ai · anthropic
0 mean radius0 max radius1 articles
Semantic Site Map
Free, no signup. Explore how 1 published articles are distributed around this site's semantic center. Find central pages, coherent topic neighborhoods, and editorial outliers in this static, build-time report without treating distance as a quality score.
Current build-time report · eligible published corpus
Report generated Aug 28, 2026.
This is a precomputed public report — filters and sorting run in your browser, and nothing you enter is sent or stored. 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.
1included articles
0most central radius
0furthest radius
1semantic topics
Site-radius distribution
Each dot is an article. Select one to open its article-level AI + SEO Data view.
more central · 0further out · 0
The horizontal position is relative to this report's observed minimum and maximum; it is not a universal 0–100 grade.
Topic coherence
Lower mean distance means the articles in that model-created cluster sit closer together.
Article evidence
Search, filter, and sort the complete included set.
Showing 1 articles
No articles match this search and topic filter. Clear the search or choose All topics.
| Article | Site radius | Topic | Topic radius | Closest articles |
|---|---|---|---|---|
| The Modern-Day Michelin Guide: Anthropic's Massive Programmatic SEO Opportunity#1 of 1 by centrality | 0 | Single site cluster | — | The Modern-Day Michelin Guide: Anthropic's Massive Programmatic SEO Opportunity |
How to use it
- Start with the distribution to see the site's overall semantic spread.
- Compare topic mean and maximum radii to find tight and broad editorial neighborhoods.
- Filter the evidence table by topic, then inspect the most central and furthest pages.
- Open an article to see its nearest semantic neighbors and article-level context.
- Compare semantic structure with the Internal Link Cluster Visualizer; semantic similarity and actual internal links answer different questions.
What the results mean
- Site radius is cosine distance from the centroid of every included article embedding, scaled for readability.
- Centrality rank orders the included articles from the smallest site radius to the largest.
- Topic radius is an article's distance from its assigned topic centroid.
- Topic mean/max summarizes how tightly the topic's member embeddings group together.
- Closest articles are the three highest cosine-similarity neighbors in the included corpus.
Features
- Build-time analysis of every eligible published article.
- Site-wide distribution, topic coherence cards, and complete evidence table.
- Search, topic filtering, sorting, and CSV export.
- Direct handoff to article evidence and the link-graph visualizer.
Limitations
This release analyzes PatrickStox.com only; it does not crawl or accept uploads. Results depend on the model, corpus, text cleanup, tags, and cluster count. Embedding similarity can miss factual, navigational, commercial, and business distinctions, and it cannot tell whether two pages should link.
Frequently asked questions
Can I run Semantic Site Map on my own website here?
Not in this release. This public explorer exposes the build-time analysis for PatrickStox.com. It is a transparent case study, not a general-purpose crawler or upload analyzer.
Does a larger radius mean an article is bad or off-topic?
No. It means the article is farther from the site-wide semantic centroid. A page can be far from the global center while fitting tightly inside a coherent topic cluster.
How is Semantic Site Map calculated?
Published, human-reviewed article text is embedded with all-MiniLM-L6-v2Embeddings are dense numerical vectors — lists of floating-point numbers — that represent the meaning of text in a high-dimensional space. Semantically similar content lands close together, so search and AI systems can match by meaning, not just keywords.. Cosine distance from the normalized site centroid becomes the site radius. Deterministic spherical k-means supplies topic centroids and within-topic distance.
Why are some articles excluded?
Preview, draft, example, and AI-labeled content is excluded. The report covers only published editorial articles that count toward site-wide aggregates.
Is Semantic Site Map a Google ranking signal?
No. It is a site-planning diagnostic built from one embedding model. Google does not provide or endorse this metric.
Site passport Local context for this saved site
Local data
Saved targets, named lists, and recent check summaries remain only in this browser.
Rate this tool
Feature requests for Siteradius
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.
You won't be emailed about that request anymore.
Loading…
➕ Request a feature
New requests are reviewed before they appear here.
Where this tool helps
Common use cases
See the editorial portfolio's semantic spread
Use the relative distribution to distinguish pages near the corpus centroid from pages farther from the current site-wide center.
Compare topic cohesion before expanding a cluster
Review each model-created topic's mean and maximum radius to find tight neighborhoods and broader clusters that deserve editorial review.
Investigate corpus-relative outliers
Treat transparent site-radius and topic-radius review cues as prompts for investigation, not automatic deletion, consolidation, or quality judgments.
Find nearby articles for an editorial handoff
Use the complete evidence table and closest-article links to inspect related coverage while keeping semantic similarity distinct from observed internal links.
Create a reviewable planning export
Search, filter, sort, and download the currently visible rows so an editor can preserve the analyzed corpus and decide the next supported action.
Watch the full workflow
Semantic Site Map walkthrough
Read the transcript
Semantic Site Map
This beginner walkthrough explains the basic terms first, then uses a clearly labeled sample report to explore topic groups, find unusual pages, search and filter the evidence, and download a reproducible review list.
Step 1
Semantic Site Map is a visual way to explore a library of articles. It estimates which pages are close to the overall subject center, groups similar pages into topics, and names nearby articles. That can help an editor understand the shape of a site before deciding what deserves a closer look.
Step 2
Here are four terms. An embedding is a list of numbers that represents the meaning of some text. A centroid is the calculated center of all those article representations. Cosine distance measures how different one article is from that center. Site radius is that distance, scaled to make the report easier to read. It is not a Google metric or an S-E-O score.
