Languages

Internal Link Cluster Visualizer

Free, no signup. Turn a crawl export, page inventory, or a bounded raw-HTML crawl into a navigable link map. See observed internal authority, root reachability, and explainable missing-link opportunities.

Built by Patrick Stox

Try:

Use the explicit “Export graph for Cluster Visualizer” download in Scout Site Audit Free. The JSON is read locally and is never embedded in a URL or stored in this browser.

Accepted columns and pipe format

Edge CSV: Source,Destination, source,target, or a Screaming Frog All Outlinks export. Optional Source Entities and Target Entities columns accept semicolon-separated names. Quoted values and extra columns are accepted. A headerless two-column CSV also works.

Page inventory: one row per page as URL | title | comma-separated terms | comma-separated linked URLs | optional entities. This format can represent a page with no links, so orphan checks are possible.

Runs entirely in your browser — nothing you paste is uploaded or stored. Supplied graph and entity columns are analyzed in this browser. This mode never crawls, calls an AI model or knowledge graph, or writes links. Anonymous run-level outcome counters may be used for aggregate research; URLs, domains, IPs, and identifiers are never included, and no statistic is released below 100 runs.

Feedback
Report a bug

Found something broken in Cluster Visualizer? Let us know what happened — this goes straight to a private triage queue, not a public list.

What will be sent 
 No tool inputs, uploads, pasted source, complete results, query parameters, or URL fragments are attached automatically. You can edit or remove the selected passage above. Browser and anti-abuse metadata is processed for spam prevention. 
  
 
  
 
 
  
 
 
   
  
     
 Local data 
Saved targets, named lists, and recent check summaries remain only in this browser.
 
   
 
      
 
How to use it
  1. Use supplied mode for a page inventory or edge export; use crawl mode for a bounded public raw-HTML graph.
  2. Choose the correct parser or let the tool detect it, then analyze the graph.
  3. Select the click-depth root and read coverage warnings before metrics.
  4. Use the map for patterns, the table for exact evidence, and the suggestion table for human link review.
  5. Export page evidence and suggestions when you need an auditable handoff.
 
Example result Built-in page-inventory example
The built-in seven-page inventory includes a homepage, four editorial pages, a tool, and an About page. The deterministic analysis reports 7 supplied pages, 7 unique internal links, root click depths, normalized PageRank, and evidence-based segments. The unlinked “Topic cluster strategy” row can be evaluated as an orphan because pipe format explicitly supplies the complete example inventory.
These counts describe the included fixture only; they are not a live site measurement.
 
What the results mean
  • PageRank — relative link equity normalized within this graph.
  • Click depth / unreachable — shortest observed path from the selected root, or no observed path.
  • Orphan — evaluated only when the supplied inventory can represent pages with no edges.
  • Entity evidence — a shared supplied name or bounded local title/heading heuristic, shown separately with provenance.
  • Missing-link score — transparent lexical, segment, modest entity-overlap, and graph-deficit factors, not a requirement.
 
How it works
The parser normalizes up to 250 supplied pages and 3,000 edges, or the bounded crawler collects up to 150 raw-HTML pages from 1,000 discovered URLs. Graph analysis deduplicates edges, calculates iterative normalized PageRank and shortest root paths, groups pages from URL/term evidence, and scores absent observed edges. Entity overlap contributes at most 10% of topical strength; no bulk AI or knowledge-graph request occurs.
 
Features
  • Pipe inventory, generic edge CSV, optional entity columns, and Screaming Frog outlink support.
  • Bounded exact-host root or sitemap crawl with robots and trap controls.
  • Interactive canvas plus accessible sortable evidence tables.
  • Normalized PageRank, click depth, reachability, segment counts, and scoped orphan state.
  • Explainable review-only missing-link suggestions and two CSV exports; no link writes.
  • Handoff to Semantic Site Map for a public example of semantic clusters, which are deliberately kept distinct from this tool's observed links and lexical/entity clues.
 
Limitations
Supplied graphs are only as complete as the input. Root crawls cannot discover isolated pages, sitemap crawls cover only the inventory supplied, and raw HTML omits JavaScript links. PageRank is internal to this graph; entity names can be ambiguous; and suggestions do not inspect paragraph context. “Existing-link absence” means absent from this analyzed graph only. Crawl caps, redirects, policies, or trap sampling can make all results partial.
 
 
Does this tool crawl my website?
Supplied-graph mode never makes a request. Crawl mode retrieves up to 150 public same-site raw-HTML pages through the bounded shared crawlerA crawler — also called a spider or bot — is an automated program that fetches web pages, extracts their links, and queues new URLs to visit. Search engines use crawlers to discover and download content for their index.; it does not render JavaScript or write to the audited site. Results remain in this tab unless you export them.
Can an outlinks export identify orphan pages?
Not by itself. A page with no links may be absent from an outlinks export entirely. Orphans are evaluated only for the pipe-format page inventory, and the selected crawl root is never called an orphan merely because nothing links to it.
Are the segments semantic topic clusters?
No. The segments use visible URL directories first and supplied or derived lexical terms only when a directory is unavailable. They are navigation evidence, not an embedding or search-engine interpretation.
What does PageRank mean in this report?
It is normalized only within the supplied or bounded observed graph. It helps compare nodes in that input and is not Google PageRankPageRank is Google's original recursive link-graph algorithm: a page's score depends on the scores of the pages linking to it, and in the published model each page's score is split across its outbound links (the simplified version: links are weighted votes). Google says it's evolved since launch but still part of its core ranking systems. or an external authority metric.
Should I add every suggested missing link?
No. Suggestions combine lexical/segment overlap with visible graph deficits. Add a link only when the source passage gives readers a natural reason to visit the target.
   
