signal of wasted potential: the page exists, but the graph routes no equity to it. RankForge lists orphans explicitly and, where the page is worth keeping, recommends contextual links from relevant, already-authoritative pages.

Link depth

Because authority decays per hop, the click-depth of a page from strong entry points is inseparable from its authority. RankForge computes depth from the homepage and other high-authority hubs, and surfaces the distribution — a long tail of valuable pages at depth 4+ is both a crawl-frequency problem and an authority problem at once. Fixing depth is often the single highest-leverage authority move on a large site.

Putting it together, the pipeline is: crawl the site → normalize and consolidate URLs (canonicals, redirects, duplicates) → build the directed link graph with each edge tagged by type (contextual / nav / footer / sidebar) and rel attributes → run a weighted, damped PageRank over that graph → read out per-page authority, plus where it pools (hubs) and where it leaks (dead ends, nofollow traps, redirect loss).

  crawl ──> normalize URLs ──> build typed graph ──> weighted
            (canonical,          (edges tagged          PageRank
             redirect,            contextual/nav/        (damped)
             dedupe)              footer/sidebar,           |
                                  follow/nofollow)          v
                                                    per-page authority
                                                    + pools / leaks /
                                                      orphans
                                                          |
                                                          v
                                                  ranked link fixes

The point of modelling authority flow isn't the score — it's that the model tells you exactly which link to add. Because RankForge holds the full graph, it can identify pages that are authority-rich and topically relevant to a starved target, and recommend a specific contextual link: from this page, to that page, with this anchor text. That's a link that shortens depth, passes real (non-diluted, followed) equity, and reinforces a topical relationship — the three things that move the model and, with it, rankings.

It also tells you what to stop doing: prune the redirect chains, fix the canonicals leaking equity, and trim mega-menu bloat that's diluting every link. Read the companion pieces on building topic clusters and improving crawl depth for the tactical playbooks, or run the checker below to see your own authority map.