Step 3
Use this explorer to understand the spread of a content library, compare broad and tightly focused topics, find pages that differ from most of the site, discover semantically nearby articles for manual review, and export a focused planning list. Each result is a clue to investigate, not an instruction to delete, merge, or rewrite anything.
Step 4
This badge says the walkthrough uses a deterministic demonstration corpus with synthetic metrics. That means everyone sees the same safe example. The public release analyzes eligible, published PatrickStox dot com articles at build time. It does not crawl your website and there is no upload box in this version.
Step 5
The example includes twelve articles across three semantic topics. The most central site radius is twelve, and the furthest is sixty-two. Those endpoints describe only this included collection. Add, remove, or substantially edit articles and the center can move, so save the report date and scope with any notes you make.
Step 6
The review panel flags Translation versus Localization as unusually far from the site center. It separately flags JavaScript S-E-O and Translation versus Localization inside their assigned topics, and notes that the three topic centers are fairly distinct. These thresholds are calculated from this corpus. A flag means inspect the page and its context; it does not mean the page is bad.
Step 7
Each dot represents an included article. Dots farther left are closer to the site-wide center, while dots farther right are more different from that center. A page can sit far right because it covers a valuable specialty. The line shows relative semantic spread, not quality, rankings, traffic, expertise, or business value.
Step 8
Topic radius compares an article with the center of its own topic. The mean describes the group on average, and the maximum shows its widest member. Here, Technical S-E-O has mean seven point three and max seventeen. Content Strategy is seven point eight and sixteen. International S-E-O is broader at mean ten and max twenty-four.
Step 9
Select Show included articles to open a topic card. Now you can see the four Technical S-E-O pages and each page’s site radius. The clustering model creates membership; a common article tag supplies the readable topic name. Treat that label as a navigation aid, because an editor may describe the same group differently.
Step 10
The evidence table has three simple controls. Search finds a title or topic. Topic limits the table to one group. Sort can show the most central pages, the furthest pages, titles alphabetically, or the tightest topic fit. The line below the controls reports how many rows are currently visible.
Step 11
Choose Furthest first. Translation versus Localization moves to the top with site radius sixty-two and centrality rank twelve of twelve. Its topic radius is twenty-four. This is a convenient starting point for review, but the numbers do not tell us whether the page should change or whether it serves a useful specialized intent.
Step 12
Choose International S-E-O from the Topic menu. The count changes to four of twelve, and only that group remains. Comparing a page with its own neighborhood can be more useful than comparing it with the entire site, especially when the website intentionally covers several different subject areas.
Step 13
Search for translation and one row remains. The row keeps the site radius, the assigned topic, the topic radius, and three closest articles together. If a search returns nothing, the page explains how to recover by clearing the search or choosing All topics, so filtering never silently looks like missing source data.
Step 14
The Hreflang row lists International S-E-O Audit, International S-E-O Checklist, and Translation versus Localization as its closest semantic neighbors in this sample. That means their text representations are similar inside this corpus. It does not prove that links exist, that a link should be added, or that the pages satisfy the same search intent.
Step 15
With International S-E-O still selected, choose Download C-S-V. The tool exports exactly the four visible articles plus a header row, not the hidden eight. The file contains article, site radius, topic, topic radius, and closest articles. Keep the report date, fixture or corpus scope, and original page U-R-L with the download.
Step 16
The report embeds eligible article text with all Mini L-M L six V two. It normalizes those vectors, calculates cosine distance from the site centroid for site radius, and uses deterministic spherical K means for topic membership and topic centroids. The same input produces the same grouping, but another model, corpus, cleanup method, tag set, or cluster count can produce different results.
Step 17
Start with the distribution, compare topic cards, and filter the evidence table. Then open an interesting article and read it. Check its intended audience, search intent, performance, and nearby pages. Finally, compare this semantic view with the Internal Link Cluster Visualizer, because text similarity and the links that actually connect pages answer different questions.
Step 18
The main features are build-time analysis of eligible published articles, a site-wide distribution, topic-coherence cards, corpus-relative review cues, a complete evidence table, search, topic filtering, four sort orders, visible-row C-S-V export, direct article links, and a handoff to the observed internal-link graph.
Step 19
This release is a PatrickStox dot com case study, not a general crawler or upload analyzer. Preview, draft, example, and A-I-labeled content is excluded from the live public report. Embeddings can miss factual, navigational, commercial, and business differences. The map cannot measure quality, predict rankings or traffic, prove a link exists, or decide whether pages should link, merge, move, or be removed.
Use semantic distance to ask a question—not to issue a verdict.
Begin with one interesting page, read it in context, compare its semantic neighbors with the links that actually exist, and decide whether any change would help a reader. The public explorer is free, requires no signup, and makes its important boundaries visible.
Common issues & how to fix them
- Needs review Eligible corpus is too small for comparison Fix: Publish or approve enough eligible human-reviewed articles for a meaningful comparison, then rebuild the report.
- Warning Article is far from the corpus topic center Fix: Review whether the article intentionally expands the site’s scope; connect it to the relevant topic or revise its focus if not.
- Warning Article is distant from its nearest content cluster Fix: Review the article against its assigned cluster and clarify its topic or cluster placement when the distance is unintended.
- Information Corpus is split across weakly connected topic groups Fix: Create deliberate editorial bridges between genuinely related topic groups instead of forcing unrelated clusters together.
- Warning Corpus has low topical coherence Fix: Review the broadest clusters and split or refocus pages only where the model’s evidence matches editorial intent.