Next stepSEO Opportunity Finder — run the full end-to-end workflow.

People also explore
ToolLink AnalyzerToolScout Site Audit FreeToolSemantic Site MapArticleInternal LinksArticleOrphan PagesArticleWebsite Structure
 
 
 
Feature requests for Cluster Visualizer
 

Upvote what you want most. New ideas can be submitted from the floating Feedback menu;
      requests appear here once approved, and the most-wanted rise to the top.

   
Loading…
  
  ➕ Request a feature   
New requests are reviewed before they appear here.
 
     
  
  
 
 
 
 
   
       
 
Where this tool helps
Common use cases
Map an internal-link cluster
Visualize a bounded supplied link graph to understand how pages and topic groups connect.
Find orphaned or isolated pages
Spot URLs with no meaningful internal paths before they disappear from normal discovery flows.
Review internal authority flow
Use the PageRank view to find pages receiving too little or disproportionately high internal link equity.
Discover relevant missing links
Review lexical suggestions as editorial candidates, then validate each link in page context before adding it.
Watch the full workflow
Internal Link Cluster Visualizer walkthrough
Read the transcript

      
      

        
Internal Link Cluster Visualizer

        

        
This visualizer maps an internal-link graph, calculates normalized PageRank and click depth, groups pages from observable U-R-L and term evidence, and proposes review-only missing links. I’ll analyze the built-in fictional inventory, explain the map and tables, show alternate inputs and crawl scope, cover limitations, and outline a responsible link-review workflow.

      

      

        
Step 1

        

        
Use this tool during information-architecture reviews, migrations, topic-cluster work, or internal-link audits. It converts supplied or bounded observed links into a navigable map and exact evidence tables.

      

      

        
Step 2

        

        
Supplied graph mode is private, browser-only analysis of a page inventory, edge C-S-V, or Scout export. Crawl mode retrieves up to one hundred fifty public same-site raw-H-T-M-L pages with explicit coverage limits.

      

      

        
Step 3

        

        
Pipe inventory can represent pages with no links, so scoped orphan checks are possible. Edge C-S-V represents observed source-to-destination links. Optional entity columns remain supplied evidence, not knowledge-graph verification.

      

      

        
Step 4

        

        
The page opens with a seven-page fictional inventory, so selecting Analyze produces real local graph calculations without a network request. Your own supplied input supports up to two hundred fifty pages and three thousand observed links.

      

      

        
Step 5

        

        
Graph evidence reports included pages, unique links, reachability, segments, and suggestion counts. Coverage warnings explain parser limits, missing inventory, crawl caps, redirects, or policies that narrow every downstream claim.

      

      

        
Step 6

        

        
Node size reflects PageRank normalized only within this graph, and color reflects evidence-based segments. Select or drag nodes for exploration; use the table, not the picture alone, for exact values.

      

      

        
Step 7

        

        
Click depth is the shortest observed path from the selected root. Unreachable means no path was observed in this input. A root crawl cannot discover isolated pages and therefore cannot prove true orphan status.

      

      

        
Step 8

        

        
The page table provides title, segment, normalized PageRank, inlinks, outlinks, and click depth. Search, filter by segment, sort a metric, select a row for evidence, or export the bounded page C-S-V.

      

      

        
Step 9

        

        
Suggestions combine lexical and segment overlap with graph deficits and a modest entity-evidence contribution. Each row means only that the edge was absent from this graph; it is never a requirement to add the link.

      

      

        
Step 10

        

        
Crawl mode stays on the same host, honors matching robots Allow and Disallow rules, uses trap guards, and fetches raw H-T-M-L at a bounded rate. This walkthrough does not start a crawl or contact a site.

      

      

        
Step 11

        

        
Supplied inputs may be incomplete, root crawls miss isolated pages, sitemap inventories are not whole-site proof, and raw H-T-M-L omits JavaScript links. PageRank is input-relative, entity names can be ambiguous, and suggestions do not inspect paragraph context.

      

      

        
Step 12

        

        
Export page and suggestion evidence, open each proposed source page, and add a link only when the surrounding passage creates a natural reader need. Reanalyze after changes and compare coverage before claiming improvement.

      

      

        
Use graph evidence to ask better internal-link questions.

        

        
Start with a representative inventory or bounded crawl, read coverage before metrics, and investigate important, deep, unreachable, or underlinked pages. Review every proposed source-to-target pair in paragraph context, export the evidence for collaboration, and recrawl after changes. The graph describes observed links; it does not measure Google PageRank or semantic authority.

      

    
 
         
 
Common issues & how to fix them
 
     
  •  Warning Page has no internal links from the cluster Fix: Add at least one crawlable contextual link to the orphan from a relevant indexed cluster page, and link back where it helps navigation. 
  •  Warning Page is isolated from its topic cluster Fix: Connect the page to its nearest topic cluster with descriptive links from and to the cluster hub or closely related pages. 
  •  Warning Page has weak internal-link connectivity Fix: Add a small number of relevant contextual links from stronger cluster pages and replace low-information anchors with topic-specific text. 
  •  Warning Cluster depends heavily on one bridge page Fix: Add direct links between the subgroups currently dependent on the bridge so one page is not the sole path connecting the cluster. 
  •  Warning Topic graph is split into disconnected components Fix: Choose the intended hub and add relevant cross-component links so the disconnected topic groups form one navigable cluster. 
  •  Information Page cannot be assigned to a topic cluster Fix: Assign the page to a clear topic and connect it to that topic’s hub; if it serves no distinct intent, consolidate or remove it. 
    •